Remote Sensing of Environment
|
|
- Edward Haynes
- 5 years ago
- Views:
Transcription
1 RSE-07469; No of Pages 6 Remote Sensing of Environment xxx (2009) xxx xxx Contents lists available at ScienceDirect Remote Sensing of Environment journal homepage: Efficient radiative transfer model inversion for remote sensing applications John Hedley a,, Chris Roelfsema b, Stuart R. Phinn b a School of Biosciences, University of Exeter, EX4 4PS, UK b Center for Remote Sensing and Spatial Information Science, School of Geography, Planning and Environmental Management, University of Queensland, Brisbane, Australia article info abstract Article history: Received 13 May 2009 Received in revised form 16 July 2009 Accepted 18 July 2009 Available online xxxx Keywords: Radiative transfer Inversion Look-up table (LUT) Successive approximation A simple method for efficient inversion of arbitrary radiative transfer models for image analysis is presented. The method operates by representing the shape of the function that maps model parameters to spectral reflectance by an adaptive look-up tree (ALUT) that evenly distributes the discretization error of tabulated reflectances in spectral space. A post-processing step organizes the data into a binary space partitioning tree that facilitates an efficient inversion search algorithm. In an example shallow water remote sensing application, the method performs faster than an implementation of previously published methodology and has the same accuracy in bathymetric retrievals. The method has no user configuration parameters requiring expert knowledge and minimizes the number of forward model runs required, making it highly suitable for routine operational implementation of image analysis methods. For the research community, straightforward and robust inversion allows research to focus on improving the radiative transfer models themselves without the added complication of devising an inversion strategy Elsevier Inc. All rights reserved. 1. Introduction Inversion of physics-based radiative transfer models is an area of rapid development in remote sensing of both aquatic and terrestrial environments (Liang, 2007). In these approaches a parameterized forward model for reflectance such as Hydrolight (for aquatic applications) or PROSPECT, PROSAIL or DART (for terrestrial canopies) takes a series of parameters describing the optical properties of participating media or canopy structure and the model defines a mapping from parameter space to spectral space (Fig. 1a) (Darvishzadeh et al., 2008; Gastellu-Etchegorry et al., 2003; Mobley et al., 2005; Zhang et al., 2008). Image analysis then uses a search algorithm to find the parameter space location which minimizes the distance of the corresponding spectral space location from the image pixel reflectance, measured by a distance metric such as Euclidean distance. Two distinct approaches to the search algorithm can be identified: 1) pre-calculation of reflectance look-up tables (LUTs) by repeated runs of the forward model with systematically differing parameter values (Mobley et al., 2005), optionally with interpolation across tabulated points (Gastellu-Etchegorry et al., 2003; Liang et al., 2006) and 2) spectral matching by successive approximation using optimization techniques such as the Downhill Simplex or Levenberg Marquardt (L M) algorithm (Brando et al., 2009; Goodman & Ustin, 2007; Klonowski et al., 2007; Lee et al., 1999; Wolfe, 1978), which is feasible only if the forward model is sufficiently fast to permit many runs Corresponding author. address: j.d.hedley@exeter.ac.uk (J. Hedley). for each image pixel. With either approach, for a complex radiative transfer model, both the forward modeling and the inversion may take substantial computational effort. Routine or large-scale operational image analysis by physics-based methods therefore demands the development of efficient approaches for both modeling and inversion. Often the forward model is simplified or approximated to facilitate inversion. In contrast, we develop a generic inversion technique which allows the use of any forward model and hence model accuracy need not be compromised. A key problem in the efficient representation of the parameter to reflectance mapping function is that it is not generally clear a priori how to subdivide the parameter space in order to evenly populate spectral space from the forward model. For example, modeled shallow water reflectance typically has an inverse exponential relationship with depth (Mobley, 1994), so a regular subdivision of a depth parameter would over-sample the deep water spectral space region with similar reflectance spectra produced from unnecessary forward model runs. At the same time the shallow water spectral space region would be relatively under-sampled with a higher discretization error in the tabulated reflectances (Fig. 1b). The interaction between multiple parameters (e.g. depth vs. water clarity) makes the general problem of efficient and accurate LUT construction very hard to tackle by analytical means. To address this problem we present an automated construction procedure for adaptive look-up trees (ALUTs) that evenly distributes and minimizes the discretization error in spectral space and facilitates fast inversion (Fig. 1). The construction technique can be applied to any forward radiative transfer model that is parameterized by a series of real and integer values. The following sections describe 1) the requirements of a forward model to which the ALUT construction /$ see front matter 2009 Elsevier Inc. All rights reserved. doi: /j.rse
2 2 J. Hedley et al. / Remote Sensing of Environment xxx (2009) xxx xxx a set of set of m real-valued parameters, α 1, α 2 α m, and any number of discrete integer parameters, i 1, i 2, etc. As a first step, any discrete valued parameters are combined into a single integer parameter by i=i 1 +i 2 N 1 +i 3 N 1 N 2 + so that i=1, 2 N, where N 1, N 2 are the maximum value each integer parameter can take and N is the product of the maximum values. This gives a set N of mapping functions from the real-valued parameter space to remote sensing reflectance in spectral space, f i : R m YR n hα 1 ; α 2 N α m iir rs ð1þ Fig. 1. Adaptive LUT construction for a forward model of one continuous parameter, α, mapping to two spectral bands, hb 1 ; b 2 i. (a) The mapping function from parameter space to spectral space which the LUT seeks to accurately represent. (b) Discretization by regular steps in parameter space over-samples some regions of spectral space and under-samples others. (c) and (d) iterative adaptive construction fills in the gaps to approximate an even distribution of points in spectral space and efficiently represent the function. method can be applied 2) the adaptive construction algorithm, and 3) the methodology for building and searching the tree representation of the ALUT. The subsequent sections present an illustrative example of ALUT construction using a shallow water radiative transfer model (Lee et al., 1999) and its application to hyperspectral airborne imagery of a coral reef. 2. Methods 2.1. Mapping from parameter space to spectral space The adaptive LUT construction can be applied to any forward model for n-band remote sensing reflectance modeled dependent on where f i is the forward model for particular discrete value of i and each f i consists of an assumed continuous mapping from m- dimensional parameter space to n-dimensional spectral space, with R rs being the spectral remote sensing reflectance in n-vector form. The separation of discrete and continuous parameters is necessary because the adaptive subdivision can only be applied to the continuous parameters. In essence, a series of separate ALUTs are constructed, one for each value of i, this accommodates forward models that contain discrete parameters, for example where a model component is based on a choice of one from a set reflectance spectra. The only additional requirement is that a priori limits for the continuous parameters are defined, so that α MIN i V α i V α MAX i, this is the same as is required for any LUT methodology and is also often used in successive approximation techniques to constrain the mapping function domain Adaptive look-up-table construction ALUT construction begins with an initial regular subdivision of parameter space (as shown for 1-dimensional parameter space in Fig. 1b), then an iterative algorithm repeatedly assesses which regions in parameter space correspond to currently under-sampled regions of spectral space, the most effective parameter is subdivided in that region, the new points are added to the ALUT and the iteration continues (Fig. 1c,d). The key to implementing the subdivision algorithm for m- dimensional parameter space is that the ALUT structure is represented as a hierarchical voxelation of the parameter space, where voxels are m-dimensional parameter bounding boxes, the 2 m vertices of which are current points in the LUT (Fig. 2). To assess the local point distribution in spectral space, change in R rs across the voxel is assessed by taking the mean distance in spectral space when changing Fig. 2. Hierarchical voxelation of parameter space for a forward model of three continuous parameters hα 1 ; α 2 ; α 3 i. Black dots show the parameter space location of points in the ALUT where R rs has been calculated, open circles show where subdivision is occurring and points need to be calculated for the next iteration.
3 J. Hedley et al. / Remote Sensing of Environment xxx (2009) xxx xxx 3 a parameter from its lower bound on one side of a voxel to its upper bound on the other side. This occurs along 2 m 1 edges of the voxel, one edge for each high-low combination of the parameters other than the one being assessed. Using Euclidean distance as the distance metric, the mean change in R rs for parameter i is Di ðþ= 1 2m X 1 2 m 1 j =1 jr rs ði; j; 0Þ R rs ði; j; 1Þj ð2þ with R rs (i, j, 0 1) being the remote sensing reflectance corresponding to the vertices at the low and high parameter value end of the jth voxel edge which describes a change in parameter i. The mean change in R rs for each parameter, D(i), is calculated for each parameter in each voxel when the voxel is created. The values are stored in a list which is continually updated and sorted during the construction process. Voxels at the top of the list represent the parameter space region which currently has the greatest discretization error in spectral space. The LUT is grown by taking the voxel at the top of the list and bisecting it into two new voxels at the midpoint of the parameter value that represents the greatest mean spectral change across the voxel (Fig. 2). When a voxel is subdivided 2 m 1 new vertices are required, their parameter locations are the midpoints of the subdivided edges, and 2 m 1 forward model calculations are required to generate the corresponding R rs spectra. The mean spectral space distances for each parameter, D(i), are calculated for the new voxels (Eq. (2)) and added to the sorted list, and the construction procedure continues. To initialize the ALUT construction in m-dimensional space a regular grid of voxels is created with corners at the extremes of the parameter limits, typically this grid can just be a single voxel (Fig. 2a). When the forward model contains a discrete parameter, i.e. NN1, the ALUT construction algorithm is repeated independently N times, with the forward mapping function being f i, i=1,2, N. Essentially N separate ALUTs are constructed, these are merged into a single treestructured LUT as described in the next section Search optimization by binary space partitioning trees when all tabulated points below the current node have been checked or ruled out. During the search, the Euclidean distance to the current best found R rs vector is maintained. At each node the algorithm will descend down a branch only if the image reflectance to match is on the same side of the node's dividing plane as the branch, or is closer to the dividing plane than the current best matched Euclidean distance. Only then can points on the far side of the dividing plane possibly be a better match than the current best match. Note that the choice of n p, maximum sample size for the PCA, may have a small effect on inversion time but has no effect on accuracy. The suggested value of 30,000 is probably suitable for all applications. 3. Example application The following sections illustrate an example ALUT inversion of a 5-parameter shallow water radiative transfer model applied to a hyperspectral airborne image of a tropical coral reef (Fig. 3). Speed and accuracy of ALUT bathymetric retrievals is compared to two implementations of the Levenberg Marquardt successive approximation algorithm, with tide-corrected sonar depth transects used for ground-truth Imagery and forward model The adaptive LUT construction and inversion algorithm was applied to a pixel 17-band CASI image ( nm, FWHM nm) acquired in 2002 of Heron Reef, Australia (23.44 S, E, Fig. 3) with the ALUT constructed to represent a variant of Lee's semianalytic forward model for shallow water remote sensing reflectance (Lee et al., 1998, 1999) where subsurface remote sensing reflectance is estimated from subsurface solar zenith angle (θ w =32 ) subsurface view angle (θ=0 ), and depth in meters, H, by ( " # )! r rs ðλþ r dp rs ðλþ 1 exp 1 + DC uðλþ κλ ð ÞH cosθ w cosθ ( " # ) ð3þ + 1 π ρλ ð Þexp 1 cosθ + DBu ð λ Þ κλ ð ÞH w cosθ To provide a spectral matching search algorithm that is more efficient than an exhaustive search the storage of all the reflectance vectors is post-processed into the structure of a binary space partitioning (BSP) tree. This is achieved by the following algorithm: 1. A list is initialized containing all R rs vectors in the ALUT linked to their corresponding generating parameter values. 2. A principal components analysis (Manly, 1994) is performed on a random selection of n p of the R rs vectors in the list (example: n p =30,000), or the entire list if the current number of vectors in the list is less than n p. 3. A dividing plane in spectral space is established, such that the normal to the plane is the first principal component axis and the plane passes through the mean of the selected R rs vectors. 4. The list of points is subdivided into two lists dependent on which the side of the dividing plane the R rs points lie on. 5. The procedure is applied recursively to the two new lists, starting at Step 2, unless the number of points in the remaining lists is below a predetermined limit (example: 20 points) or the tree depth reaches maximum limit (example: 200 subdivisions). In the resulting tree every node has an associated dividing plane and subdivides into two branches, the actual tabulated reflectances are stored in groups of 20 or less in the terminal nodes (leaves). Searching the BSP tree to find the closest match to an image reflectance is efficient since the global optimum can be found without checking every tabulated R rs vector. The search traverses the tree recursively downwards to the leaves, stepping back up one level Fig. 3. CASI Image of Heron Reef and location of sonar transects. The full image (top panel) is pixels of 1 m 1 m in 17 bands nm with variable band widths from 10 to 20 nm. The images here are monochrome versions of an RGB image based on bands centered at 665, 564 and 480 nm. The central dark region in (1) is the island itself.
4 4 J. Hedley et al. / Remote Sensing of Environment xxx (2009) xxx xxx where the subsurface remote sensing reflectance for optically deep water, r dp rs (λ), and D C u (λ) and D B u (λ), can be estimated from r dp rs ðλþ ½0: :170uðλÞŠuðλÞ D C p uðλþ 1:03 ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1+2:4uðλÞ D B p uðλþ 1:04 ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1+5:4uðλÞ where uðλþ = b b ðλþ= ½aðλÞ + b b ðλþš; κλ ð Þ = aðλþ + b b ðλþ where a(λ) and b b (λ) are the total absorption and backscattering coefficients and estimated as the sum of the effects of water constituents by, aðλþ = a w ðλþ + ½a 0 ðλþ + a 1 ðλþlnpšp + G exp ½ 0:015ðλ 440ÞŠ ð8þ b b ðλþ = b bw ðλþ + Xð400=λÞ Y ð9þ ð4þ ð5þ ð6þ ð7þ The pure water absorption and backscatter a w (λ) and b bw (λ) are considered known (Buiteveld et al., 1994; Pope & Fry, 1997), P, G, X and H arefreeparameterstobesolvedbyinversion,thesedescribe absorption due to phytoplankton, CDOM absorption, particulate backscatter and depth respectively. The particulate backscatter slope is fixed, Y=1, and the phytoplankton spectral absorption parameters, a 0 (λ) and a 1 (λ), are tabulated in (Lee et al., 1998). For the substrate cover reflectance in Eq. (3), ρ(λ), a linear mixture model is used(wettle & Brando, 2006; Brando et al., 2009) whereρλ ð Þ = E i1 ðλþm + E i2 ðλþð1 mþ is a linear mix, in proportion m, of any two benthic reflectances, E i1 (λ), E i2 (λ), taken from a set of thirteen typical reef benthic reflectances including several of sand, live and dead coral, algae and seagrass, so i 1 =2 13 and i 2 =1 i 1 1 giving 78 pair combinations, describable by a single integer parameter, i=1 78. The forward model outputs (Eq. (3)) were translated to vectors of above-surface remote sensing reflectance, R rs using R rs ðþ= j 0:5r rs λ j = 1 1:5rrs λ j where λj is the CASI band j wavelength center (Lee et al., 1999). The CASI image had been previously corrected to units of above-surface diffuse reflectance, R(0 + ) Joyce (2004) and so was converted to above-surface remote sensing reflec- tance using the relation R rs =0:54R 0 þ = kq where Q=4 (Mobley, 1994) and k is a spectrally independent factor to convert above-surface to below-surface diffuse reflectance, k R 0 þ = R 0 ð Þ 0:7, estimated from multiple runs of PlanarRad, an open-source planeparallel radiative transfer model for directional radiance similar to the commercial software Hydrolight (Hedley, 2008; Mobley, 1994) ALUT and Levenberg Marquardt inversion The structure of the forward model to be encoded by the ALUT in the form of Eq. (1) is therefore f i : R 5 YR 17 hp; G; X; H; miir rs for i =1N 78 ð10þ The lower and upper limits of the parameters used to establish the single initial parameter voxel (Fig. 2a) were: 0 P 1.5, 0 G 0.08, 0 X 0.007, 0 H 20, 0 m 1. Hence the maximum retrievable depth is 20 m, the ranges of G and X were chosen for relatively clear waters while the upper bound of P, the component of a(440) due to phytoplankton, was set substantially higher than would normally be expected in these waters to illustrate features of the ALUT construction (Fig. 4). Two ALUTs of spectral space points were generated from the above parameterization, this size of ALUT being the maximum the computer's memory could accommodate. The first, henceforth called Automatic ALUT, followed the previously described procedure exactly with automatic subdivision of parameters based on the distribution of points in spectral space. However, the strong effect of water absorption on reflectance above 700 nm results in an extreme treatment of depth parameter subdivision: shallow depths are generated with steps as fine as 5 mm, while other parts of the ALUT contain no depths between 10 and 20 m (Fig. 4a). This is an accurate picture of the effect of depth on the spectral shape and magnitude of reflectance, but raises the question as to whether accuracy could be improved by a more practical discretization of the depth parameter. To answer this, a second ALUT was constructed where depth subdivision was forced to a minimum of five subdivisions and constrained to eight, so throughout the Detailed Depth ALUT depth is at worst represented to an accuracy of 0.63 m all the way to 20 m, but no finer than 0.08 m in shallowest regions (Fig. 4b). Fig. 4. Example slices through (a) the automatically generated ALUT and (b) ALUT generated with constraints on the depth parameter subdivision. The diagrams show two axes in parameter space, depth and the phytoplankton parameter, P. The values for other parameters are fixed at G=0, X=0, m=0 and the bottom reflectance is pure sand. Boxes represent cross-sections of voxels (Fig. 2) and line intersections are the point locations in the ALUT where R rs has been calculated.
5 J. Hedley et al. / Remote Sensing of Environment xxx (2009) xxx xxx 5 Fig. 5. Depth estimates versus sonar derived bathymetry, n=5057. Dashed lines are the regression slope plus and minus one standard error, Δy is the mean of the absolute value of the errors between estimated and sonar bathymetry. To compare the ALUT inversion approach to alternate methods the model was also inverted by two publicly available C code implementations of the L M algorithm, the Netlib Cephes package 1 and Levmar Results and discussion The Levmar L M optimization and both ALUT inversions resulted in almost identical accuracies for bathymetric retrieval assessed over 5057 sonar data points (Fig. 5). In all three cases r 2 =0.91, the slight difference between the L M and ALUT inversions with respect to regression slopes (1.078 vs. ~0.93) and in the mean of absolute depth error (0.90 vs. ~0.77) can be partly accounted for by the lack of an upper depth parameter constraint on the L M inversion. Constraining ALUT inversion to depths less than 20 m is essentially the incorporation of a priori information, since it was known all sonar depths were less than 20 m. However, incorporating the same upper bounds on the L M inversion actually produced worse results, due to algorithm convergence issues in the vicinity of the constraints. Since the ALUT BSP tree search algorithm returns the global optimum it is free from convergence issues. While the L M successive approximation was able to produce slightly closer spectral matches to the image reflectances than the ALUT (Fig. 6) this did not translate into higher bathymetric accuracy. Both methods were able to obtain spectral matches close to or within image noise levels, estimated from the covariance matrix of pixels over deep homogeneous water (Fig. 6). Hence the ability to produce a spectral match was not limiting for either approach. The Automatic ALUT inversion displayed clear discretization of depth retrievals (Fig. 5b) due to the depth parameter subdivision structure of the ALUT (Fig. 4a). However while the Detailed Depth ALUT produced a cosmetically superior scatterplot (Fig. 5c), accuracy assessed by the three metrics of r 2, regression slope and mean absolute error was virtually identical to the Automatic ALUT (Fig. 5b). The distribution of spectral match errors was also virtually identical (Fig. 6). This indicates that the automatically generated structure of the ALUT does give an efficient representation of the models capability to retrieve environmental parameters. The very similar spread of the L M and ALUT bathymetric retrievals points to a deficiency in the underlying forward radiative transfer model or image noise as the ultimate accuracy limiting factor. 1 URL: 2 URL: lourakis/levmar. Fig. 6. Spectral matching errors over pixels containing sonar data points, expressed as distribution of Euclidean distance between image and matched spectral reflectances, n=4146. Environmental and image noise is estimated as the noise-equivalent deltareflectance, NEΔR rs, by the covariance of spectral reflectance over a homogeneous deep water region, and then by calculating the radius of the bounding sphere of 100 simulated reflectances with noise component derived from the covariance matrix. Match errors below this limit are thus within the error that image noise alone may produce. The computation time for ALUT pixel inversion compares favorably with the two L M implementations, and the initial ALUT construction time is negligible compared to the full image analysis (Table 1). Due to the BSP tree search algorithm, the ALUT inversion time is dependent on how closely the forward model represents the spectra to be inverted. Hence the ALUT inversion is an order of magnitude faster when inverting synthetic spectra generated from the forward model with random parameter values (Table 1). Table 1 Per-pixel inversion time from the two L M implementations and the ALUT algorithm. Algorithm Table Per-pixel inversion time Whole image construction Image Synthetic Analysis (min) (ms) (ms) (h) Levmar L M n/a 580 Cephes L M n/a ALUT ~24 Inversion times for the image transect pixels and synthetic reflectances generated randomly from the forward model are shown. Hyphens indicate analyses that were not done. The whole image consists of approx pixels.
6 6 J. Hedley et al. / Remote Sensing of Environment xxx (2009) xxx xxx Increased inversion efficiency with an improved forward model is a desirable property and implies the ALUT algorithm could attain even better whole-image performance than the results presented here. Faster successive approximation algorithms may exist, while the comparison with two freely available implementations is not definitive the ALUT approach clearly offers competitive inversion times. Beyond efficient inversion, the primary advantage of the ALUT methodology is ease of use. The ALUT construction algorithm has only one user parameter, the number of spectral space points to generate. In fact no decision on this needs to be made, since the hierarchical structure of the ALUT means an existing ALUT can always have more points added to it later on and the final structure will be no different to having built an ALUT of that size in the first place. This is invaluable when using a computationally expensive forward model such as Hydrolight, since a small basic ALUT can be built initially for testing image analysis and then refined later without making any previously modeled spectra redundant. Building a conventional look-up table by regular parameter step sizes may result in an unpredictable total construction time, and early termination will result in a partially constructed LUT of no use for inversion. In contrast, ALUTs are fully constructed at all times during construction, so can be built to specific processing time limit, and early termination yields a fully usable ALUT for image inversion. The primary weakness of the ALUT algorithm is that in a worst case scenario the number of points required to describe the shape of the mapping function (Eq. (1)) to a specific level of discretization error in spectral space may rise exponentially as the dimensionality increases. This means the ALUT approach may be inefficient for forward models with a large number of free parameters to be inverted. However, note that it is the dimensionality and space-filling properties in spectral space that are relevant, not the number of parameters per se. A 10-parameter forward model may not necessarily fill out 10-dimensions in spectral space very effectively, due to correlation between spectral bands for example. This potential limitation will therefore be application-specific. 5. Conclusion A generic method that can invert any radiative transfer model that calculates remote sensing reflectance from a series of integer and realvalued parameters has been presented. The method is based on the hierarchical construction of a look-up-table which is post-processed into a binary space partitioning tree. The method has the following key advantages 1) is simple to use with no configuration parameters requiring expert knowledge 2) minimizes the number of forward runs required offering high efficiency for computationally expensive models 3) offers fast inversions, and 4) permits efficient timemanagement by allowing preview analysis before the full table is constructed. These features make the approach highly suitable for operational implementation of image analysis methods. Straightforward and robust inversion allows work to focus on the more important objective of improving the forward radiative transfer model employed. Acknowledgments This work was funded by the UK's Natural Environment Research Council under grant NE/E015654/1, and also benefited from discussions held during meetings of the Office of Naval Research and Australian Research Council funded Working Group for Optically Shallow Water Remote Sensing. Funding for the Heron Reef CASI dataset and field validation work was provided by Australian Research Council grants to S. Phinn. Image corrections and pre-processing were conducted by Karen Joyce. References Brando, V. E., Anstee, J. M., Wettle, M., Dekker, A. G., Phinn, S. R., & Roelfsema, C. (2009). A physics based retrieval and quality assessment of bathymetry from suboptimal hyperspectral data. Remote Sensing of Environment, 113, Buiteveld, H., Hakvoort, J., & Donze, M. (1994). The optical properties of pure water. In J. S. Jaffe (Ed.), Ocean Optics XIIProc. SPIE, Vol (pp ). Darvishzadeh, R., Skidmore, A., Schlerf, M., & Atzberger, C. (2008). Inversion of a radiative transfer model for estimating vegetation lai and chlorophyll in a heterogeneous grassland. Remote Sensing of Environment, 112(5), Gastellu-Etchegorry, J. P., Gascon, F., & Estève, P. (2003). An interpolation procedure for generalizing a look-up table inversion method. Remote Sensing of Environment, 87 (1), Goodman, J., & Ustin, S. L. (2007). Classification of benthic composition in a coral reef environment using spectral unmixing. Journal of Applied Remote Sensing, 1(1), Hedley, J. (2008). A three-dimensional radiative transfer model for shallow water environments. Optics Express, 16(26), Joyce, K., A method for mapping live coral cover using remote sensing. Ph.D. thesis, The University of Queensland. Klonowski, W. M., Fearns, P. R., & Lynch, M. J. (2007). Retrieving key benthic cover types and bathymetry from hyperspectral imagery. Journal of Applied Remote Sensing, 1(1), Lee, Z., Carder, K., Mobley, C., Steward, R., & Patch, J. (1998). Hyperspectral remote sensing for shallow waters. I. A semianalytical model. Applied Optics, 37(27), Lee, Z., Carder, K., Mobley, C., Steward, R., & Patch, J. (1999). Hyperspectral remote sensing for shallow waters: 2. Deriving bottom depths and water properties by optimization. Applied Optics, 38(18), Liang, S. (2007). Recent developments in estimating land surface biogeophysical variables from optical remote sensing. Progress in Physical Geography, 31(5), Liang, S., Zhong, B., & Fang, H. (2006). Improved estimation of aerosol optical depth from MODIS imagery over land surfaces. Remote Sensing of Environment, 104(4), Manly, B. F. J. (1994). Multivariate statistical methods, 2nd Edition. Chapman and Hall. Mobley, C. D. (1994). Light and Water. San Diego: Academic Press. Mobley, C. D., Sundman, L. K., Davis, C. O., Bowles, J. H., Downes, T. V., Leathers, R. A., Montes, M. J., Bissett, W. P., Kohler, D. D. R., Reid, R. P., Louchard, E. M., & Gleason, A. (2005). Interpretation of hyperspectral remote-sensing imagery by spectrum matching and look-up tables. Applied Optics, 44(17), Pope, R. M., & Fry, E. S. (1997). Absorption spectrum ( nm) of pure water. II. Integrating cavity measurements. Applied Optics, 36(33), Wettle, M., & Brando, V. E. (2006). Sambuca: Semi-analytical model for bathymetry, unmixing and concentration assessment. Tech. Rep. CSIRO Land and Water Science Report 22/06 Canberra, Australia: CSIRO. Wolfe, M. A. (1978). Numerical methods for unconstrained optimization. New York: Van Nostrand Reinhold Company. Zhang, Y., Chen, J. M., Miller, J. R., & Noland, T. L. (2008). Leaf chlorophyll content retrieval from airborne hyperspectral remote sensing imagery. Remote Sensing of Environment, 112(7),
Algorithm Comparison for Shallow-Water Remote Sensing
DISTRIBUTION STATEMENT A: Approved for public release; distribution is unlimited. Algorithm Comparison for Shallow-Water Remote Sensing Curtis D. Mobley Sequoia Scientific, Inc. 2700 Richards Road, Suite
More informationOptimizing Machine Learning Algorithms for Hyperspectral Very Shallow Water (VSW) Products
Optimizing Machine Learning Algorithms for Hyperspectral Very Shallow Water (VSW) Products W. Paul Bissett Florida Environmental Research Institute 10500 University Center Dr. Suite 140 Tampa, FL 33612
More informationContinued Development of the Look-up-table (LUT) Methodology For Interpretation of Remotely Sensed Ocean Color Data
Continued Development of the Look-up-table (LUT) Methodology For Interpretation of Remotely Sensed Ocean Color Data Curtis D. Mobley Sequoia Scientific, Inc. 2700 Richards Road, Suite 107 Bellevue, WA
More informationContinued Development of the Look-up-table (LUT) Methodology For Interpretation of Remotely Sensed Ocean Color Data
Continued Development of the Look-up-table (LUT) Methodology For Interpretation of Remotely Sensed Ocean Color Data W. Paul Bissett Florida Environmental Research Institute 10500 University Center Dr.,
More informationMultiple Optical Shallow Water Inversion Methods for Bathymetry, In-Water Optics, and Benthos Mapping : How Do They Compare?
Multiple Optical Shallow Water Inversion Methods for Bathymetry, In-Water Optics, and Benthos Mapping : How Do They Compare? Dekker A.G. (1,2), *Phinn S.R. (2), Anstee J. (1), Bissett P. (3), Brando V.E.
More informationSAMBUCA Semi-Analytical Model for Bathymetry, Un-mixing, and Concentration Assessment. Magnus Wettle and Vittorio Ernesto Brando
SAMBUCA Semi-Analytical Model for Bathymetry, Un-mixing, and Concentration Assessment Magnus Wettle and Vittorio Ernesto Brando CSIRO Land and Water Science Report 22/06 July 2006 Copyright and Disclaimer
More informationCHALLENGES GLOBAL APPROACH TO:
CHALLENGES GLOBAL APPROACH TO: SPECTRAL LIBRARIES(MACROPHYTES, MACRO_ALGAE, SEAGRASSES, MUD, SAND, RUBBLE, DETRITUS, BENTHIC MICRO_ALGAE PHYTOPLANKTON) INLAND BIO-OPTICAL DATABASES-OPEN SOURCE INLAND WATER
More informationA Look-up-Table Approach to Inverting Remotely Sensed Ocean Color Data
A Look-up-Table Approach to Inverting Remotely Sensed Ocean Color Data Curtis D. Mobley Sequoia Scientific, Inc. Westpark Technical Center 15317 NE 90th Street Redmond, WA 98052 phone: 425-867-2464 x 109
More informationChallenges in detecting trend and seasonal changes in bathymetry derived from HICO imagery: a case study of Shark Bay, Western Australia
Challenges in detecting trend and seasonal changes in bathymetry derived from HICO imagery: a case study of Shark Bay, Western Australia Rodrigo Garcia 1, Peter Fearns 1, Lachlan McKinna 1,2 1 Remote Sensing
More informationThree dimensional light environment: Coral reefs and seagrasses
Three dimensional light environment: Coral reefs and seagrasses 1) Three-dimensional radiative transfer modelling 2) Photobiology in submerged canopies 3) Sun glint correction of high spatial resolution
More informationAnalysis of Hyperspectral Data for Coastal Bathymetry and Water Quality
Analysis of Hyperspectral Data for Coastal Bathymetry and Water Quality William Philpot Cornell University 453 Hollister Hall, Ithaca, NY 14853 phone: (607) 255-0801 fax: (607) 255-9004 e-mail: wdp2@cornell.edu
More informationHyperspectral Remote Sensing
Hyperspectral Remote Sensing Multi-spectral: Several comparatively wide spectral bands Hyperspectral: Many (could be hundreds) very narrow spectral bands GEOG 4110/5100 30 AVIRIS: Airborne Visible/Infrared
More information2017 Summer Course Optical Oceanography and Ocean Color Remote Sensing. Overview of HydroLight and EcoLight
2017 Summer Course Optical Oceanography and Ocean Color Remote Sensing Curtis Mobley Overview of HydroLight and EcoLight Darling Marine Center, University of Maine July 2017 Copyright 2017 by Curtis D.
More informationDERIVATIVE BASED HYPERSPECRAL ALGORITHM FOR BATHYMETRIC MAPPING
DERIVATIVE BASED HYPERSPECRAL ALGORITHM FOR BATHYMETRIC MAPPING David D. R. Kohler, William D. Philpot, Civil and Environmental Engineering, Cornell University Ithaca, NY14853 Curtis D. Mobley Sequoia
More informationThe Unsolved Problem of Atmospheric Correction for Airborne Hyperspectral Remote Sensing of Shallow Waters
The Unsolved Problem of Atmospheric Correction for Airborne Hyperspectral Remote Sensing of Shallow Waters Curtis Mobley Vice President for Science and Senior Scientist Sequoia Scientific, Inc. Bellevue,
More informationShallow-water Remote Sensing: Lecture 1: Overview
Shallow-water Remote Sensing: Lecture 1: Overview Curtis Mobley Vice President for Science and Senior Scientist Sequoia Scientific, Inc. Bellevue, WA 98005 curtis.mobley@sequoiasci.com IOCCG Course Villefranche-sur-Mer,
More informationOcean color algorithms in optically shallow waters: Limitations and improvements
Ocean color algorithms in optically shallow waters: Limitations and improvements Kendall L. Carder *a, Jennifer P. Cannizzaro a, Zhongping Lee b a University of South Florida, 140 7 th Ave. S, St. Petersburg,
More informationREMOTE SENSING OF BENTHIC HABITATS IN SOUTHWESTERN PUERTO RICO
REMOTE SENSING OF BENTHIC HABITATS IN SOUTHWESTERN PUERTO RICO Fernando Gilbes Santaella Dep. of Geology Roy Armstrong Dep. of Marine Sciences University of Puerto Rico at Mayagüez fernando.gilbes@upr.edu
More informationHyperspectral Remote Sensing of Coastal Environments
Hyperspectral Remote Sensing of Coastal Environments Miguel Vélez-Reyes, Ph.D. Laboratory for Applied Remote Sensing and Image Processing Center for Subsurface Sensing and Imaging Systems University of
More information2017 Summer Course on Optical Oceanography and Ocean Color Remote Sensing. Introduction to Remote Sensing
2017 Summer Course on Optical Oceanography and Ocean Color Remote Sensing Introduction to Remote Sensing Curtis Mobley Delivered at the Darling Marine Center, University of Maine July 2017 Copyright 2017
More informationSEA BOTTOM MAPPING FROM ALOS AVNIR-2 AND QUICKBIRD SATELLITE DATA
SEA BOTTOM MAPPING FROM ALOS AVNIR-2 AND QUICKBIRD SATELLITE DATA Mohd Ibrahim Seeni Mohd, Nurul Nadiah Yahya, Samsudin Ahmad Faculty of Geoinformation and Real Estate, Universiti Teknologi Malaysia, 81310
More informationWater Column Correction for Coral Reef Studies by Remote Sensing
Sensors 2014, 14, 16881-16931; doi:10.3390/s140916881 Review OPEN ACCESS sensors ISSN 1424-8220 www.mdpi.com/journal/sensors Water Column Correction for Coral Reef Studies by Remote Sensing Maria Laura
More informationInversion of irradiance and remote sensing reflectance in shallow water between 400 and 800 nm for calculations of water and bottom properties
Inversion of irradiance and remote sensing reflectance in shallow water between 400 and 800 nm for calculations of water and bottom properties Andreas Albert and Peter Gege What we believe to be a new
More informationLecture 7. Spectral Unmixing. Summary. Mixtures in Remote Sensing
Lecture 7 Spectral Unmixing Summary This lecture will introduce you to the concepts of linear spectral mixing. This methods is sometimes also called: Spectral Mixture Analysis (SMA: Wessman et al 1997)
More informationLab 9. Julia Janicki. Introduction
Lab 9 Julia Janicki Introduction My goal for this project is to map a general land cover in the area of Alexandria in Egypt using supervised classification, specifically the Maximum Likelihood and Support
More informationAlgorithm Acceleration for Geospatial Analysis
Algorithm Acceleration for Geospatial Analysis GTC 2012 / May 14-18 / San Jose, CA James Goodman, PhD, PE President / CEO HySpeed Computing LLC jgoodman@hyspeedcomputing.com Matthew Sellitto & David Kaeli
More informationRemote Sensing of Environment
Remote Sensing of Environment 113 (2009) 1025 1045 Contents lists available at ScienceDirect Remote Sensing of Environment journal homepage: www.elsevier.com/locate/rse A forward image model for passive
More informationSpectral Classification
Spectral Classification Spectral Classification Supervised versus Unsupervised Classification n Unsupervised Classes are determined by the computer. Also referred to as clustering n Supervised Classes
More informationWorkhorse ADCP Multi- Directional Wave Gauge Primer
Acoustic Doppler Solutions Workhorse ADCP Multi- Directional Wave Gauge Primer Brandon Strong October, 2000 Principles of ADCP Wave Measurement The basic principle behind wave the measurement, is that
More informationClustering and Visualisation of Data
Clustering and Visualisation of Data Hiroshi Shimodaira January-March 28 Cluster analysis aims to partition a data set into meaningful or useful groups, based on distances between data points. In some
More informationRetrieval of optical and microphysical properties of ocean constituents using polarimetric remote sensing
Retrieval of optical and microphysical properties of ocean constituents using polarimetric remote sensing Presented by: Amir Ibrahim Optical Remote Sensing Laboratory, The City College of the City University
More informationChapter 4. Clustering Core Atoms by Location
Chapter 4. Clustering Core Atoms by Location In this chapter, a process for sampling core atoms in space is developed, so that the analytic techniques in section 3C can be applied to local collections
More informationClassification of Hyperspectral Breast Images for Cancer Detection. Sander Parawira December 4, 2009
1 Introduction Classification of Hyperspectral Breast Images for Cancer Detection Sander Parawira December 4, 2009 parawira@stanford.edu In 2009 approximately one out of eight women has breast cancer.
More informationMotivation. Aerosol Retrieval Over Urban Areas with High Resolution Hyperspectral Sensors
Motivation Aerosol etrieval Over Urban Areas with High esolution Hyperspectral Sensors Barry Gross (CCNY) Oluwatosin Ogunwuyi (Ugrad CCNY) Brian Cairns (NASA-GISS) Istvan Laszlo (NOAA-NESDIS) Aerosols
More informationREMOTE SENSING OF VERTICAL IOP STRUCTURE
REMOTE SENSING OF VERTICAL IOP STRUCTURE W. Scott Pegau College of Oceanic and Atmospheric Sciences Ocean. Admin. Bldg. 104 Oregon State University Corvallis, OR 97331-5503 Phone: (541) 737-5229 fax: (541)
More informationOcean Optics Inversion Algorithm
Ocean Optics Inversion Algorithm N. J. McCormick 1 and Eric Rehm 2 1 University of Washington Department of Mechanical Engineering Seattle, WA 98195-26 mccor@u.washington.edu 2 University of Washington
More informationBottom Albedo Images to Improve Classification of Benthic Habitat Maps
Bottom Albedo Images to Improve Classification of Benthic Habitat Maps William J Hernandez, Ph.D Post-Doctoral Researcher NOAA CREST University of Puerto Rico, Mayaguez, Puerto Rico, Global Science and
More informationSENSITIVITY ANALYSIS OF SEMI-ANALYTICAL MODELS OF DIFFUSE ATTENTUATION OF DOWNWELLING IRRADIANCE IN LAKE BALATON
SENSITIVITY ANALYSIS OF SEMI-ANALYTICAL MODELS OF DIFFUSE ATTENTUATION OF DOWNWELLING IRRADIANCE IN LAKE BALATON Van der Zande D. (1), Blaas M. (2), Nechad B. (1) (1) Royal Belgian Institute of Natural
More informationSections 3-6 have been substantially modified to make the paper more comprehensible. Several figures have been re-plotted and figure captions changed.
Response to First Referee s Comments General Comments Sections 3-6 have been substantially modified to make the paper more comprehensible. Several figures have been re-plotted and figure captions changed.
More informationInteractive comment on Quantification and mitigation of the impact of scene inhomogeneity on Sentinel-4 UVN UV-VIS retrievals by S. Noël et al.
Atmos. Meas. Tech. Discuss., www.atmos-meas-tech-discuss.net/5/c741/2012/ Author(s) 2012. This work is distributed under the Creative Commons Attribute 3.0 License. Atmospheric Measurement Techniques Discussions
More informationInteractive comment on Quantification and mitigation of the impact of scene inhomogeneity on Sentinel-4 UVN UV-VIS retrievals by S. Noël et al.
Atmos. Meas. Tech. Discuss., 5, C741 C750, 2012 www.atmos-meas-tech-discuss.net/5/c741/2012/ Author(s) 2012. This work is distributed under the Creative Commons Attribute 3.0 License. Atmospheric Measurement
More informationIntroduction of spatial smoothness constraints via linear diffusion for optimization-based hyperspectral coastal ocean remote-sensing inversion
JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 113,, doi:10.1029/2007jc004441, 2008 Introduction of spatial smoothness constraints via linear diffusion for optimization-based hyperspectral coastal ocean remote-sensing
More informationCHRIS Proba Workshop 2005 II
CHRIS Proba Workshop 25 Analyses of hyperspectral and directional data for agricultural monitoring using the canopy reflectance model SLC Progress in the Upper Rhine Valley and Baasdorf test-sites Dr.
More informationPrincipal Component Image Interpretation A Logical and Statistical Approach
Principal Component Image Interpretation A Logical and Statistical Approach Md Shahid Latif M.Tech Student, Department of Remote Sensing, Birla Institute of Technology, Mesra Ranchi, Jharkhand-835215 Abstract
More informationAn Accurate Method for Skew Determination in Document Images
DICTA00: Digital Image Computing Techniques and Applications, 1 January 00, Melbourne, Australia. An Accurate Method for Skew Determination in Document Images S. Lowther, V. Chandran and S. Sridharan Research
More informationAnalysis of the In-Water and Sky Radiance Distribution Data Acquired During the Radyo Project
DISTRIBUTION STATEMENT A. Approved for public release; distribution is unlimited. Analysis of the In-Water and Sky Radiance Distribution Data Acquired During the Radyo Project Kenneth J. Voss Physics Department,
More informationImprovements to the SHDOM Radiative Transfer Modeling Package
Improvements to the SHDOM Radiative Transfer Modeling Package K. F. Evans University of Colorado Boulder, Colorado W. J. Wiscombe National Aeronautics and Space Administration Goddard Space Flight Center
More informationB553 Lecture 12: Global Optimization
B553 Lecture 12: Global Optimization Kris Hauser February 20, 2012 Most of the techniques we have examined in prior lectures only deal with local optimization, so that we can only guarantee convergence
More informationAn analytical model for subsurface irradiance and remote sensing reflectance in deep and shallow case-2 waters
An analytical model for subsurface irradiance and remote sensing reflectance in deep and shallow case-2 waters A. Albert German Aerospace Center DLR), Remote Sensing Technology Institute, D-82230 Wessling,
More informationUncertainties in the Products of Ocean-Colour Remote Sensing
Chapter 3 Uncertainties in the Products of Ocean-Colour Remote Sensing Emmanuel Boss and Stephane Maritorena Data products retrieved from the inversion of in situ or remotely sensed oceancolour data are
More informationSELECTION OF THE OPTIMAL PARAMETER VALUE FOR THE LOCALLY LINEAR EMBEDDING ALGORITHM. Olga Kouropteva, Oleg Okun and Matti Pietikäinen
SELECTION OF THE OPTIMAL PARAMETER VALUE FOR THE LOCALLY LINEAR EMBEDDING ALGORITHM Olga Kouropteva, Oleg Okun and Matti Pietikäinen Machine Vision Group, Infotech Oulu and Department of Electrical and
More informationTOA RADIANCE SIMULATOR FOR THE NEW HYPERSPECTRAL MISSIONS: STORE (SIMULATOR OF TOA RADIANCE)
TOA RADIANCE SIMULATOR FOR THE NEW HYPERSPECTRAL MISSIONS: STORE (SIMULATOR OF TOA RADIANCE) Malvina Silvestri Istituto Nazionale di Geofisica e Vulcanologia In the frame of the Italian Space Agency (ASI)
More informationEstimating oceanic primary production using. vertical irradiance and chlorophyll profiles. from ocean gliders in the North Atlantic
Estimating oceanic primary production using vertical irradiance and chlorophyll profiles from ocean gliders in the North Atlantic Victoria S. Hemsley* 1,2, Timothy J. Smyth 3, Adrian P. Martin 2, Eleanor
More informationIdentifying Layout Classes for Mathematical Symbols Using Layout Context
Rochester Institute of Technology RIT Scholar Works Articles 2009 Identifying Layout Classes for Mathematical Symbols Using Layout Context Ling Ouyang Rochester Institute of Technology Richard Zanibbi
More informationQuality assessment of RS data. Remote Sensing (GRS-20306)
Quality assessment of RS data Remote Sensing (GRS-20306) Quality assessment General definition for quality assessment (Wikipedia) includes evaluation, grading and measurement process to assess design,
More information2017 Summer Course on Optical Oceanography and Ocean Color Remote Sensing. Apparent Optical Properties and the BRDF
2017 Summer Course on Optical Oceanography and Ocean Color Remote Sensing Curtis Mobley Apparent Optical Properties and the BRDF Delivered at the Darling Marine Center, University of Maine July 2017 Copyright
More informationThe latest trend of hybrid instrumentation
Multivariate Data Processing of Spectral Images: The Ugly, the Bad, and the True The results of various multivariate data-processing methods of Raman maps recorded with a dispersive Raman microscope are
More informationCluster Analysis for Microarray Data
Cluster Analysis for Microarray Data Seventh International Long Oligonucleotide Microarray Workshop Tucson, Arizona January 7-12, 2007 Dan Nettleton IOWA STATE UNIVERSITY 1 Clustering Group objects that
More informationPredicting Atmospheric Parameters using Canonical Correlation Analysis
Predicting Atmospheric Parameters using Canonical Correlation Analysis Emmett J Ientilucci Digital Imaging and Remote Sensing Laboratory Chester F Carlson Center for Imaging Science Rochester Institute
More informationFifteenth ARM Science Team Meeting Proceedings, Daytona Beach, Florida, March 14-18, 2005
Assessing the Impact of the Plane-Parallel Cloud Assumption used in Computing Shortwave Heating Rate Profiles for the Broadband Heating Rate Profile Project W. O Hirok Institute for Computational Earth
More informationLinking sun-induced fluorescence and photosynthesis in a forest ecosystem
Linking sun-induced fluorescence and photosynthesis in a forest ecosystem COST ES1309 Tagliabue G, Panigada C, Dechant B, Celesti M, Cogliati S, Colombo R, Julitta T, Meroni M, Schickling A, Schuettemeyer
More informationAPPLICATION OF SOFTMAX REGRESSION AND ITS VALIDATION FOR SPECTRAL-BASED LAND COVER MAPPING
APPLICATION OF SOFTMAX REGRESSION AND ITS VALIDATION FOR SPECTRAL-BASED LAND COVER MAPPING J. Wolfe a, X. Jin a, T. Bahr b, N. Holzer b, * a Harris Corporation, Broomfield, Colorado, U.S.A. (jwolfe05,
More informationGradient-Free Boundary Tracking* Zhong Hu Faculty Advisor: Todd Wittman August 2007 UCLA Department of Mathematics
Section I. Introduction Gradient-Free Boundary Tracking* Zhong Hu Faculty Advisor: Todd Wittman August 2007 UCLA Department of Mathematics In my paper, my main objective is to track objects in hyperspectral
More informationAN IMPROVED HYBRIDIZED K- MEANS CLUSTERING ALGORITHM (IHKMCA) FOR HIGHDIMENSIONAL DATASET & IT S PERFORMANCE ANALYSIS
AN IMPROVED HYBRIDIZED K- MEANS CLUSTERING ALGORITHM (IHKMCA) FOR HIGHDIMENSIONAL DATASET & IT S PERFORMANCE ANALYSIS H.S Behera Department of Computer Science and Engineering, Veer Surendra Sai University
More informationSupplementary Figure 1. Decoding results broken down for different ROIs
Supplementary Figure 1 Decoding results broken down for different ROIs Decoding results for areas V1, V2, V3, and V1 V3 combined. (a) Decoded and presented orientations are strongly correlated in areas
More informationVisualization and Analysis of Inverse Kinematics Algorithms Using Performance Metric Maps
Visualization and Analysis of Inverse Kinematics Algorithms Using Performance Metric Maps Oliver Cardwell, Ramakrishnan Mukundan Department of Computer Science and Software Engineering University of Canterbury
More informationStudy on LAI Sampling Strategy and Product Validation over Non-uniform Surface. Lingling Ma, Xiaohua Zhu, Yongguang Zhao
of Opto Electronics Chinese of Sciences Study on LAI Sampling Strategy and Product Validation over Non-uniform Surface Lingling Ma, Xiaohua Zhu, Yongguang Zhao of (AOE) Chinese of Sciences (CAS) 2014-1-28
More informationFigure 1: Workflow of object-based classification
Technical Specifications Object Analyst Object Analyst is an add-on package for Geomatica that provides tools for segmentation, classification, and feature extraction. Object Analyst includes an all-in-one
More informationENVI Tutorial: Vegetation Hyperspectral Analysis
ENVI Tutorial: Vegetation Hyperspectral Analysis Table of Contents OVERVIEW OF THIS TUTORIAL...1 HyMap Processing Flow...4 VEGETATION HYPERSPECTRAL ANALYSIS...4 Examine the Jasper Ridge HyMap Radiance
More informationFuzzy Entropy based feature selection for classification of hyperspectral data
Fuzzy Entropy based feature selection for classification of hyperspectral data Mahesh Pal Department of Civil Engineering NIT Kurukshetra, 136119 mpce_pal@yahoo.co.uk Abstract: This paper proposes to use
More information1. Estimation equations for strip transect sampling, using notation consistent with that used to
Web-based Supplementary Materials for Line Transect Methods for Plant Surveys by S.T. Buckland, D.L. Borchers, A. Johnston, P.A. Henrys and T.A. Marques Web Appendix A. Introduction In this on-line appendix,
More informationFourier analysis of low-resolution satellite images of cloud
New Zealand Journal of Geology and Geophysics, 1991, Vol. 34: 549-553 0028-8306/91/3404-0549 $2.50/0 Crown copyright 1991 549 Note Fourier analysis of low-resolution satellite images of cloud S. G. BRADLEY
More informationAttenuation of visible solar radiation in the upper water column: A model based on IOPs
Attenuation of visible solar radiation in the upper water column: A model based on IOPs ZhongPing Lee, KePing Du 2, Robert Arnone, SooChin Liew 3, Bradley Penta Naval Research Laboratory Code 7333 Stennis
More informationAnalysis of the In-Water and Sky Radiance Distribution Data Acquired During the Radyo Project
DISTRIBUTION STATEMENT A. Approved for public release; distribution is unlimited. Analysis of the In-Water and Sky Radiance Distribution Data Acquired During the Radyo Project Kenneth J. Voss Physics Department,
More informationCOMP30019 Graphics and Interaction Scan Converting Polygons and Lines
COMP30019 Graphics and Interaction Scan Converting Polygons and Lines Department of Computer Science and Software Engineering The Lecture outline Introduction Scan conversion Scan-line algorithm Edge coherence
More informationINF 4300 Classification III Anne Solberg The agenda today:
INF 4300 Classification III Anne Solberg 28.10.15 The agenda today: More on estimating classifier accuracy Curse of dimensionality and simple feature selection knn-classification K-means clustering 28.10.15
More informationENVI Classic Tutorial: Multispectral Analysis of MASTER HDF Data 2
ENVI Classic Tutorial: Multispectral Analysis of MASTER HDF Data Multispectral Analysis of MASTER HDF Data 2 Files Used in This Tutorial 2 Background 2 Shortwave Infrared (SWIR) Analysis 3 Opening the
More informationAn Intelligent Clustering Algorithm for High Dimensional and Highly Overlapped Photo-Thermal Infrared Imaging Data
An Intelligent Clustering Algorithm for High Dimensional and Highly Overlapped Photo-Thermal Infrared Imaging Data Nian Zhang and Lara Thompson Department of Electrical and Computer Engineering, University
More informationGEOG 4110/5100 Advanced Remote Sensing Lecture 2
GEOG 4110/5100 Advanced Remote Sensing Lecture 2 Data Quality Radiometric Distortion Radiometric Error Correction Relevant reading: Richards, sections 2.1 2.8; 2.10.1 2.10.3 Data Quality/Resolution Spatial
More informationA New Online Clustering Approach for Data in Arbitrary Shaped Clusters
A New Online Clustering Approach for Data in Arbitrary Shaped Clusters Richard Hyde, Plamen Angelov Data Science Group, School of Computing and Communications Lancaster University Lancaster, LA1 4WA, UK
More informationJPEG compression of monochrome 2D-barcode images using DCT coefficient distributions
Edith Cowan University Research Online ECU Publications Pre. JPEG compression of monochrome D-barcode images using DCT coefficient distributions Keng Teong Tan Hong Kong Baptist University Douglas Chai
More informationCOMBINING HIGH SPATIAL RESOLUTION OPTICAL AND LIDAR DATA FOR OBJECT-BASED IMAGE CLASSIFICATION
COMBINING HIGH SPATIAL RESOLUTION OPTICAL AND LIDAR DATA FOR OBJECT-BASED IMAGE CLASSIFICATION Ruonan Li 1, Tianyi Zhang 1, Ruozheng Geng 1, Leiguang Wang 2, * 1 School of Forestry, Southwest Forestry
More informationSearch direction improvement for gradient-based optimization problems
Computer Aided Optimum Design in Engineering IX 3 Search direction improvement for gradient-based optimization problems S Ganguly & W L Neu Aerospace and Ocean Engineering, Virginia Tech, USA Abstract
More informationMETRIC PLANE RECTIFICATION USING SYMMETRIC VANISHING POINTS
METRIC PLANE RECTIFICATION USING SYMMETRIC VANISHING POINTS M. Lefler, H. Hel-Or Dept. of CS, University of Haifa, Israel Y. Hel-Or School of CS, IDC, Herzliya, Israel ABSTRACT Video analysis often requires
More informationQuantifying the Dynamic Ocean Surface Using Underwater Radiometric Measurements
DISTRIBUTION STATEMENT A. Approved for public release; distribution is unlimited. Quantifying the Dynamic Ocean Surface Using Underwater Radiometric Measurements Dick K.P. Yue Center for Ocean Engineering
More informationLecture on Modeling Tools for Clustering & Regression
Lecture on Modeling Tools for Clustering & Regression CS 590.21 Analysis and Modeling of Brain Networks Department of Computer Science University of Crete Data Clustering Overview Organizing data into
More informationDetermining satellite rotation rates for unresolved targets using temporal variations in spectral signatures
Determining satellite rotation rates for unresolved targets using temporal variations in spectral signatures Joseph Coughlin Stinger Ghaffarian Technologies Colorado Springs, CO joe.coughlin@sgt-inc.com
More informationScanner Parameter Estimation Using Bilevel Scans of Star Charts
ICDAR, Seattle WA September Scanner Parameter Estimation Using Bilevel Scans of Star Charts Elisa H. Barney Smith Electrical and Computer Engineering Department Boise State University, Boise, Idaho 8375
More informationFast Fuzzy Clustering of Infrared Images. 2. brfcm
Fast Fuzzy Clustering of Infrared Images Steven Eschrich, Jingwei Ke, Lawrence O. Hall and Dmitry B. Goldgof Department of Computer Science and Engineering, ENB 118 University of South Florida 4202 E.
More informationECLT 5810 Clustering
ECLT 5810 Clustering What is Cluster Analysis? Cluster: a collection of data objects Similar to one another within the same cluster Dissimilar to the objects in other clusters Cluster analysis Grouping
More informationCentre for Digital Image Measurement and Analysis, School of Engineering, City University, Northampton Square, London, ECIV OHB
HIGH ACCURACY 3-D MEASUREMENT USING MULTIPLE CAMERA VIEWS T.A. Clarke, T.J. Ellis, & S. Robson. High accuracy measurement of industrially produced objects is becoming increasingly important. The techniques
More informationAlgorithms for 3D Isometric Shape Correspondence
Algorithms for 3D Isometric Shape Correspondence Yusuf Sahillioğlu Computer Eng. Dept., Koç University, Istanbul, Turkey (PhD) Computer Eng. Dept., METU, Ankara, Turkey (Asst. Prof.) 2 / 53 Problem Definition
More informationSilhouette-based Multiple-View Camera Calibration
Silhouette-based Multiple-View Camera Calibration Prashant Ramanathan, Eckehard Steinbach, and Bernd Girod Information Systems Laboratory, Electrical Engineering Department, Stanford University Stanford,
More informationD-Optimal Designs. Chapter 888. Introduction. D-Optimal Design Overview
Chapter 888 Introduction This procedure generates D-optimal designs for multi-factor experiments with both quantitative and qualitative factors. The factors can have a mixed number of levels. For example,
More informationAdaptive Robotics - Final Report Extending Q-Learning to Infinite Spaces
Adaptive Robotics - Final Report Extending Q-Learning to Infinite Spaces Eric Christiansen Michael Gorbach May 13, 2008 Abstract One of the drawbacks of standard reinforcement learning techniques is that
More informationCHAPTER 6 MODIFIED FUZZY TECHNIQUES BASED IMAGE SEGMENTATION
CHAPTER 6 MODIFIED FUZZY TECHNIQUES BASED IMAGE SEGMENTATION 6.1 INTRODUCTION Fuzzy logic based computational techniques are becoming increasingly important in the medical image analysis arena. The significant
More information4.12 Generalization. In back-propagation learning, as many training examples as possible are typically used.
1 4.12 Generalization In back-propagation learning, as many training examples as possible are typically used. It is hoped that the network so designed generalizes well. A network generalizes well when
More informationRADIANCE IN THE OCEAN: EFFECTS OF WAVE SLOPE AND RAMAN SCATTERING NEAR THE SURFACE AND AT DEPTHS THROUGH THE ASYMPTOTIC REGION
RADIANCE IN THE OCEAN: EFFECTS OF WAVE SLOPE AND RAMAN SCATTERING NEAR THE SURFACE AND AT DEPTHS THROUGH THE ASYMPTOTIC REGION A Thesis by JULIE MARIE SLANKER Submitted to the Office of Graduate Studies
More informationJournal of Theoretical and Applied Mechanics, Sofia, 2015, vol. 45, No. 2, pp
Journal of Theoretical and Applied Mechanics, Sofia, 2015, vol. 45, No. 2, pp. 59 74 SOLID MECHANICS WAVE PROPAGATION DUE TO AN EMBEDDED SEISMIC SOURCE IN A GRADED HALF-PLANE WITH RELIEF PECULIARITIES.
More informationHyperspectral remote sensing for shallow waters: 2. Deriving bottom depths and water properties by optimization
Hyperspectral remote sensing for shallow waters: 2. Deriving bottom depths and water properties by optimization Zhongping Lee, Kendall L. Carder, Curtis D. Mobley, Robert G. Steward, and Jennifer S. Patch
More information