ABSTRACT. data in 3D form from samples in both reflection and transmission modes. The current

Transcription

1 ABSTRACT The hyperspectral imaging system has the capability to collect spatial and spectral data in 3D form from samples in both reflection and transmission modes. The current software used with the existing lab-based system allows only data acquisition and control of acquisition parameters; a software interface is needed to compute unknown optical properties such as absorption and scattering coefficients for the analyzed samples. The Inverse Adding-Doubling method, which is implemented using the C programming language, has been widely used with the double integrating spheres optical system at single wavelengths when the light source is laser. The double integrating spheres system collects cumulative diffuse reflection and transmission data from samples. The hyperspectral imaging system provides additional spectral information when the incident light is a broadband source. In this project, the Inverse Adding-Doubling method is adapted and integrated for use on the data acquired by the hyperspectral imaging system to calculate selected unknown optical properties using a GUI environment. In addition, the developed GUI provides a user interface for interactively finding the unknown extrinsic optical properties such as reflection and transmission coefficients and anisotropy by using theories and methods such as Monte Carlo simulations, Mie Theory and the Adding-Doubling method, which are based on solving forward problems, and have been adapted for the HIS data. ii

2 TABLE OF CONTENTS ABSTARCT ii TABLE OF CONTENTS...iii LIST OF FIGURES v LIST OF TABLES...vii 1. Background and Rationale Hyperspectral Imaging Goals and Objectives Implementing Different Analysis Tools for the Hyperspectral Data Solving Inverse Problems Inverse Adding-Doubling Method Solving Forward Problems Adding-Doubling Method Monte Carlo Simulations Time Monte Carlo Simulation Small Monte Carlo Simulation Tiny Monte Carlo Simulation Mie Theory LabVIEW Programming Acquiring Spectral Data using the HIS and LabVIEW LabVIEW Design and Implementation LabVIEW Design Interface Software Engineering Techniques Requirements Software Tools and General Requirements Functional Vs Non Functional Requirements Structured Requirements 21 iii

3 3.5 Analysis Design Results, Testing and Validation User Interface Inverse Adding-Doubling Method Monte Carlo Simulations Time Monte Carlo Simulation Small Monte Carlo Simulation Tiny Monte Carlo Simulation Adding-Doubling Method Mie Theory Conclusions Future Work Acknowledgements...57 Bibliography and References...58 iv

4 LIST OF FIGURES Figure 1: Hyperspectral Imaging System in Reflection Mode..2 Figure 2: Hypercube Image.2 Figure 3: a. Hypercube Image.3 b. Spectral Signature from a Single Pixel Location...3 Figure 4: Sample Spectral graphs from a different pixel in the Hyperspectral ImageCube 5 Figure 5: Labview Implemented Functionality Diagram for the Integrated Software Tool System Figure 6: Tools Hierarchical Diagram...17 Figure 7: Use case Diagram for Integrated Software Tools.24 Figure 8: Sequence Diagram for IAD Method...25 Figure 9: Sequence Diagram for Monte Carlo Simulations...26 Figure 10: Sequence Diagram for Adding-Doubling Method...27 Figure 11: Sequence Diagram for Mie Theory..28 Figure 12: User Interface Front Panel 29 Figure 13: User Interface Block Diagram..30 Figure 14: Inverse Adding-Doubling Method Front Panel 32 Figure 15: Inverse Adding-Doubling Method Block Diagram..33 Figure 16: Inverse Adding-Doubling Method Front Panel (input) 34 Figure 17: Inverse Adding-Doubling Method (input text file)..35 Figure 18: Inverse Adding-Doubling Method Front Panel (output)..36 Figure 19: Inverse Adding-Doubling Method (output text file) 37 v

5 Figure 20: Time Monte Carlo Simulation Front Panel...38 Figure 21: Time Monte Carlo Simulation Block Diagram 39 Figure 22: Time Monte Carlo Simulation Front Panel (input)..39 Figure 23: Time Monte Carlo Simulation Front Panel (output) 40 Figure 24: Small Monte Carlo Simulation Front Panel 41 Figure 25: Small Monte Carlo Simulation Block Diagram...42 Figure 26: Small Monte Carlo Simulation Front Panel (input)...42 Figure 27: Small Monte Carlo Simulation Front Panel (output)...43 Figure 28: Tiny Monte Carlo Simulation Front Panel..44 Figure 29: Tiny Monte Carlo Simulation Block Diagram...45 Figure 30: Tiny Monte Carlo Simulation Front Panel (input)...45 Figure 31: Tiny Monte Carlo Simulation Front Panel (output)...46 Figure 32: Adding- Doubling Method Front Panel 47 Figure 33: Adding- Doubling Method Block Diagram...48 Figure 34: Adding- Doubling Method Front Panel (input) 49 Figure 35: Adding- Doubling Method Front Panel (output)..50 Figure 36: Mie Theory Front Pane...51 Figure 37: Mie Theory (input text file)...52 Figure 38: Mie Theory Block Diagram...53 Figure 39: Mie Theory Front Panel (input) 54 Figure 40: Mie Theory Front Panel (output)...54 vi

6 LIST OF TABLES Table 1: Selected Optical Properties that can be computed with the Inverse Adding-Doubling Method.7 Table 2: Selected Optical Properties for Adding-Doubling Method..8 vii

7 1. Background and Rationale Hyperspectral imaging is a relatively new tool used to explore the spectral content of a variety of materials at different spatial locations. Hyperspectral imaging has found applications in agriculture, medicine, environmental analysis, and quality control, to name only a few. Many applications of hyperspectral imaging can be found in remote sensing. Remote sensing hyperspectral imaging systems are space or airborne, and are expensive to deploy and almost impossible to control by the end user. In contrast, a labbased hyperspectral imaging system allows hyperspectral image acquisition with the required input parameters and under controlled conditions. A lab-based Hyperspectral Imaging System (HIS) has been set up in the Hyperspectral Optical Property Instrumentation (HOPI) Laboratory at Texas A&M University-Corpus Christi (Figure 1). This system has the capability to capture hyperspectral datacubes in the nm spectral region. The system consists of a hyperspectral imaging spectrograph attached to a 14-bit CCD camera with a fiber optic broadband light source as the illuminator. Although different light sources can be used, the broadband nature of the used light source allows the extraction of different wave bands through the spectrograph, and, thus, provides extended characterization of the analyzed samples. The significance of this system is its capability to capture 3D spectral and spatial data that can be analyzed to extract both conceptual as well as spectral information about the underlying samples. The hypercube image of a leaf sample obtained using the lab-based hyperspectral imaging system is shown in Figure 2. 1

8 Figure 1: Hyperspectral Imaging System in Reflection Mode Figure 2: Hyper Cube Image [Mehrubeoglu 2011] 2

9 1.1 Hyperspectral Imaging (HSI) Hyperspectral imaging (HSI) is a technique for acquiring 3D images called hyperspectral cubes or datacubes [Khoobehi 2004][Bahram 2004][Beach 2007]. Each hyperspectral cube has both spatial and spectral information and each pixel of the image frame has a spectral depth. The X-Y horizontal plane represents the spatial image and the vertical Z-axis direction provides wavelength-dependent spectral data (Figure 3). Figure 3 shows the spatial and spectral aspects of a hyperspectral image cube from a remote sensing system. Unlike single spectral information, HIS provides multi-dimensional spectral information for the imaged data, which can then be analyzed for localized (at each pixel/pixel group) optical properties such as absorption coefficient and scattering coefficient of the image pixels representing sample s properties at different wavelengths. Figure 3: a. Hypercube Image b. Spectral Signature from a Single Pixel Location [David 2009] 3

10 1.2 Goals and Objectives The main purpose of the hyperspectral imaging system is to obtain transmission and reflection mode measurements from samples to determine the optical and hyperspectral properties of different media. The hardware system is only capable of acquiring, storing, and viewing the hyperspectral image data. So there is a significant need to integrate the HIS with the above mentioned software tools to process data from the datacubes, and compute optical properties. It is not feasible to compute optical properties by hand for such massive amounts of data. The goal of this project is to create an integrated system that performs sample data analysis to calculate optical properties, such as scattering and absorption coefficients. In addition, other simulation capabilities have been added to determine reflection and transmission properties; heat produced when light travels into the sample; effects of radius of particles on generated heat; statistical computations of backscattered light vs. time. The objectives of the project are investigating and integrating related software tools, creating graphical user interface using LabVIEW programming, and testing and validation of the integrated system. 2. Implementing Different Analysis Tools for the Hyperspectral Data The Inverse Adding-Doubling program, which is currently being used for the double integrating spheres system to find the optical properties of the different samples at a single wavelength is modified for use in the hyperspectral imaging system. The Inverse Adding-Doubling program calculates the optical properties such as absorption and 4

11 scattering coefficients by using the observed measurements (diffuse reflection & transmission spectra) from the hyperspectral imaging system. The diffuse reflection and transmission spectra show the intensity of light received by the CCD camera in each mode after the light interacts with the samples under analysis and then is separated into different wave bands (Figure 4). Figure :4 Sample spectral graphs from a different pixel in the Hyperspectral ImageCube If the optical properties of the samples are known ahead of time, then using the tools for solving forward problems, reflection and transmission values can be calculated. This will be useful for calibrating the system with known standards. Such forward problems are solved through Monte Carlo simulations, Mie Theory and Adding-Doubling programs. 5

12 A GUI interface is created that encapsulates all the programs mentioned including the Inverse Adding-Doubling (IAD) method to find the unknown optical properties, Monte Carlo (MC) simulation, Mie Theory (MT) and Adding-Doubling (AD) method to find extrinsic properties with the known intrinsic optical properties. 2.1 Solving Inverse Problems The Inverse Adding-Doubling Method Inverse Adding-Doubling method (IAD) is used to generate the optical properties of light scattering and absorbing materials [Cheong 1990][Guang 2005][Scott 2010][Jian 2010][Lee 2010]. Through the IAD method, optical properties such as absorption coefficient and scattering coefficient can be calculated from measured total reflectance and total transmittance. The flexibility of the IAD method allows the determination of different number of parameters depending on the experiment, from among wavelength, measured reflectance, and measured transmittance. The equations used in Inverse Adding-Doubling method are summarized in Table 1. In this project, Prahl s IAD C- based program is interfaced with the HIS data through LabVIEW graphical programming [Scott 2010]. 2.2 Solving Forward Problems The Adding-Doubling Method The Adding-Doubling method is the solution of the radiative transport equation for anisotropic scattering and mismatched boundaries [Scott 1995]. This method is used to calculate the total reflection and total transmission of the sample. These values should be considered as important while measuring the optical properties of the sample. The 6

13 advantages of the Adding-Doubling method are that only integrations over angle are required to calculate total diffuse reflection and transmission values. Table 1: Selected Optical Properties that can be computed with the Inverse Adding- Doubling Method [Scott 2010] Interpretation of results can be made at each step, the method is equivalent for isotropic and anisotropic scattering, and the results are obtained for all angles of incidence used in the integration. The equations for Adding-Doubling method are summarized in the Table 2. The Adding-Doubling method is one of the best methods to determine the optical properties of turbid media [Cheong 1990][Scott 1995]. This method should be used such that the data processing with the radiative transport equation continues until the reflection 7

14 and transmission values matched those obtained experimentally. The advantage over existing methods is increased accuracy and flexibility in modeling turbid samples with intermediate albedos, mismatched boundary conditions, and anisotropic scattering. The disadvantage of this method is its entirely numerical nature. The Adding-Doubling method is equivalent for both the isotropic and anisotropic scattering [Cheong 1990][Scott 1995][Vladdmir 1996][Jianwei 2006]. Table 2: Selected Optical Properties for Adding-Doubling Method [Scott 1995] Monte Carlo Simulations Monte Carlo methods present a variety and an exhaustive solution for modeling of photon propagation through the simulation of photon transport. During modeling of photon transport when a scattering event occurs on a tissue, there is a deflection in the trajectory angle of the photon. In Monte Carlo method, the path of the photon movement subsequent to the scattering could be predicted as probability distributions derived as per the rules of photon transport. The modeling of this activity is determined from the 8

15 radiative transfer equation which ascertains the motion of photons using the differential equation. The multiple scattering events of light in turbid media are based on the radiative theory. Scattering is caused due to the presence of spherical particles in the birefringent medium; therefore, Mie Theory can be used to describe the scattering events. The birefringent medium is homogeneous, which means it has the same birefringent orientation and the birefringent value at two different locations and orientations in the medium. The transport path length (S) between two scattering events can be described by using the equation, S = -ln(ε)µ,.(1) Where ε is the random variable from 0 to 1 and the µ T is the sum of the absorption coefficient and the scattering coefficient. At each scattering level the photon will lose some energy and the lost energy can be represented using the equation, W = W * (µ a / µ T ),... (2) Where µa the absorption coefficient, W is the energy of the photon before scattering, W is the lost energy at each scattering level. The photon will have a selected but random polar angle θ and azimuthal angle Ø when the scattered event occurs. The angle Ø is positive for counter-clockwise. The Stokes vector of the photon packet in the new coordinate system can be represented by the equation, S = M (θ) R (Ø) S, (3) Where M (θ), R (Ø) are the two matrices, S represents the Stokes vector before scattering and S is the vector after scattering occurs. The θ and Ø values can be based on the probability distribution function. 9

16 Due to anisotropic background and refractive index around the sphere particles, there is a slight change in the polar angle and the azimuthal angle, which are neglected, and only the angles between the two scattering events are considered [Tatarskii 1998][Concannon 1995][Gang 2010]. is the double refraction of the considered medium between these two scattering events that can be seen with the linear retarder parameters. = ( n) ( / λ ), (4) n is the difference between the maximum and minimum refractive indices of the plane perpendicular to the propagation direction of the photon. λ is the wavelength of the light in the sample medium. The energy of the photon possesses after n scattering events in the birefringent medium can be calculated using the following formula, [µ s / (µ a + µ s ) ] n, (5) In Monte Carlo simulations large number of photons should be used to get the average detection signals. At the particular detection point on the surface of the birefringent medium, the Stokes vectors of the back scattered light will be the sum of all the photon packets that exited the medium from the particular point Time Monte Carlo Simulation Time Monte Carlo simulations simulate light propagation from a point source in an infinite medium with isotropic scattering [Scott 2007]. This simulation should be run to obtain the relation between the time and backscattered light when the light is incidentally shone onto the medium. The objective of the time resolved Monte Carlo technique is to simulate the propagation of polarized light in birefringent turbid media. 10

17 The calculated quantities include the reflection Mueller matrices, the transmission Mueller matrices, and the degree of polarization (DOP) [Xueding 2002]. They described the techniques to differentiate between weakly scattered and multiply scattered photons. The propagation of polarized light in turbid media is a complex process. Parameters, such as the size, shape, and density of the scatter as well as the polarization state of the incident light plays major role in the results. The number of scattering events is related to the optical path length and the time of propagation. A time-resolved study is needed to understand the evolution of polarization in turbid media. The direct tracing method was used in this simulation implementation. The Stokes vector and the local coordinates of each incident photon packet are traced statistically. At each scattering event, the incoming Stokes vector of the photon packet is first transformed into the scattering plane through a rotation operator and then converted by, S = M (θ) S,.(6) M is the single scattering Mueller matrix, given by the Mie Theory and S is the Stokes vector, S is the transformed Stokes vector. Time-resolved simulation is a useful tool for understanding the physical processes of polarization propagation in turbid media [Xueding 2002] Small Monte Carlo Simulation Small Monte Carlo simulation simulates light propagation from normal irradiation of a semi-infinite medium with anisotropic scattering [Scott 2007]. This simulation should be run to obtain the relation between the depth and heat produced due to scattering of the light through the medium. 11

18 Tiny Monte Carlo Simulation Tiny Monte Carlo simulation simulates the time resolved backscattering of a semi-infinite medium with anisotropic scattering. This simulation should run to obtain the relation between the radius and heat produced due to scattering of the light through the medium Mie Theory Mie Theory is used for the solution to the Maxwell s equation, which describes the scattering of electromagnetic radiation by a sphere [Lihong 2007][Mie 2008]. Presently, the term Mie solution is also used in broader contexts, for example when discussing solutions of Maxwell's equations for scattering by stratified spheres or by infinite cylinders, or generally when dealing with scattering problems solved using the exact Maxwell s equations in cases where one can write separate equations for the radial and angular dependence of solutions. The Mie solutions are in terms of infinite series and include calculation of scattering phase function, extinction, scattering, and absorption efficiencies, and other parameters such as asymmetry of parameter or radiation torque. Mie Theory and Rayleigh Theory are based on Maxwell s equations; model the scattering of a plane monochromatic optical wave by a single particle. If the particle size is greater than the wavelength, the wave is diffracted. The Rayleigh Theory is applicable only to particles that are smaller than the optical wavelength. Similarly Mie Theory is valid for the homogeneous isotropic spheres of any size. The Mie Theory reduces to Rayleigh Theory when the particle size is much smaller than the wavelength [Lihong 2007]. 12

19 Mie Theory computes the scattering of light by the particles when the particle size is less than the optical wavelength. Polar coordinates are used for the scattering of light. If the incident light propagates through the azimuthal (Z) axis, the scattered is located at the origin. Field point is located at (r, θ, Ø). The distribution of scatter light intensity for un-polarized light intensity is represented as, I(r, θ) = (((1 + cos 2 θ) K 4 α 2 ) / 2r 2 ) * I 0,.(7) Where α represents the polarization of the particle, I 0 denotes the incident light intensity, and K represents the propagation constant. θ denotes the polar angle and Ø denotes the azimuthal angle. N b denotes the refractive index of the background medium and λ represents the wavelength in vacuum. K= 2πn b / λ,...(8) Substituting the K value in the above equation should produce the relation I(r, θ) which is directly proportional to the 1 / λ 4. The short wavelengths have higher scatter than the long wavelength. For instance, blue light, which has a shorter wavelength, is scattered more than red light, which has a longer wavelength, by the same particles. σs is the scatter cross section, which is given by, σ s = 8πK 4 α 2 / 3,..(9) The polarization (α) of the sphere with radius a is represented as, α = ((n 2 rel 1) / ( n 2 rel +2) )* ( a 3 ),.(10) Where, n rel is the relative refractive index of the particle and is calculated as, n rel = n s / n b, (11) Where, n s and n b denote the refractive index of the sphere, the refractive index of the background respectively. 13

20 Substituting α in the above equation (9) results in, σ s = 8πa 2 x 4 / 3 n 2 rel -1/ n 2 rel+2 2,.(12) And the size parameter x is defined as the x= K*a. Substituting these values results in, σ s = a 6 /λ 4,.(13) From the above equations the scattering efficiency can be obtained, which is represented as, Q s = 8x 4 / 3 n 2 rel 1/ n 2 rel+2 2,...(14) The scattering efficiency is only dependent on x and n rel, where x is product of propagation constant and sphere radius. If n rel tends to unity, the above equation can be modified to obtain the following, Q s = 32x 4 / 27 n rel -1 2,.(15) 3. LabVIEW Programming LabVIEW is a graphics-based software suitable for developing the applications needed for the HIS and optical property measurements. The main advantage with the LabVIEW software is the ability to design applications without writing the logic for all the functions that are to be implemented. LabVIEW has predefined functions. The main challenge is to find the right functions from the function palettes for the desired application, and determining their proper sequence and connection to be compatible with the data types. LabVIEW is a graphical programming environment used by millions of engineers and scientists to develop sophisticated measurement, test, and control systems using graphical icons and wires that resemble a flowchart [LVM 2010]. It allows integration with many hardware devices and provides built-in libraries for advanced 14

21 analysis and data visualization for creating virtual instrumentation [LVM 2010]. The LabVIEW platform is independent of operating system versions. Some additional information from the listed reference is as follows [LVM 2010]: 1. It allows development of Virtual Instruments (VIs) through its programming environment. 2. It consists of an interactive user interface (front panel). 3. It requires programming through a dataflow diagram (block diagram) that serves as the source code. 4. SubVIs can be called by higher level VIs. Additional Features of LabVIEW programming can be listed as follows [LVM 2010]: 1. Easy GUI guided programming. 2. Built-in hardware integration facility. 3. Easy built-in analysis and signal processing. 4. Compatibility with different operating systems. 5. Compatibility with multiple programming languages. 6. Ability to program multiple cores. 3.1 Acquiring Spectral Data using the HIS and LabVIEW The HIS has the capability to acquire images that are over 3 GB in size, if the system is run at full resolution in both the spectral and spatial domains (811 x 1600 x 1200 x 14 bits/pixel) (See Figure 4 for a sample leaf image; left graph). The spectral information (diffuse reflectance or transmission) contains the intensity of light at the measured wavelength bands that can be saved in a text file and be exported using the LabVIEW software. The exported data can be saved into a notepad or similar file (in.rxt 15

22 format for this project). The spectral information at a given spatial pixel location in the hyperspectral image, shown in Figure 4 (middle), will be used to compute optical properties at measured wavelengths. The next section describes the LabVIEW design and implementation to use this data to compute optical properties. Therefore, the programs to measure optical properties must be adapted to accommodate this spectral data format, and LabVIEW designed to retrieve and output data accordingly. 3.2 LabVIEW Design and Implementation Figure 5 shows the LabVIEW Implemented Functionality Diagram for the Integrated Software Tool System. Figure 5: Labview Implemented Functionality Diagram for the Integrated Software Tool System 16

23 3.3 LabVIEW Design Interface Figure 6 shows the LabVIEW design interface for tools hierarchical diagram. This diagram depicts the hierarchy of the overall system functionality. The user can choose the kind of analysis to be performed using the front Design Interface. The user can choose to compute the Inverse Problems or the Forward Problems from the menu. Within the interface of Forward Problems the user can perform any of the Monte Carlo simulations, Adding-Doubling method or Mie Theory analysis. Figure 6: Tools Hierarchical Diagram 17

24 3.4 Software Engineering Techniques Requirements Software Tools and General Requirements The main components of this projects are adaptation of software tools such as Inverse Adding-Doubling method, Adding-Doubling method, Monte Carlo simulations such as Time Monte Carlo, Small Monte Carlo, Tiny Monte Carlo, and, finally, Mie Theory. In addition, the most significant contribution to this project is the Design of the interface to use the tools using LabVIEW programming. For this project, first suitable tools to find the optical properties of materials were identified. The existing method to compute the optical properties is Inverse Adding- Doubling method designed for the double integrating spheres system. In its existing state, this method is only capable of calculating the optical properties of a single wavelength at a time. But the hyperspectral imaging system has the capability to capture data at multiple wavelengths. So for this project, the IAD method was adapted to from the double integrating spheres application to iteratively compute the optical properties for multiple wavelengths during the same execution of the program. While the Inverse Adding-Doubling method is used to solve inverse problems to find unknown optical properties of any media by using the observed data from the HIS system, forward problems result in the computation of sample properties from unknown optical properties. The HIS requires Monte Carlo simulation tools, which are used to create the solutions by using the probability functions instead of solutions generated by using the integral equations. These simulations include Time Monte Carlo, Small Monte Carlo and Tiny Monte Carlo. The Adding-Doubling method is required to find the sample 18

25 properties such as reflection and transmission by normal or diffuse illumination of the light. The Mie Theory is required to calculate other sample properties such as anisotropy, extinction coefficient and scattering coefficient based on the scattering angles of the particles in the sample. For this project, the finished product is an application with six LabVIEW VI s. The user needs to run the usermenu.vi continuously and select one of the tools from all the available tools. Then the program directs the user to the selected tool. Now the user is able to give input parameters and run the tool. Based on the entered user input parameters, the output will be displayed as standard output values as well as in the graphs if applicable (if there are enough data points as a result). When the user runs the tool, the corresponding output files and text files for the graphs are created in the background. The user needs to click on the STOP button to stop executing the tool Functional and Non-Functional Requirements The functional requirements of the project are as follows: 1. The Inverse Adding-Doubling tool is used to compute the optical properties of the sample such as absorption coefficient and scattering coefficient. The user must identify the input as a text file with the reflection and transmission values and the corresponding wavelengths. 2. Time Monte Carlo simulation is used to simulate the light propagation in anisotropic medium and it provides the simulation between times vs. back scattered light. It requires the user to input parameters such as scattering coefficient, absorption coefficient, anisotropy factor, and refractive index. 19

26 3. Small Monte Carlo is used to simulate light propagation in anisotropic medium and it results in the simulation of depth vs. heat. This tool requires the user to input the parameters such as scattering coefficient, absorption coefficient, anisotropy factor, and refractive index. 4. Tiny Monte Carlo is used to simulate light propagation in isotropic medium and it provides the simulation between radiuses vs. heat. It requires user to input the parameters such as scattering coefficient and absorption coefficient. 5. The Adding-Doubling method is used to find sample properties such as total reflection for normal illumination, total reflection for diffuse illumination, total transmission for normal illumination and total transmission for diffuse illumination. It requires user to input the parameters such albedo, optical thickness, refractive index and anisotropy factor. 6. Mie Theory is used to calculate the sample properties such as anisotropy, scattering coefficient and extinction coefficient based on the angles of the scattering particles, it requires user to give input as a rich text formatted (.rxt) file with radius, wavelength, real index, imaginary index, density, and number of angles. 7. The requirement for this project needs Lab-VIEW software. The software must be integrated into the HIS interface and allow independent individual access to each of the tools. Among many non-functional requirements, the project must be very userfriendly. The software tools will very likely be used by people with little more experience than using a computer for everyday tasks. 20

27 Structured Requirements For the integration of software tools, the following must be considered: The interface should allow user to select the tool to compute the optical and sample properties of different media. The user needs to 1. locate the directory for the Usermenu.vi on the computer. 2. run the Usermenu.vi. 3. continuously run the Usermenu.vi program by clicking the RUN CONTINUOSLY button right next to the arrow mark in the left corner of the front panel. 4. select the tool using tool palette, so the tool palette is available in VIEW: Click on VIEW and select the tool palette from the options available. 5. click on the HAND mark button in the tool palette to select the tool from the user menu. 6. select one of the tool and click on OK to run the tool. This will direct the user to the associate tool. In order to run the selected tool and allow the selected tool to process the input/output based on the developed design, the user needs to 1. again click on the HAND mark tool to provide the input values. 2. enter the input parameters such as values or text files as input in corresponding text boxes or corresponding place holders. 3. click on the ARROW mark which is available in the top left corner of the front panel. 21

28 Within the tools, 4. the LabVIEW block diagram will find the corresponding VI in the background and run the associated program. 5. the output text file is created in the folder(the directory where the corresponding background VI run s) if necessary for the tool; otherwise the output will be displayed to the standard output. 6. the graphs will be displayed to the output screen. 3.5 Analysis The requirements for this project are to compute the optical properties of hyperspectral images. These images are obtained from the instrumentation in the lab. In order to fulfill this requirement we have to assess the pathway of receiving the input from the instrumentation into our proposed system. For Inverse Adding-Doubling method and Mie Theory the input is refined manually from the instrumentation and is made compatible with the code in the proposed system. The input received comprises of two independent files which are merged into text files which serve as the input into the code into our proposed system. For Monte Carlo and Adding-Doubling requirements, the input to the proposed system is in form of manually entered values. The input is translated further into the C code in the system and the optical properties are computed and presented to the user in a Lab view interface. Lab view is used owing to its superior use in graphical understanding of the data. 22

29 3.6 Design Figure 7 shows the various interactions of the user with the proposed system design. The user of the system can execute the Inverse Adding-Doubling methods by inputting a text and viewing the results in graphical format. Similarly Monte Carlo simulations are done when the user inputs parameters and views the results in graphical formats. Likewise input to Mie Theory and Adding-Doubling are provided by the user and the results are to be viewed in graphical format. 23

30 System Input text file for inverse adding-doubling method View output text file, Display graphs and results Input parameters for monte carlo simulations View results and graphs for monte carlo simulations User Input parameters for adding-doubling method caluclations View results to the standard output for adding-doubling method caluclations Input text file for mie theory View results to the output standard for Mie Theory Figure 7: Use case Diagram for Integrated Software Tools 24

31 USER SYSTEM 1: Select IAD method from user menu 2: Enter text file as input 3: Process input text file and write the optical properties results in the output text file 4: Generate graphs and results 5: View results on output screen Figure 8: Sequence diagram for IAD Method The Figure 8 displays a scenario of user interacting with the system for executing the Inverse Adding-Doubling method. First, the user selects the IAD method from the menu interface. Input to the system is given in the format of a text file. The system then, executes the properties and generates the graph and results to the user, onto the output screen. 25

32 USER SYSTEM 1: Select Time or Small or Tiny monte carlo simulation from the menu 2: Enter the input parameters 3: process the input data values 4: Display graphs and results 5: View results on output screen Figure 9: Sequence Diagram for Monte Carlo Simulations The Figure 9 displays a scenario of user interacting with the system for executing the Monte Carlo simulations such as Time or Small or Tiny Monte Carlo. First, the user selects the IAD method from the menu interface. Input to the system is given in the format of value parameters. The system then, executes the properties and generates the graph and results to the user, onto the output screen. 26

33 USER 1: Select Adding-Doubling method from the menu SYSTEM 2: Enter the input parameters 3: process the input data values 5: View results on output screen Figure 10: Sequence Diagram for Adding-Doubling Method Figure 10 displays a scenario of user interacting with the system for executing the Adding-Doubling method. First, the user selects the Adding-Doubling method from the menu interface. Input to the system is given in the format of value parameters. The system then, executes the properties and generates results to the user, onto the output screen. 27

34 USER 1: Select Mie Theory from the menu SYSTEM 2: Enter the text file as input 3: process the input data values 5: View results on output screen Figure 11: Sequence Diagram for Mie Theory Figure 11 displays a scenario of user interacting with the system for executing the Mie Theory. First, the user selects the Mie Theory from the menu interface. Input to the system is given in the format of a text file. The system then, executes the properties and generates results to the user, onto the output screen. 28

35 4. Results, Testing and Validation The integrated system user interface provides all the programs listed in the abstract that allow solving forward problems (Monte Carlo, Mie Theory and Adding-Doubling programs) and inverse problems (Inverse Adding-Doubling program). User can select one program from the list and give the input parameters to the program. The output is then displayed on the screen. 4.1 User Interface Figure 12 shows the front panel of the user interface. This is the first interface the user encounters where he/she has to select the tool of interest to run. Figure 12: User Interface Front Panel 29

36 Figure 13 shows the underlying block diagram for the user interface in Figure

37 Figure 13: User Interface Block Diagram 4.2 Inverse Adding-Doubling Method When the user selects the Inverse Adding-Doubling method, the text file (.rxt) name is entered as the input to the program. The input file consists of information about the wavelength, measured transmittance and measured diffuse reflectance of the sample at each pixel. When the user runs the program, output text file is created with the same file name but the extension for the output file is.txt. The output data file contains the estimated absorption coefficient and reduced scattering coefficient of the sample at different wavelengths for the given pixel. The program also creates the two excel (.xls) files; one contains the data about the wavelength and absorption coefficient and the second contains the data about the wavelength and reduced scattering coefficient. The interface also provides the option to select the graphs, where one graph displays the results as wavelength on the x-axis and absorption coefficient on the y-axis. The second graph is to display the results as wavelength on the x-axis and reduced scattering coefficient on they-axis. The front panel, front panel for input, input file, output file, front panel for output and the block diagram for the Inverse Adding-Doubling program are shown in Figures

38 Figure 14 is the interface after the user clicks on the Inverse Adding-Doubling method icon on the main menu for Inverse Adding-Doubling method. Figure 14: Inverse Adding-Doubling Method Front Panel 32

39 Figure 15 shows the underlying block diagram for the Inverse Adding-Doubling method in Figure 14. Figure 15: Inverse Adding-Doubling Method Block Diagram 33

40 Figure 16 is the Inverse Adding- Doubling method interface after the user provides/enter the text file name. Figure 16: Inverse Adding- Doubling Method Front Panel (input) 34

41 Figure 17 is the input file and it also shows how the user needs to provide the data and format of the text file for Inverse Adding-Doubling method. Figure 17: Inverse Adding-Doubling Method (text input file) 35

42 Figure 18 shows the output results and graphs for the user input in Figure 17. Figure 18: Inverse Adding- Doubling Method Front Panel (output) 36

43 Figure 19 is the output file generated for the user in the same directory where the user provided the input file, but the output file is the same name as input file (.rxt) but with the different format (.txt). Figure 19: Inverse Adding-Doubling Method (output data file) 4.3 Monte Carlo Simulation Methods The Monte Carlo simulation method has three sub-programs: 1. Time Mote Carlo Simulation 2. Small Monte Carlo Simulation 3. Tiny Monte Carlo simulations 37

44 4.3.1 Time Monte Carlo Simulation When the user selects the Time Monte Carlo simulation program, the user needs to input the parameters such as absorption coefficient, scattering coefficient, anisotropy and refractive index. The output will be displayed to the standard output screen. A timemc.xls file will also be created, which is used to display the resulting graph between the time and backscattered light when the user selects to display the time Monte Carlo graph. The front panel, front panel for input, front panel for output, and the block diagram for the Time Monte Carlo simulation program are shown in Figures Figure 20 is the interface after the user clicks on the Time Monte Carlo simulation icon on the main menu for Time Monte Carlo simulation. Figure 20: Time Monte Carlo Simulation Front Panel 38

45 Figure 21 shows the underlying block diagram for Time Monte Carlo simulation in Figure 20. Figure 21: Time Monte Carlo Simulation Block Diagram Figure 22 is the Time Monte Carlo simulation interface after the user provides the input parameters. 39

46 Figure 22: Time Monte Carlo Simulation Front Panel (input) Figure 23 shows the output results and graphs for the user input in Figure 22. Figure 23: Time Monte Carlo Simulation Front Panel (output) Small Monte Carlo Simulation When the user selects the Small Monte Carlo simulation program, the user needs to input the parameters such as absorption coefficient, scattering coefficient, anisotropy factor, and refractive index. The output is displayed as standard output values on the screen, and the small-mc.xls file is created. The small-mc.xls file is used to display the resulting graph displaying the relation between depth and heat when the user selects to display the small Monte Carlo graph. The front panel, front panel for input, front panel for output and the block diagram for the small Monte Carlo simulation method are shown in Figures

47 Figure 24 is the interface after the user clicks on the Small Monte Carlo simulation icon on the main menu for Small Monte Carlo simulation. Figure 24: Small Monte Carlo Simulation Front Panel 41

48 Figure 25 shows the underlying block diagram for Small Monte Carlo simulation in Figure 24. Figure 25: Small Monte Carlo Simulation Block Diagram Figure 26 is the Small Monte Carlo simulation interface after the user provides the input parameters. Figure 26: Small Monte Carlo Simulation Front Panel (input) 42

49 Figure 27 shows the output results and graphs for the user input in Figure 26. Figure 27: Small Monte Carlo Simulation Front Panel (output) Tiny Monte Carlo Simulation When the user selects the Tiny Monte Carlo simulation program, the user needs to input the parameters such as absorption coefficient, scattering coefficient. The output will be displayed on the output screen. The small-mc.xls file is created, which is used to display the resulting graph showing the relation between time and backscattered light when the user selects to display the Small Monte Carlo graph. The front panel, front panel for input, front panel for output, and the block diagram for the Tiny Monte Carlo simulation method are shown in Figures

50 Figure 28 is the interface after the user clicks on the Tiny Monte Carlo simulation icon on the main menu for Tiny Monte Carlo simulation. Figure 28: Tiny Monte Carlo Simulation Front Panel 44

51 Figure 29 shows the underlying block diagram for Tiny Monte Carlo simulation in Figure 28. Figure 29: Tiny Monte Carlo Simulation Block Diagram Figure 30 is the Tiny Monte Carlo simulation interface after the user provides the input parameters. 45

52 Figure 30: Tiny Monte Carlo Simulation Front Panel (input) Figure 31 shows the output results and graphs for the user input in Figure 30. Figure 31: Tiny Monte Carlo Simulation Front Panel (output) 4.4 Adding-Doubling Method When the user selects the Adding-Doubling method, he/she needs to enter the input parameters such as albedo, optical thickness, anisotropy factor, and refractive index. The output is then displayed as standard output values on the screen. The output parameters displayed by the Adding-Doubling method include total reflection for normal illumination, total transmission for normal illumination, total reflection for diffuse illumination, and total transmission for diffuse illumination. The front panel, front panel for input, front panel for output, and the block diagram for the Adding-Doubling method are shown in Figures

53 Figure 32 is the interface after the user clicks on the Adding-Doubling method icon on the main menu for Adding-Doubling method. Figure 32: Adding-Doubling Method Front Panel 47

54 Figure 33 shows the underlying block diagram for Adding-Doubling method in Figure 32. Figure 33: Adding-Doubling Method Block Diagram 48

55 Figure 34 is the Adding-Doubling method interface after the user provides the input parameters. Figure 34: Adding-Doubling Method Front Panel (input) 49

56 Figure 35 shows the output results and graphs for the user input in Figure 34. Figure 35: Adding-Doubling Method Front Panel (output) 4.5 Mie theory: When the user selects Mie Theory, he/she needs to enter the input file name (e.g., mie.rxt). The input parameters in the file include the radius, wavelength, real index, imaginary index, density and the number of angles. The angles represent equally spaced angles between 0 and 360 degrees. The output will be displayed on the screen. The output parameters displayed by the program include anisotropy factor, extinction coefficient, scattering coefficient along with the input parameters specified by the user in the input file. The front panel, front panel input, front panel output, and the block diagram of the Mie Theory tool are shown in Figures

57 Figure 36 is the interface after the user clicks on the Mie Theory icon on the main menu for Mie theory. Figure 36: Mie Theory Front Panel 51

58 Figure 37 is the input file and it also shows how the user needs to provide the data and format of the input text file for Mie Theory. Figure: 37 Mie Theory (text input file) 52

59 Figure 38 shows the underlying block diagram for Mie Theory in Figure 36. Figure 38: Mie Theory Block Diagram 53

60 Figure 39 is the Mie Theory interface after the user provides/enters the text file name. Figure 39: Mie Theory Front Panel (input) Figure 40 shows the output results and graphs for the user input in Figure 39. Figure 40: Mie Theory Front Panel (output) 54

61 5. Conclusions Two sets of algorithms have been adapted and implemented in LabVIEW for use with the data from the hyperspectral imaging system. These algorithms solve inverse problems (Inverse Adding-Doubling program) and forward problems (Adding-Doubling program, Monte Carlo simulations, Mie Theory) to determine both intrinsic and extrinsic optical properties of analyzed materials. Originally, the HIS was only capable of acquiring and viewing hyperspectral images. With these algorithms, which have been integrated into the hyperspectral imaging systems environment as an added capability, individual spectra associated with a selected pixel can be analyzed, and intrinsic optical properties, such as absorption coefficient and scattering coefficient (Inverse Adding- Doubling method) can be computed from hyperspectral data collected in reflection and transmission mode. Similarly extrinsic optical properties such as reflectance and transmission characteristics (Adding-Doubling method) can be determined from scattering coefficient, absorption coefficient, anisotropy factor, and refractive index. Monte Carlo programs can be used to compute other extrinsic properties, such as depth vs. heat, or particle radius vs. backscattered light, or radius vs. heat, from known intrinsic parameters. Mie Theory is useful to compute the extinction coefficient, anisotropy factor, from density and angle measurements (number of angles). As brand new capabilities for the analysis of hyperspectral image data, these programs will be used to determine not only the optical properties of materials, but also monitor changes associated with these optical properties, which, in turn, will be useful for applications such as environmental monitoring, detection of cancer cell state and response to treatment, and other applications in sciences and engineering. The original codes, which have been written in 55

62 C, have been integrated into the LabVIEW interface, the main interface with the hardware. The programs have been tested with the hyperspectral data and previously published properties for testing and validation. The results showed the successful execution of the programs with expected outcomes; therefore, the LabVIEW-based tool is now ready to be commissioned in the lab as the added capability of the system. 6. Future Work Future research will focus on additional automation of pixel/spectrum selection and optical property computations which will increase the efficiency of data analysis; currently, the user must select the pixel spectrum to be analyzed; then, the reflectance and transmission measurements associated with this pixel must be manually written to a file to be read by the Inverse Adding-Doubling program. A remote access to the hardware has been developed by another student. Future work will also include combining the work presented here and the remote accesses interface such that optical properties can be computed remotely. Automation of data writing from the click of a pixel so that the data can be easily read by the integrated programs will be an added benefit to the HOPI project. A third integration effort is necessary to merge the data obtained from two independent optical systems, namely, the hyperspectral imaging system described here, and the double integrating spheres system which is also set up in the lab for data collection. These two systems are currently independent, but obtain similar features from samples. Merging the capabilities of these two systems will result in a more robust system where the results can be compared and verified independently; therefore, the future work includes integrating the designed system with the existing instrumentation of 56