Austalian Jounal of Basic and Applied Sciences, 5(1): 80-85, 011 ISSN 1991-8178 An Assessment of the Efficiency of Close-Range Photogammety fo Developing a Photo-Based Scanning Systeminthe Shams Tabizi Minaet in Khoy City 1 Reza DavaiMajd, Akba Jafai 1, Depatment of Geomatic Engineeing, Khoy Banch, Islamic Azad Univesity, Khoy, Ian. Abstact: Acquiing 3D data and modeling the suface of objects has attacted the attention of a boad ange of uses and eseaches due to thei numeous applications. Nowadays, econstucting 3D sufaces is one of the impotant topics in machine vision due to its vaious applications in industy and cultual heitage. Thus, many eseach studies have been caied out on suface econstuction andextaction of 3D data fom the suface of objectsand ae efeed to as Shape fom X, whee X can be steeo images, stuctued light, textue, focus, shading, eflection, o silhouette (Salvi, J., et al., 004).One of the common methods fo measuing 3D sufaces is close-ange photogammety whee accuate, dense point clouds ae extacted and the suface of inteest is econstucted using advanced coelation-based image scanne/matche. Key wods: Lase scanne, photo-based 3D scanning, coelation matching. INTRODUCTION These days, with the development of boadband intenet and the ability of gaphical display on compute sceens, digital3d models can easily be put on view.display of vitual museums, histoical castles, and antiquities ae of the most impotant applications of digital models.lase scanning and photo-based scanning ae not alwayssufficient fo developing digital models, especially in tems of geometic accuacy, photoealism, potability, and cost. Thus, a photo-based scanning systemwas used and put to test in the pesent eseach. MATERIALS AND METHODS Calibation of Sony DCR 0 The field test was used accoding to the following figue to calibate the camea used in the eseach. This field test consists of 90 tagets of which 4 tagets wee designed withalgoithms fo automatic detection.on-thejob calibation was used because of the non-meticdigital camea and lack of geometic stength. The following figue was pinted on an A0 sheet and was installed on a flat gound at the site of the minaet. Fig. 1: The calibation netwok The imaging stations wee chosen based on figue and fou photos wee captued fom each of these stations. Since the tagets of the field test do not yet have any 3D coodinates, the cameas wee calibated without contol points as a fee netwok using the self-calibation method. Using this method, calibation paametes, exteiooientation of the camea, and the 3D coodinates of the points wee computed simultaneously. A elationship is established between image space and object space fo data pocessing and detemining the calibation paametes using equation 1 (Atkinson, K.B., 1998; Rocchini, C., et al., 001). Coesponding Autho: Reza Davai Majd, Khoy Banch, Islamic Azad Univesity, Khoy, Ian. 80
Aust. J. Basic & Appl. Sci., 5(1): 80-85, 011 Fig. : Imagingstations in the field test. m ( X Xc) m ( Y Yc) m ( Z Zc) ( ) ( ) ( ) 11 1 13 x x c m 31 X Xc m 3 Y Yc m 33 Z Zc m ( X Xc) m ( Y Yc) m ( Z Zc) ( ) ( ) ( ) 1 3 y y c m 31 X Xc m 3 Y Yc m 33 Z Zc x y ( x ( x ( y x x P [ 1 j y P [ j ) ( y ) ( x 1 ) 1 ( K ( K ( y y 3 1 j x 3 1 j y K K ) 5 j ) ] P 5 j K j ( x K 1 j ) ] P ( x 7 3 j x 7 3 j ) x ) )( y )( y y y ) B 1 j ) B j x B x j y (1) In the above equation, and denote calibation paametes and detemining the type and numbe of these paametes depend on the geometic stength of the camea. Typically, the most impotant calibation paametes ae adial and tangential distotions inthe lens as well as shift of images points, focal length, and affine paametes. Fig. 3: Anoveall view of Sony DCR 0 camea. Afte pocessing, the following data wee computed: Fig. 4: Sony DCR 0 camea calibation esults. 81
Aust. J. Basic & Appl. Sci., 5(1): 80-85, 011 Fig. 5: Calibation accuacy of Sony DCR 0 camea. The calculated calibation eo is 1.5 pixels (each pixel is equal to one micon) and consideing the focal length and the appoximate 3-mete distance of photos fom taget sheets, the image scale is 1.1000 and this value ceates a 1.5 mm eo on the gound. Designing a Photogammetic Netwok aound the Minaet and Setting the Tagets: To achieve acceptable esults in detemining the paametes, fist a peliminay analysis has to be done to detemine the appoximate locations fom which the camea has to take photogaphs and this pocess is efeed to as netwok design. The diving factos in netwok design involve factos affecting the distance between the camea and the object and factos affecting visibility, and these factos ae explicated below (Atkinson, K.B., 1998). Factos Affecting the Camea-Object Distance: The factos affecting the distance between the camea and the object ae again of two types factos that pevent the camea fom movingneae to the object which ae depth of field (DOP) and the numbe of image points, and factos that pevent the camea fom moving fathe fom the object which ae imageesolution, camea s field of view (FOV), image scale, and woking space constaint. Each of these factos will be explained below. The minimum and maximum distance is obtained by equations and 3. () (3) Depth of Field The depth of view of a camea depends on the F-numbe (D), pixel dimensions (c), focal length (f), and the distance fom which the camea has a good focus (p). Equation 4 is used to detemine the distance between the camea and the object as a esult of the depth of field. (4) Image Scale Equation 4.3 is used to detemine image scale whee a geate scale is used so that the points will be eadily visible on the image. (5)
Aust. J. Basic & Appl. Sci., 5(1): 80-85, 011 whee is the numbe of image obsevations, is netwok stength, is the main distance, and is the accuacy of image points. Afte examining the above factos, image scale and image distance wee computed 1.1000 and 3-4 metes espectively. The imaging netwok was designed accoding to the figue below whee imaging stations wee placed at a distance of 1 mete to ceate paallel (steeo) photos and thee was a 1.30 mete distance between each pai of photos. Fo convenience, imaging stations wee set up on the gound at 4-mete distances with the minaet. The photos wee taken in both convegent and paallel modes whee convegent photos wee used fo solving oientations and paallel photos wee used fo geneating point clouds and the suface textue of the minaet. Fig. 6: A schematic epesentation of the imaging netwok set up aound the Shams Tabizi Minaet. Fo accuate and quick oientationof photos and attaching the models togethe, some tagets wee used as displayed in the following figue. Consideing the gand size of the Shams Tabizi Minaet located in Khoy City, modeling stages wee divided into seveal smalle pats and eventually, the final model was obtained by combining the ceated models. Tagets wee attached aound the minaet with 1.30 mete distance and with metes heightfom each othe. Fig. 7: An example of the tagets used in the minaet. In the photogaphy, it is tied to make 4 taget sheets visible in the fist pefeence in each image as displayed in the above figue i.e. thee should be 56 tagets. Tagets wee coodinated using atotal Station systemin ode to geoefeence the models and ceate a single system of coodinates in all the models and attaching them togethe. The accuacy of the gound coodinates is 0.5 mm. Results: Data Pocessing: Afte taking the photogaphs, the following steps wee taken using the Photomodele Scanne softwae: Intoducing the computed calibation paametes. Oientation of the photos using the taget and contol points with an accuacy of one pixel (at least 3 points wee used in each model). Identifying the ange of point clouds in each photo. Choosing the photos with pope B/H atio (the distance between two imaging stations with espect to the distance fom the minaet, that is 0.5-0.5 (Photometic Pty Ltd, Eos Systems, site addess: http://www.photomodele.com). 83
Aust. J. Basic & Appl. Sci., 5(1): 80-85, 011 Detemining the sampling distance fo ceating points with diffeent distibutions. Detemining the appopiate seach window dimensions fo coelation matching and geneating point clouds fom photo pais. Removing possible point cloud noises. Tiangulation of points and ceating a continuous suface with ealistic-textueof points. Recalling all the models fo geneating the geneal model. Fig. 8: An example of two photos captued fo ceating point clouds with 0.5 B/H. Fig. 9: Examples of a wie-fame model and a photoealistic -textue model with sampling distance of 10 mm of a pat of the minaet. To examine the accuacy of the geneated model, 10 contol points wee consideed as checkpoints and by compaing the eal gound coodinates and the coodinates extacted fom the model, the RMS of checkpoints wee computed as 1.5 mm. Conclusion: The pesent aticle ties to test a photo-based 3D scanningsystem on an antiquity and to evaluate its accuacy, and to demonstate that this system simila to lase scanning systems can be applied in most 84
Aust. J. Basic & Appl. Sci., 5(1): 80-85, 011 pects fo documenting antiquities and modeling uban funitue. Moeove, by choosing a high-esolution digital camea and taking paallel o nealypaallel photos, one can geneate point clouds with ealistic textue. REFERENCES Atkinson, K.B., 1998. Close ange photogammety and machine vision. Whittles Publishing, Scotland, pp: 384. Guhing, J., C. Bemme, J. Bohm and D. Fitsch, 000. Data pocessing and calibation of coss-patten stipe pecto. IAPRS, Amstedam, 33(5). Paa, G., F. Rottensteine and W. Pölzleitne, 001. Image Matching Stategies. In: Kopatsch W.G. and Bischof H. (Eds.), Digital Image Analysis. Selected Techniques and Applications. Spinge-Velag New Yok, ISBN 0-387-95066-4., pp: 393-410. Photometic Pty Ltd, Eos Systems, site addess: http://www.photomodele.com. Rocchini, C., P. Cignoni, C. Montani and P. Pingi, 001. A low cost 3D scanne based on stuctued light.euogaphics, 0(3). Salvi, J., J. Pag'es and J. Batlle, 004. Patten codification stategies in stuctued light systems. Patten Recognition, 37(4): 87-849. 85