28 2 21 3 JOURNAL OF APPLIED SCIENCES Electronics and Information Engineering Vol. 28 No. 2 Mar. 21 255-8297(21)2-136-6 4379. 1... TP751 P231.5 A Detection and Compensation of Shadows in High Resolution Remote Sensing Images Using PCA WANG Yue, WANG Shu-gen School of Remote Sensing and Information Engineering, Wuhan University, Wuhan 4379, China Abstract: From the shadow region properties, a new algorithm of shadow detection and compensation based on principal component analysis (PCA) is proposed for high resolution remote sensing images. The ratio between the original blue band and the first principal component is computed. Shadow and non-shadow regions are segmented using the histogram-thresholding method. A morphological algorithm is used to produce an accurate shadow mask, and a dilation operation is performed to find the surrounding homogeneous region of a shadow. Each shadow region is matched to its adjacent homogeneous region to achieve shadow compensation. Finally, image smoothing and inverse PCA transformation are performed on the compensated image. Experimental results show that the algorithm is effective. Keywords: remote sensing, principal component analysis, shadow detection, shadow compensation.. DSM(digital surface model) [1-2]... Susuki [3]. Pooya Sarabandi C1C2C3 [4]. IKONOS 2 (G) 3 (R) 4 (NIR) RGB HIS HIS I IKONOS 21-1-29 21-3-1 E-mail: Redflower_yue@126.com E-mail: wangsg@whu.edu.cn
2 137 [5]. [6].. [7]. [8-9].. (PCA) PCA HIS C1C2C3.. RGB 1.. 1 1.1 Hotelling. [1-11]. R G B 3. RGB 1) RGB M N 3 x = [x R x G x B ] T. RGB 3 3 C x (1) C x = 1 K 1 K (x k m x )(x k m x ) T (1) k=1 m x x k K = MN. 2) e i λ i (i = 1, 2,, MN) C x λ 1 > λ 2 > > λ MN, e 1, e 2,, e MN. C x e 11 e 1,MN A =.. (2) e MN,1 e MN,MN 3) (3) y = A(x m x ) (3) y = [y 1, y 2, y 3 ] T, 3. A y 1 9%. λ 1 y 1 1. 1.2 [7]. R G B. C1C2C3 RGB. C1C2C3. [3] C1 C2 C3 C3 C1 C2. C3 (4) B C3 = arctan( max{r, G} ) (4)
138 28 C3 R B B 1 y 1 R B. B 1 y 1. B 1 y 1. 1 (b) B 1 y 1. (a). C3 (e) B/y 1 (d). 1-4 2.5 2. 1.5 1..5 (a) (a) Original image (c) C3 (c) C3 image 2. 1.5 1..5 (b) B/y 1 (b) B/y 1 ratio image 1-4 5 1 15 2 25 (d) B/y 1 (d) B/y 1 histogram 5 1 15 2 25 (e) C3 (f) B/y 1 (e) C3 histograme (f) Detected shadow on B/y 1 image (g) C3 (h) (g) Detected shadow on C3 image (h) Division image of shadow orientation W 4 5 3 N S 2 1 8 6 7 1 B/y 1 C3 Figure 1 Comparison results of shadow detection between on B/y 1 image and on C3 image E C3 B/y 1 Ostu. C3 B/y 1 C3 B/y 1. G R B C3 B/y 1 B/y 1. 1. 8-.. 2 T. T T. 3. 2. [3].. [12].. 2.1.. se
2 139.. 8 1(h) 1(a) 1. 3. 1(f) 1 1 1 s = 1 1 1 (5)... 2.2 1 m shadow, m nonshadow, I = A mnonshadow m shadow I shadow (6) A.6 1; I shadow I... 1 2 3. 1 2 3 RGB.. 3 MATLAB R29a. 2 3 1 2 3. 2(a) 1 1 18 1 64,.25 m. 2(b). T 1 A.8. 2 (d) 2 2 777 3 17,.25 m,. T 1 A.8. 2(g) 3 quickbird 2 61 2 46,.6 m. 2(h). T 5 A.7. (η DR ) (η FR ). η DR η FR η FR = η DR = N TP N TP + N FP 1% N FN N TP + N FN + N FP 1% (7) N TP (true positive) N FN (false negative) N FP (false positive). 1.
14 28 (a) (b) (c) (a) Original image (b) Result image of shadow detection (c) Result image of shadow compensation (d) (e) (f) (d) Original image (e) Result image of shadow detection (f) Result image of shadow compensation (g) (h) (i) (g) Original image (h) Result image of shadow detection (i) Result image of shadow compensation 2 Figure 2 Final results of shadow detection and compensation 1 Table 1 Shadow detection accuracy measurments N TP N FN N FP N DR/% N FR/% (a) 125 12 14 89.93 7.94 (d) 332 43 28 92.22 1.67 (g) 15 1 13 88.98 7.81. 1 3 1 3 2(a) 2(d) 2(g) 4 6.. 1. 2(f) 2(h)..9.8.7.6.5.4.3.2.1 5.3.25.2.15.1.5.3.25.2.15.1.5 5 5 5 1 15 2 25 3 5 1 15 2 25 3 5 1 15 2 25 3
2 141.6.5.4.3.2.1-5 5 1 15 2 25 3.7.6.5.4.3.2.1 5 5 1 15 2 25 3.1.9.8.7.6.5.4.3.2.1 5 5 1 15 2 25 3 3 Figure 3 Comparison between histograms of 4 shadow regions and results after shadow compensation.... : [1] Dare P M. Shadow analysis in high-resolution satellite imagery of urban areas[j]. Photpgrammetric Engineering and Remote Sensing, 25, 71: 169-177. [2] Young S D, Kakarlapudi S, Uijt de haag M. A shadow detection and extraction algorithm using digital elevation models and x-band weather radar measurements[j]. International Journal of Remote Sensing, 25, 26: 1531-1549. [3] Susuki A, Shio A, Arai H, Ohtsuka S. Dynamic shadow compensation of aerial images based on color and spatial analysis[c]//barcelona, Catalonia, Spain: Proceedings of the 15th International Conference on Patten Recognition, 2, 1 317-32. [4] Sarabandi P, Yamazaki F, Matsuokav M, Kiremidjian A. Shadow detection and radiometric restoration in satellite high resolution images[c]//proceedings of the IEEE International Geoscience and Remote sensing symposium, 24, 6 3733-3747. [5]. IKONOS [J]. 24, (4): 29-32. Huang hao, Zhang youjing, Ma xuemei. A shadow automated extraction method of IKONOS images based on image fusion[j]. Remote Sensing Information, 24, (4): 29-32. (in Chinese) [6]. [J]. 28, 41(7): 8-84. He Kai, Zhao Hongying, Liu Jingjing Wang Chengyou. Shadow removal of city s remote sensing image based on fractal and texture analysis[j]. Journal of Tianjin University, 28, 41(7): 8-84. (in Chinese) [7]. [J]. 23, 28(5): 514-516. Wang Shugen, Guo Zejin, Li Deren. Shadow compensation of color aerial images[j]. Geomatics and Information Science of Wuhan Uinversity, 23, 28(5): 514-516. (in Chinese) [8]. QuickBird [J]. 28, 44(7): 3-32. Zheng Qianying, Wang Qinmin. A shadow restoration method of QuickBird satellite imagery[j]. Computer Engineering and Applications, 28, 44(17): 3-32. (in Chinese) [9] Chen Y, Wen D, Jing L, Shi P. Shadow information recovery in urban areas from very high resolution satellite imagery[j]. International Journal of Remote Sensing, 27, 28(15): 3249-3254. [1]. [J]. 24, 2(7): 8-13. Dong dongxia. A method of determining the color image scluster number based on prinicipal component transform[j]. Microcomputer Applications, 24, 2(7): 8-13. (in Chinese) [11]. IKONOS [J]. 28, 33(9): 947-95. Li Cunjun, Liu Liangyun, Wang Jihua, Zhao Chunjiang. Ikonos shadow extraction in urban region based on principal component fusion information distortion[j]. Geomatics and Information Science of Wuhan Uinversity, 28, 33(9): 947-95. (in Chinese) [12] Tsai V J D. A comparative study on shadow compensation of color aerial images in invariant color models[j]. IEEE Transactions on Geoscience and Remote Sensing, 26, 44(6): 1661-1667. ( : )