Adaptive digital technique for discriminating between shadow and water bodies in the high resolution satellite imagery

This research presents a new algorithm for classification theshadow and water bodies for high-resolution satellite images (4-meter) of Baghdad city, have been modulated the equations of thecolor space components C1-C2-C3. Have been using the color spacecomponent C3 (blue) for discriminating the shadow, and has beenused C1 (red) to detect the water bodies (river). The new techniquewas successfully tested on many images of the Google earth andIkonos. Experimental results show that this algorithm effective todetect all the types of the shadows with color, and also detects thewater bodies in another color. The benefit of this new technique todiscriminate between the shadows and water in fast Matlab program.


Introduction
The shadows are physical phenomena observed in most natural scenes [1].Shadows in images lead to undesirable problems on image analysis.Moreover, shadows imply a geometric relationship between objects, light source, and viewpoint.That is why much attention has been paid to the area of shadow detection and removal over the past decades.The shadow can be divided into two major classes (self-shadow and cast shadow).A self-shadow occurs in the portion of an object that is not illuminated by direct light.A cast shadow is the area projected by the object in the direction of direct light [2].

B)) B))
)) e Eqs.where, Src (x, y) the gray value of the pixel position (x,y) of the original imaged (x, y), the grayscale value in pixel position (x, y) after normalization, and min (src), max (src) denote the minimum and maximum gray value in the original image respectively.

Methodology and results
The spatial detection has been applied to detect the shadow and river or any water bodies for high-resolution images.
The proposed algorithm applies color analysis, for the water detection and approximate segmentation the river shape.The approach has been tested with high resolution color images, which extracted from Google earth images suitable of the algorithm has been contrasted with a visual localization of the river entailed in a given area, and shadow, these are the steps of the program.
Step (3) Apply the automatic threshold on Fig. 3e, for first condition and using 3x3 or 5x5 mask to get the image in Fig. 4a.
Step ( 5 g the previo onos (4 me g. 5. No.31, PP. 8 ous steps on eter) to get   Image 4, the analysis of detection algorithms is shown in Table1.Also the analysis of numerous the shadow detection algorithms is shown in Table 2.The algorithms include the algorithm TP, FN, FP, and TN are obtained in accordance with the referenced original image of the corresponding pixels to the total number of pixels in the image.FP value in our method has a notable reduction and indicates that it reduces the error of the proportion due to detecting the non-shadows as shadows.Analyzing PA, CA and OA value, can conclude that PA values in the images change a little and the other values are all improved.The improvement of OA value directly illustrates the effect of the algorithm.

Conclusions
This is a new algorithm for shadow and river detect, are proposed here using of the normalization of the propose new equations CC3 component to detect the shadow use CC1 to detect the water bodies in different color in Fig. 4 a, the shadow is black and the water is green with white background.In Fig. 4b, the shadow is red and the water is yellow with white background, while in Fig. 4c the shadow is black and the water is green with a cyan background.In Fig. 5b, the shadow is black and the water is blue.In Fig. 5c, the shadow is red and the water is magenta.In Fig. 5d, the shadow is white and the water is yellow.Experimental results show that the algorithm detects the shadow more
Fig. 5: (a) o e the above

Fig. 6 :
Fig. 6: nt of results ment metho ly three as e correct sha to the truly ers Accurac t of the co lgorithm to