Research And Implementation On Shadow Processing Method Of Remote Sensing Images

With the development of remote sensing technology and the improvement of satellite spatial resolution, remote sensing images are used more and more widely. However, during the process, of image acquirement, the image will inevitably be subject to a number of factors. Especially in the high-density urban area, shadows generated by buildings block the information on the image, resulting in image degradation phenomenon. Shadow is caused by lack of light radiation, which means the loss of image information. It means the loss of image information and affects image matching, pattern recognition, feature extraction, and digital photogrammetry. So it is necessary to research and treatment to shadow. And a large number of scholars are concerned about this matter. In general, the processing technology of shadow consists of two parts:shadow detection and shadow elimination.For the shadow phenomenon prevalent in remote sensing image, this paper analysis and deals with this matter, in order to detect shadows accurately and eliminate shadows effectively. In the course of the study, firstly convert images from the RGB color space to HSV color space and C1C2C3color space based on high tonal values、high saturation values、low brightness value、 color invariant features, and the blue ray value fluctuation ratio characteristics of the shaded area in remote sensing images. And propose a shadow detection algorithm based on HSV-C1C2C3color models, which combined spectrum technology, dual-threshold detection as well as the principle of mathematical morphology. Then compare this method to histogram threshold detection method and mathematical morphology detection method. The results show that the proposed detection algorithm has the low complexity and high detection accuracy. Secondly, after shadow detection, calculate light color of shadow and non-shadow areas according to the theory of color constancy and General Grey-World principle. So put forward a shadow elimination algorithm based on General Grey-World color constancy after determining the optimal solution of the Minkowski index P in the light of the mean and variance of the non-shaded area. In order to verify the effectiveness of the proposed algorithm, compare this elimination algorithm with the traditional algorithm, Wallis algorithm and Retinex algorithm. Experiments show that the proposed elimination algorithm does not change the non-shaded area, has a good shadow removal effect and litter color deviation which makes up for the shortcomings of the traditional methods. Finally, the paper achieves shadow processing system of remote sensing image, completes the entire implementation process of the shadow processing on the platform of MATLAB/GUI.