The Research Of Shadow Detection Algorithms Based On Color Aerial Images

During the imaging process of color aerial digital images, the quality of them were degraded more or less by the imaging technology, imaging environment or other limits. Shadow is one of these degraded phenomenon. It will affect many remote sensing image processes, such as image mosaic, pattern recognition, objects automatic detection, and etc. So it is indispensable to pro-process the shadow area of the images.The pre-processing of image shadow includes two steps. Shadow detection is the first step and shadow removal is second step. It is a new research field and a few people set foot in it. So the emphasis of this thesis is shadow detection. The research content and innovation arc embodied by following ways:1. Because the research content is the shadow of color aerial digital images, this thesis drastically summarized all existing characteristics of color spaces, then lists the transformation formula with the basic color space—RGB color space, introduce the conception of color constancy in the end.2. The characteristic of shadow is the theory base of shadow detection algorithms. So this thesis analyzes the definition, geometrical character and spectral character of shadow in detail.3. Summarizing shadow detection algorithms that brought forth by other people, draws some ideology and ways from these algorithms.4. Brings forth two new shadow detection algorithms: one is based on total variation theory and the other is based on mean shift theory. From the examination results, it is feasibility to apply these two algorithms to detect shadow area on the digital images automatically.5. This thesis uses C++ language to realize the algorithms that we bring forth.From this thesis, we draw a conclusion: the shadow detection algorithms only tightly based on to gain better detection results. So we must find more inherent characteristics of shadow for the sake of getting better shadow detection results. And it is also a direction of this research field. The research goal of shadow detection is offer research content for the research of shadow removal. The final goal is to enable many remote sensing image processes to get better results.