Identify, Target High-resolution Satellite Imagery-based Power Plant

The technology of object automatic recognition on remote-sensing image is a very active research field in pattern recognition and image processing these years. The objects of power-station— reservoir, condensing tower, coal bridge, chimney and workshop-building are typical artificial-object in land. The research of the automatic recognition of the objects of power-station has great meaning to the military.Our work is based on the 1-metre-resolution remote-sensing image, and adopts the knowledge-driven strategy for object recognition. We set up the knowledge base of each kind of object and design the algorithm according to their feature. They are the recognition of reservoir and condensing tower based on gray and shape features, the recognition of coal bridge based on analysis of lines of the edge, the recognition of chimney using the features of location and self-structure, and the building recognition based on the theory of perceptual organization.The processing of each object mainly follows the three-layers processing flow—the primary, intermediate and advanced steps. They are image segmentation, feature extraction and model-verification. Each layer is guided by the corresponding knowledge.We make many studies and comparisons of segmentation and feature extraction. And we propose many practical algorithms in the research. For example, a binarization algorithm based on Canny operator, a merging algorithm of circle based on statistical powers from Hough transform, an algorithm for line extraction, an universal algorithm for troughs extraction from a histogram. And we improved the algorithms for the generation and verification of the roof-rectangle, and so on.