Strategic Target Recognition In Remote Sensing Images Were Studied

In recent years, more and more military images need to be processed to achieve some military objective; therefore, the recognition of strategic targets from remote sensing images becomes increasingly important. Moreover, military targets with huge amounts of image data need to be recognized and processed in a quick and real-time manner; hence, parallel programming is absolutely necessary.This thesis focus attention on modeling, classification and automatic recognition of the strategic targets from remote sensing images, as well as the parallelism of some modules. In particular, the thesis emphasizes on a thorough research on the modeling and the recognition algorithm of large strategic targets such as seaports, roads (highway), and road intersections, etc..For the seaport targets detection, this thesis adopts a segmental method based on the gray-level and the gradient; it works much better than the traditional method and is more suitable for parallelism. In addition, this thesis selects the method that can detect the port element with a " " shape, while neglect the size and direction of the port contour.For the road targets detection, on the basis of the extraction of road features, this thesis probes road models with different satellite resolutions and scene types, as well as the automatic road detection and recognition. Thus, it proposes an expert system that can recognize all types of roads.For the junction targets detection, this thesis offers the structure and the classification of various types of junctions (intersections), and researches different types of algorithm. Integrating the traditional methods of junction detection and corner detection, this thesis then suggests a simple detection method that can use the results of road detection.