Algorithm Design And System Implementation For Space Target Recognition Based On Image Fusion

Fusion recognition of space target is based on several sensors, using multi-source information fusion theory and method to recognize the type of space target, such as satellites, orbital transfer vehicles, space robotics and spacecraft etc. It is a promising research topic but still only just started, so is urgent to carry out extensive basic theory and technology research.This paper has summarized the existing research results in image pre-processing, target feature extraction, multi-source information fusion, etc. And then carry out depth and systematic study in space targets feature extracting and feature-level and decision-level fusion. Several research contents have been included:1).The classic feature extraction methods related with space target have been studied and summarized. In order to make full use of their infrared and visible features, while with better performance and accuracy, a combination of several feature extraction methods has been proposed for fusion recognition.2).The related feature-level and decision-level fusion algorithm have been studied. A variety of fusion methods were analyzed and summarized, feature-level including: serial integration and PCA; decision-level including: voting system, bayes reasoning, evidence theory, fuzzy integral. 3).The system implement of semi-physical simulation has been studied, a design scheme for semi-physical simulation system has been proposed, and the relevant software platform for space target recognition has been developed.