Research On Space-based Visible-light Space Target Detection And Recognition

Detection and recognition of space target is one of the major means of obtaining space situational information and it serves as a significant basic capability of space controlling. Space-based visible light surveillance, as a way of space target detection and recognition, has more potential for future development due to its many advantages, which include higher energy efficiency, better concealment, excellent real-time performance and ease for miniaturization, etc.This paper mainly focuses on researching pre-processing of deep space imaging and detection and recognition of space targets, covering the following:1) Modeling method is used to analyze the image-forming principle and characteristics of typical space-based visible-light images, meanwhile several common denoising methods are introduced to do the image pre-processing. After comparing the results of the experimental simulations, the best noise reduction method for this topic is determined to be applied in the latter stage of target detection and recognition.2) The article expounds how Multistage Hypothesis Testing(MHT) helps to obtain the possible motion trail of the small dim targets in space environment. Based on this, the iterative optimization method is then used to improve estimation accuracy of the target motion parameters and trail and decrease the calculation and storage capacity needed in MHT. Experimental simulation shows that with this method we could acquire better target motion trail and higher target positioning precision.3) Due to the fact that in deep space the imaging of starts and small space targets are similar and it will interfere target recognition, this paper puts forward a star map identification algorithm based on ratio of centroid polar moment Regarding the ratio of centroid polar moment of the matching triangle as the auxiliary matching characteristics in the navigation feature library helps improve the recognition speed and accuracy as well. The experimental simulation indicates that this method can not only achieve the desired results but also effectively avoid the redundancy in the star map matching process.