Based On Support Vector Machine X-ray Source Astronomical Image Point Source Recognition Method Research

The technology of autonomous navigation based on X-ray source celestial location has become a forward-looking astronomical navigation method,which can provide navigation information such as position,speed,attitude and time for spacecraft.Spacecraft autonomous navigation is not only the meaning of its lack of ground equipment support,but also in the autonomous navigation technology can greatly improve the spacecraft's mobility,concealment and anti-jamming.When the X-ray source astronomical position is used to locate the spacecraft,the point source recognition technology in X-ray source astronomical image becomes the primary problem,and the point source recognition technology of X-ray source astronomical image is of great significance.This dissertation is based on the measured image of space X-ray source,the research work carried out in the following three aspects: the acquisition of X-ray source raw data,the processing method of X-ray source astronomical image and the point source extraction of X-ray source astronomical image.In the acquisition of X-ray source raw data,this dissertation will use Chandra satellite observation as a data base.First of all,this dissertation will introduce the data format of the observation briefly.Then,according to the original data content of this dissertation,this dissertation will introduce the modules needed in the cave of Chandra satellite data processing software.Finally,the actual observation case will be given as an example,this dissertation will give the processing flow and processing results.The processing method of X-ray source astronomical image includes two parts: X-ray source astronomical image preprocessing and the extraction of potential point centroid.Image preprocessing includes two parts: image filtering and threshold segmentation.This dissertation first analyze and compare the filtering and threshold segmentation methods commonly used in image processing,and then analyze three commonly used centroid extraction algorithms.Finally,according to the X-ray astronomical image processing results,adaptive Wiener filter,adaptive threshold segmentation method and based on the peak point source centroid extraction method are selected as image processing method.Aiming at the phenomenon of diffusion point source in the X-ray source astronomical image,a method based on digital morphology is proposed.Taking the actual observation cases as an example,the number of potential point sources extracted in the image can be reduced by 30%-40%,which greatly reduces the workload of the latter point source recognition algorithm.The accuracy of the point source extracted by the image processing method can only reach 58.62%,which proves that there are still many pseudo-point sources in the potential point source.Therefore,this dissertation proposes a point source recognition algorithm based on support vector machine to further screen the potential point source.In this dissertation,principal component analysis algorithm and granularity calculation are used to optimize the original training samples,and the multi-classification problem is transformed into two classification problems by using binary tree,and the classifier itself is optimized.Through the experiment of the actual observation data,the method proposed in this paper obtains the X-ray source point source image under the premise that the recognition accuracy is consistent with the Wavdetect algorithm.The accuracy of the observation in the observation case can reach 93.16%.