Research On Recognition Method Of Typical Space Target Based On Image

The classification and recognition method of spatial typical targets is generated by the continuous exploration of space technology and the continuous development of space resources.Because of the consideration of space security,the establishment of the space surveillance system of the typical space target has important significance.The classification and recognition of spatial targets is an important part of the space surveillance system,and more and more researchers pay attention to the importance of spatial object recognition.At present,the research in this field is still in the primary stage,so it is necessary to carry out the relevant theoretical research.Based on the improvement of the detection capability of space target,and from the perspective of imaging plane target,we discussed the classification and identification of spatial typical targets under the multi-angle of view.In this paper,we proposed two methods of image classification based on local feature modeling.The main work and achievements of this paper are as follows.(1)To construct the stable expression of the image under different morphological changes,the method of image classification process based on local feature modeling was researched,as well as the analysis of key technologies: the local feature extraction,feature encoding and feature pooling.(2)Based on local feature modeling,and refer to Bag of Features model,the local image feature extraction and feature encoding problem are studied.And we proposed a method of sparse coding feature construction model based on a new SIFT algorithm,Dense SIFT.(3)On the basis of above,by the introduction of the visual saliency detection method,a weighted method of improving the feature pooling in Spatial Pyramid Representation model is introduced.Further improvement is also proposed to solve the problem of rotation.Finally,the simulation analysis of the classification and recognition is given for verification.