Research On Structural Analysis Of Spacecraft Based Optical Observations

Currently, the Space is gradually becoming a new domain of the world resource contention. Space Surveillance is an important component of national space capability, which can warn the possible collision and attach by predicting the orbit of space target.To get the features of a target from the detecting data is the base of Space Surveillance. This paper takes the spacecraft as research object and regards photoelectric detection system as data acquirer. To achieve the structural analysis based on the optical image, this paper makes a deep research from image preprocessing technique, getting the spacecraft’s components based on shape decomposition problem and the component recognition. The main work and innovations of this paper are as follows:(1) RL-IBD single frame image restoration algorithm is extended to multi-frame RL-IBD image restoration. Firstly, to ensure the reliability of the input image of multi-frame image restoration, a bad frame removing method based on the rate of change of image variance is proposed. And then, single frame RL-IBD restoration algorithm is derived from the Poisson degradation model, and extended to multi-frame image restoration. Another, experiments are done to verify the effectiveness of multi-frame restoration. Lastly, the images are enhanced through piecewise linearization and a binary image is got.(2) A hierarchical shape decomposition method based on saliency is proposed. Based on the Cognitive psychology, a new metric about the visual saliency is designed through the length of a cut and the length of the contour of a part. And then this metric is applied at the field of shape decomposition, and we get the hierarchical shape decomposition method and the components of the aircraft.(3) A recognizer based on a joint feature of the spacecraft part and support vector machine is built, and the structural analysis of the spacecraft is achieved. Firstly, this paper points that the shape of a target is a main measurement of every target. And then the wavelet moments for shape are taken out. In future, we discuss the computational method and the invariance of wavelet moments in detail, and propose a reduction method based on rough set theory against the high redundance problem of initial wavelet moments. Another, a measurement which reflects the relationship between each part and the whole target is added. After getting the joint feature of each part, a recognizer based on SVM through LibSVM is established, and then the structural graph of the spacecraft is formed. Experiments show that the proposed method is telling.