Research On Image Quality Evaluation Method For Space Object Without Reference Standard

The image quality of space objects has an important impact on the success or failure of space on orbit navigation,space object recognition and space on orbit operation.However,the objects of space on orbit tasks are often non cooperative targets,and there is no reference image as the standard.It is of great significance to study the image quality evaluation method of space objects without reference standards for future on orbit services.This paper focuses on the research of single frame image quality evaluation method without reference standard and spatial object sequence image quality evaluation method based on 3D reconstruction results.Firstly,the standard database of space object image quality evaluation is established.According to the situation of flying around in the process of space object image acquisition,the ideal image of space object and different exposure images of natural flying around are simulated,and the standard image quality is degraded according to the possible distortion of space object image acquisition.According to the objective score,the images with different degradation degree and exposure degree are classified into excellent,good,medium and poor grades and labeled,and the image quality standard database of space objects is established.Secondly,a single frame image quality evaluation method based on multidimensional image feature analysis is studied.The frequency domain features,spatial domain features and statistical features based on gray level co-occurrence matrix are extracted from the images in the standard database,and the above features are composed into feature vectors to represent the information extracted from the images.The feature vectors are put into the depth confidence network for training,and the single frame image quality evaluation model is obtained.Compared with the classical image quality evaluation method BRISQUE,the Spearman rank correlation coefficient,Kendall rank order correlation number,Pearson linear correlation coefficient and prediction accuracy of the two methods are compared,The effectiveness of the proposed image quality evaluation method is verified.Thirdly,a method of image quality evaluation based on 3D reconstruction results of space objects is proposed.Based on the analysis of comparative experiments,the optimal angle interval of sequence images in 3D reconstruction is given,and the sequence images of different quality in standard library are reconstructed with this angle interval;The accuracy of the camera position and pose inversion and the accuracy of the reconstructed point cloud are obtained by 3D reconstruction,and the quality level of the sequence image is divided;The matching points between adjacent frames of different quality sequence images are calculated to find out the standard curve that can be fitted when the matching points between adjacent frames of standard quality sequence images are found.The matching points of adjacent frames of different quality sequence images are fitted with the standard curve,and the fitting effect parameters are calculated to summarize the mapping relationship between image quality and fitting effect parameters,Then the quality evaluation model of image sequence is established.Finally,the simulation experiment system of image quality evaluation of spatial target is developed and designed.The function integration of the methods of establishing the database of image quality assessment,single frame image quality evaluation and sequence image quality evaluation is carried out,and the humancomputer interface is designed to facilitate the expansion of the subsequent spatial target quality assessment database.