Recognizing and Measuring Satellite based on Monocular Vision under Complex Light Environment

Aiming at the mission requirements of the space-borne platform approaching and measuring rotating non-cooperative target satellites under complex illumination, a method for identifying and measuring rotating satellites based on monocular vision is proposed. This method realizes the recognition and segmentation of the target satellite with the machine learning method of HOG+SVM. Besides, a method of histogram specification based on average illumination judgment is proposed to extract the inherent features on the target, and the relative position of the target satellite is estimated by PnP based on these features. Experimental results show that this method can achieve better performance in recognition and measurement of rotating target satellites under complex illumination with robustness to influencing factors such as changes of pose, scale and lighting.