Spatial Non-Cooperative Target Attitude Estimation Based on Contour Feature Template Matching

For the unique characteristics of space object images: lack of texture, scale variation due to changes in imaging distance, and complexity of the background. In this paper, a spatial non-cooperative target attitude estimation method based on contour feature template matching is proposed. According to the 3D model of the spatial target to get the 2D projected images with different viewpoints to form a template library, calculate the Hu moments of the target image and the template library image respectively, take thefirst three components of the Hu moments as the feature vectors, and then calculate the neighbourhood distances of the two feature vectors, and set a threshold to preliminarily screen out the template library of the subset of the target image that corresponds to similar poses. Construct the binary convolution matrix, calculate the binary convolution moments of the target image and the subset of template library images, convert them to the corresponding contour feature vectors, calculate the Euclidean distance and the cosine distance of these two vectors and normalise them, optimise the weight coefficients to adjust the degree of similarity through the genetic algorithm, and then find the maximum similarity of the match corresponding to the final contour similarity between the target image and the subset of template libraries corresponding to the target image and the subset of template libraries. The pose of the image is the output pose. Experiments were conducted on the simulated images, and the results show that the accuracy errors of roll angle and yaw angle are both about 6 degrees, and the accuracy error of pitch angle is about 3 degrees.