HCNN-PSI: A hybrid CNN with partial semantic information for space target recognition

Abstract

Space target recognition is the basic task of space situational awareness and has developed significantly in the last decade. This paper proposes a hybrid convolutional neural network with partial semantic information for space target recognition, which joints the global features and partial semantic information. Firstly, we propose a two-stage target detection network based on the characteristics of deep space targets. Secondly, we use the Mask R-CNN to segment the main components of the detected satellite. Thirdly, the recognized target and the segmented components are sent to the hybrid extractor to train the hybrid network. What we have done is to find the proper weights of the partial semantic information that plays different importance. The loss function of the hybrid network integrates the global-based and component-based loss with different weights. In comparison with several sets of comparative experiments, the proposed method has achieved a satisfactory result. Besides, we have simulated some real space target images by data processing and achieved a competitive performance in both the simulated dataset and the public dataset.