Online Target Recognition for Time-Sensitive Space Information Networks

Abstract

The key difficulties of online target recognition task for space information networks lie in the contradiction between time-sensitive response requirement and resource constraints(e.g., computation resource, communication resource, and training samples). To deal with the above problems, an effective online target recognizing approach is proposed, which seamlessly integrates fast online information processing task and efficient target-specific high-rate compression task. The proposed approach begins with enhancing the target-background separability by introducing intraclass and interclass couples, the new model adapted for the hospt spot image is then obtained by capturing the relation between the online target data and the massive historical data. The light-weight target-specific information is efficiently transmitted into the ground system, and the whole scene is capable of being recovered while the details of targets are being preserved. Compared with the traditional target recognition methods, the proposed approach is more promising for time-sensitive space information networks. The experiments demonstrate the effectiveness of the proposed approach.