Visualization of radiation intensity sequences for space infrared target recognition

Infrared target recognition is an important task in space-situational awareness. In the space target detection process, due to the small energy of the point target, it is easy to make the target disappear from the detection field of view under the interference of dense noise, resulting in a decline in recognition system performance. Reasonable representation of the infrared characteristics of a target is an effective means of improving the stability of recognition systems. In this study, a one-dimensional radiation intensity sequence was mapped to a two-dimensional space based on the Gramian angle field, Markov transition field, and recurrence plots to visualize the structural mode of the target radiation intensity sequence and the dynamic properties of the system generating the sequence. On this basis, a recognition framework based on convolutional neural networks was proposed to train and recognize three types of visualized signals and raw data. The experimental results showed that the proposed recognition method based on visualized signals can effectively identify the target and is robust against noise interference and missing data.