A GPU-based simulation system for infrared images of deep space targets

Weiqing Li, Ranran Xu, Mengyu Yuan

School of Computer Science& Engineering, Nanjing University of Science& Technology, Nanjing, China

In study of deep space targets recognition, infrared images of deep space targets are needed for repeat testing and evaluating. Since the limitation of deep space flight experiments, it is difficult to obtain sufficient infrared images under different conditions. Infrared image simulation technology is brought up to solve this problem efficiently. The principle of deep space targets infrared imaging was studied. Based on the infrared sensor’s optical properties, a hierarchical imaging model was built. The infrared camera and all the effects were simulated respectively, including motion trail of target and space objects, blurring, dispersion, blind elements, and noise. A mixed noise model was introduced by combining the random noise and Perling noise model. In the image simulating process, Graphic Processing Unit was used to produce noise image in real time. According to the reference photo of infrared sensors, infrared simulated images were evaluated using histogram distribution, the trend of intensity, and Signal to Noise Ratio, and the results show these images satisfied targets recognition algorithm.