基于人类视觉系统的实时红外目标检测方法
Real-Time Infrared Target Detection Method Based on Human Vision System

针对现有基于人类视觉系统红外目标检测方法存在的实时性问题，提出一种融合显著区域提取和局部对比度分析的红外目标检测新方法。首先，采用快速中值滤波和灰度合并得到二值图像并进行融合，以提高显著区域提取的精度；其次，对融合后的二值图像进行连通域分析，过滤连通区域过大和过小的区域，以减少后续检测方法的计算量；然后，在显著区域所对应的原图区域进行局部对比度分析，得到区域局部对比度图，并进行阈值分割得到弱小目标所在位置。理论分析和仿真结果表明该方法在保证检测性能的同时减少了数据运算量和储存量，提高系统的实时性。
Focusing on the real-time performance problem of existing infrared target detecting method based on human visual system, a new robust infrared target detecting method that fuses salient region extraction and local contrast analysis was proposed in this paper. Firstly, the binary image was obtained by the fast-median filtering and similarity analysis to enhance the salient region extracting accuracy. Secondly, the too large and too small area was filtered by connectivity domain analysis in binary image to reduce the amount of calculation subsequent detection algorithm. Finally, the regional local contrast diagram was obtained by local contrast analysis in significant intraregional, and the position of the dim small target was obtained by dividing threshold value. Theoretical analysis and simulation results show that the proposed method not only can ensure detection performance, but also reduce the amount of computation and data storage capacity, and improve the real-time performance of the system simultaneously. The SNR gain is up to 9.1 dB when infrared image SNR is lower