远距离红外图像目标分析及背景抑制算法

Research on Target Model in Far-distance IR Image and Background Suppression Algorithm

远距离红外图像的目标检测属于弱小目标检测问题,其图像信杂比低,背景结构边缘的抑制残差是单帧图像目标检测时引入的虚假目标的主要因素。分析了远距离红外图像的目标模型,接着根据目标模型选择合适的预测背景模版,并提出了基于该模版去最近邻算法,最后给出该背景抑制算法的仿真结果。该算法与典型的去均值算法、去中值算法相比抑制背景结构边缘更彻底和抑制后图像具有更高信噪比的优点,能为后续的阈值分割和多帧轨迹关联提供良好条件。

The detection of far-distance IR images is same as the detection of the small and small target in nature. As far-distance IR images＇ SCR is low, the leavings of background is the mayor cause of including inveracious target in background suppression. Target model in far-distance IR images is researched firstly, and an appropriate forecasting background templet is chosen on the basis of target model, then the reducing nearest neighbour algorithm is put forward. Finally, the results of evaluation are given. Comparing with classical algorithm, the leavings of background is lower and Suppressed IR images＇ SNR is higher after using the reducing nearest neighbour algorithm. Thereby it can provide better condition for segmentation and association.