使用GMM背景减除的红外伪装人体目标快速识别算法

针对红外监视系统对伪装后的人体目标检测处理速度慢、识别准确率低的现状,提出了一种基于混合高斯模型(GMM)的背景减除(BS)快速识别算法用于红外视觉监视系统伪装人体目标检测。首先,该方法采用Mean Shift算法构造初始化背景模型后,使用编码取样原则对像素分等级取样识别前景;再利用移动目标的时间-空间相干性,使用相邻像素预估法快速、精确提取目标图像。其次,设计了基于模板的目标区域聚合新算法,有效地解决了由于伪装造成的人体目标形状特征改变而导致的误检问题。实验证明:新方法切实提高了探测识别效率,降低了人体目标误判率,为实时红外监视提供了有效手段。

Fast algorithm for camouflaged people detection in infrared imagery using GMM BS

In view of the situation which the IR surveillance system is slow on the speed of target detection and low in recognition accuracy,a fast algorithm for camouflaged people detection in infrared imagery using GMM BS was proposed.First,the initialization background model was constructed by the Mean Shift algorithm.Then,two-level codes was used to sample pixel for identify moving objects.The fast mode of vicinage-estimate,which utilizes the spatio-temporal coherence of moving foreground objects,was developed for efficient silhouette detection,also.Second,a model-based grouping algorithm was proposed for fragmented targets made by camouflage.Simulation results demonstrate that the proposed algorithm provides accurate segmentation results of camouflaged people targets without flickering artifacts,while requiring a low computational load.