基于抗遮挡能力的红外图像目标跟踪研究

Research on infrared image target tracking based on anti-occlusion ability

红外图像目标跟踪过程中,受复杂背景环境影响,易被遮挡导致目标丢失,针对这一问题,研究基于抗遮挡能力的红外图像目标跟踪方法。将Mean Shift算法应用于红外图像目标跟踪过程中,在初始帧内人为确定区域选取目标模板,统计目标区域内全部像素值,构建目标模型;在后续各帧内,利用候选区域定义有可能包含移动目标的部分,构建候选目标模型;采用Bhattacharyya系数作为相似度标准,构建相似性度量函数,确定令目标模型概率密度和中心坐标位置特征分布相似度最高的点,基于此确定跟踪目标当前帧出现概率最大的位置,并将该位置作为跟踪目标的中心坐标。应用分析结果显示该方法在无遮挡与有遮挡条件下均可准确跟踪红外图像目标,且单帧计算平均时间约为18 ms,满足实时追踪需求。

In the process of infrared image target tracking,due to the influence of complex background environment,it is easy to be occluded,resulting in target loss.To solve this problem,the infrared image target tracking method based on anti-occlusion ability is studied.Put mean Shift algorithm is applied to the infrared image target tracking process,in the initial frame,the target template is selected for the determined region,and all pixel values in the target region are counted to build the target model;in the subsequent frames,the candidate region is used to define the part that may contain the moving target to build the candidate target model;Bhattacharyya coefficient is used as the similarity standard to build the similarity degree The point that makes the probability density of the target model and the feature distribution of the central coordinate position have the highest similarity is determined by the quantity function.Based on this,the position with the highest probability of the current frame of the tracking target is determined,and the position is taken as the central coordinate of the tracking target.The application analysis results show that the method can track the infrared image target accurately under the conditions of no occlusion and occlusion,and the average calculation time of a single frame is about 18/ms,which meets the real-time tracking requirements.