基于红外压缩成像的点目标跟踪方法研究

摘要

目前压缩测量的应用研究主要集中在重构图像方面，但是很多应用中最终目的是检测和跟踪。直接基于压缩测量的检测和跟踪问题尚未解决。该文首次建立一种压缩域到空间域的映射模型，并提出一种无需重构任何图像且直接从低维压缩测量中经解码进行目标跟踪的方法，并分析其应用于天基红外探测的可能性。该方法利用Hadamard测量矩阵构建红外压缩成像系统，采用自适应压缩背景差分法从低维压缩测量信息中分离背景和前景，再从压缩前景信息中解码目标空间位置，并结合数据关联和Kalman滤波算法解决了杂波环境下点目标跟踪问题。理论分析和仿真实验结果表明，该方法能利用少量压缩测量实现目标跟踪任务，并减小探测器规格及相关算法的计算复杂度和存储代价。%Currently the application research of compressive measurements is still focused on the image recovery, but the ultimate purpose is a task of target detection and tracking in many special applications. And the issue performing target detection and tracking based on compressive measurements is not yet solved. The mapping model is firstly exploited to locate the target in the spatial domain through the measurements in the compressive domain. Further, a method tracking point targets through decoding targets location in the low-dimensional compressive measurements without reconstructed image is proposed for the possible application in space based infrared detection. The method uses the Hadamard matrix to design infrared compressive imaging system, and separates the background and foreground image from the low-dimensional compressive measurements by the adaptive compressive background subtraction. With the mapping relation from the compressive domain into the spatial domain, the target location is possibly decoded. Then the task of point target tracking in the clutter environment can be done by the associated data association and Kalman filtering algorithm. The theoretical analysis and numerical simulations demonstrate the approach proposed is able to accomplish a task of target tracking only by using less compressive measurements, and reduce detector scale, computation complexity and storage cost.