基于背景重构的运动目标检测算法

Moving object detection algorithm based on background reconstruction

针对背景差分算法中的参考帧提取问题,引入动态时间弯折(DTW)算法,给出了一种新的基于块的背景重构方法。该算法根据相邻两帧图像所对应的背景区域灰度变化不大的特点,利用DTW算法从帧中提取出背景区域所对应的块,再确定出背景帧。仿真结果表明,即使是在图像存在几何畸变和部分像素点缺省的情况下该算法仍能准确地重构背景,实现对运动目标的提取。

Aimed at the problem of obtaining background frame in background subtraction algorithm, a new method based on background reconstruction is presented by using the dynamic time warping （DTW） algorithm. Based on the feature that the values of background are almost the same, and using the DTW algorithm to get the block of background from the frame, the background frame was built. Simulation results show that the background can be reconstructed even the picture is twisted or some pixel is NaN, and the target can be extracted.