基于改进背景差分法的运动物体检测的研究

Research of Moving Object Detection Based on Improved Background Subtraction Method

文中针对传统背景差分法在检测运动物体时的难点,主要是背景重构、背景实时更新、差分阈值选取,提出了一种改进的快速背景重构算法及阈值选取算法。基于传统的连续多帧图像先累加再求平均值的背景重构方法,改进的背景重构算法考虑到背景灰度区域变化不大的特点,在图像像素累加前,进行相邻两帧图像像素值相减,再与阈值作比较,小于阈值才会将前一帧像素值进行累加。针对传统的最大类间方差法是对整张图做灰度级分割,对该算法进行了改进,将图像分割成多块,使用最大类间方差法求出每个图像块的阈值,最后对阈值求取均值,并将该均值作为差分图像二值化的阈值。仿真结果显示,采用了改进算法后,运动物体的检测效果有了较好的提升。

Aiming at the traditional background difference method which has some difficulties in detecting moving objects.Main difficulties are the background reconstruction,real-time background update,difference threshold selection.This paper puts forward an improved fast background reconstruction algorithm and threshold selection algorithm.Based on the traditional method of background reconstruction for the average value of successive multi frame images,the improved background reconstruction algorithm takes into account the change of the background gray area,in front of the image pixels accumulation,the new method will be carried out in two adjacent frame image pixel values subtraction,then compare to threshold,if the result is less than the threshold,the previous frame pixel values are cumulative.Aiming at the traditional method,the maximum variance between classes is to do the gray level segmentation of the entire graph.The algorithm is improved,the image is segmented into several blocks,and the threshold value of each image block is obtained by the method of maximum between class variance.Finally,the threshold value is obtained,and the mean value is used as the threshold value of binarization.The simulation results show that the improved algorithm can obtain better testing result.