基于背景差分法的多运动目标检测与分割  

Multi-moving object detection and segmentation based on background difference method

作　　者：乐英[1] 赵志成 YUE Ying;ZHAO Zhicheng(School of Energy,Power and Mechanics Engineering,North China Electric Power University,Baoding 071000,Hebei,China)

机构地区：[1]华北电力大学能源动力与机械工程学院,河北保定071000

出　　处：《中国工程机械学报》2020年第4期305-309,共5页Chinese Journal of Construction Machinery

基　　金：国家自然科学基金资助项目(51607067);河北省自然科学基金资助项目(E2014502042)。

摘　　要：针对图像序列中的多运动目标的检测和分割问题,本文使用基于背景差分法和相邻帧差分法对多运动目标进行了提取。以179帧图像序列为实验样本对两种方法进行了对比实验,结果显示相邻帧差分法相对于背景差分法一共丢失了159 461像素。对于提取后的运动目标,处理步骤主要为图像预处理、目标检测识别和分割。在Matlab软件中实现了基于背景差分法对多运动目标的检测识别和分割,达到了预期的要求。In order to solve the problem of multi-moving target detection and segmentation in image sequences,this paper uses background difference method and adjacent frame difference method to extract multi-moving target. The results show that the adjacent frame difference method has lost a total of 159 461 pixels compared with the background difference method. For the extracted moving targets,the processing steps are as follows:image preprocessing,target detection,recognition and segmentation. In the Matlab software,the detection,recognition and segmentation of multi-moving objects based on the background difference method is realized,which meets the expected requirements.

关 键 词：多运动目标 背景差分法 目标检测 目标分割 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]