图像分块重建的红外运动目标检测算法  

Infrared image moving object detection based on image block reconstruction

作　　者：刘兴淼[1] 王仕成[1] 赵静[1] 

机构地区：[1]第二炮兵工程学院精确制导与仿真实验室,陕西西安710025

出　　处：《红外与激光工程》2011年第1期176-180,共5页Infrared and Laser Engineering

基　　金：中国博士后科学基金资助项目(20080441274);航空科学基金资助项目(20080112005)

摘　　要：背景差分法是目前运动检测中最常用的一种方法,具有检测位置准确,速度快等优点。背景提取是影响该方法检测效果的关键因素。针对背景差分算法中背景参考帧提取存在的问题,引入非相似度判别算法,给出了一种新的基于图像分块重建的红外运动目标检测方法。该算法在对图像进行合理的区域划分的基础上,利用相邻帧对应块的非相似度值判别出背景区域所对应的块,通过对各个位置的背景块进行组合确定背景帧。仿真实验表明:该算法的运算速度快、抗噪能力强,能够有效地检测出运动目标。Background subtraction is the most popular algorithm for moving object detection at present. It can detect the object position accurately and rapidly. Background extraction is the most important factor that impacts the detection results greatly. With regard to the problem of obtaining background frame in background subtraction algorithm, the dissimilarity discrimination method for detemining the blocks of background was introduced and a new infrared object detection algorithm based on image block reconstruction was presented. The algorithm first carried on the reasonable sub-block processing to the image. Then the dissimilarity of the corresponding blocks of the adjacent frame was used to discriminate the corresponding blocks of background from the frame. At last, the background frame was built through the combination of the blocks of background. Experimental results show that the algorithm has the characteristics of fast operation and strong anti-noise ability and it can detect the moving object effectively.

关 键 词：运动目标检测 背景差分 背景重建 非相似度 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]