光流向量补偿模型的运动目标检测  

Moving object detection optical flow vectors compensation model

作　　者：张秋佳[1] 王虹[2] 刘威[1] 许松江[1] 

机构地区：[1]哈尔滨理工大学应科院,黑龙江哈尔滨150080 [2]哈尔滨工业大学,黑龙江哈尔滨150001

出　　处：《激光与红外》2011年第11期1263-1266,共4页Laser & Infrared

基　　金：黑龙江省教育厅科学研究项目(No.11541057)资助

摘　　要：针对计算机视觉中的运动目标跟踪与检测问题,特别是运动目标被遮挡丢失和大幅度机动现象,提出了一种基于Lucas-Kanade光流向量补偿算法的运动目标检测方法。该方法利用Lucas-Kanade光流算法计算帧间的光流向量,并选用丢失目标的图像附近若干帧序列图像之间的光流关系,对丢失的光斑目标位置进行估计,以此对序列图像进行补全。实验表明该算法相对于其他传统光流法,提高了光流估算的精确性和可靠性,能更好地检测出运动目标。A moving object detection method based on Lucas-Kanade optical flow vectors compensation algorithm is proposed.It is used to detect and track the moving target and is especially effective when the target is sheltered and makes sharp motion.This method is realized by using Lucas-Kanade′s algorithm to calculate the optical flow vectors of frames,and then choosing the optical flow vectors of the image frames next to the lost frame to predict the position of the lost target and make a full image sequences.Results show that this algorithm can detect moving objects more accurately and reliably when compared with other traditional optical flow methods.

关 键 词：运动目标检测 光流 运动估计 补偿模型 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]