摘要
为解决经典背景差方法在运动目标检测过程中由于目标短时间滞留导致漏检问题,提出一种记忆随机平均背景差模型的背景建模方法。在记忆随机平均背景差中,将视频中首帧划分为块,计算连续两帧之差作为初始背景,从当前帧往前随机选择一定帧的均值作为背景,通过当前帧减去背景即可得到运动目标。经在国际公开数据集(SBI)上进行大量实验,验证了设计的精准性和有效性。
Object detection is an important issue in computer vision. In order to solve the problem of missing detection caused by short delay of object in the process of moving object detection of the classical background difference method,a background modeling method of random drift average background difference model is proposed. The first frame in video is divided into blocks. Next,the calculated difference between two continuous frames is taken as the initial background. Then,the mean value of a certain frame in front of the current frame is randomly selected. The moving target can be obtained by subtracting the background from the current frame. A large number of experiments have been carried out on the international public data set to verify the accuracy and effectiveness of the design.
作者
张乾
杨玉成
肖永菲
王林
ZHANG Qian;YANG Yucheng;XIAO Yongfei;WANG Lin(School of Data Science and Information Engineering,Guizhou Minzu University,Guiyang 550025,China)
出处
《现代电子技术》
北大核心
2019年第23期44-47,52,共5页
Modern Electronics Technique
基金
国家自然科学基金(61802082)
贵州省科技厅计划基金(黔科合J字[2014]2094号)
贵州省教育厅自然科学基金(黔教合KY字[2017]129)~~
关键词
计算机视觉
首帧划分
模式识别
目标检测
背景建模
性能评估
背景差模型
computer vision
first frame division
pattern recognition
object detection
background modeling
performance evaluation
background difference model