摘要
提出了基于背景分块更新的目标特征点识别、匹配和跟踪算法。该算法对视频检测区域进行分块处理,以图像帧差结果判定前景目标的状态,从而完成背景更新,可有效消除传统的基于概率模型背景更新方法的弊端。在此基础上提出了以目标位置和颜色作为特征信息的匹配、跟踪算法;并将算法成功应用于DM648硬件平台实现了4路PAL视频客流量统计。结果表明,该算法可将客流量统计准确率稳定在95%。
A new target feature matching and tracking algorithm was proposed, based on block-updating of video background. Using this algorithm, block processing the video detection zone, the correct background image was obtained according to the status of foreground target throgh the frame difference. This algorithm could effectively eliminate defects of traditional background updating method, which was based on probability model. Then, a new method to matching and tracking the feature of target, using position, luminance and colour etc. of the target was imposed. At last the algorithm was applied on the DM648 hard platform,which achieved the 4 channel PAL video passenger flow statistics. The results showed the counting precision steady at 95%.
出处
《电子技术应用》
北大核心
2013年第12期141-144,共4页
Application of Electronic Technique
基金
软件开发环境国家重点实验室开放课题(BUAA-SKLSDE-09KF-03)
江苏省大学生创新创业训练计划项目(201310291024Z)
关键词
背景更新
特征提取
目标识别
客流量统计
background update
feature tracking
target recognition
passenger flow statistics
