摘要
由于运动目标容易受到遮挡,导致识别效果不理想,提出视频前景区域运动目标姿态识别方法。采用金字塔变换对视频图像展开采样处理,通过Vibe前景检测算法完成视频的前景检测,同时在检测过程中消除了视频鬼影,提高了目标跟踪精度;采用mean-shift目标跟踪方法在视频前景区域中跟踪运动目标,提取运动目标的姿态特征,将其输入支持向量机决策函数中,完成运动目标的姿态识别。仿真结果表明,所提方法具有较高的跟踪精度和跟踪效率,且姿态识别准确率高。
Due to the susceptibility of moving targets to occlusion,the recognition effect is not ideal.Therefore,a video foreground region motion target pose recognition method is proposed.Pyramid trans form was used to sample video images at first.And then,Vibe foreground detection algorithm was adopted to complete the foreground detection.Meanwhile,the ghost of the video was elimina ted during the detection,so that the target tracking accuracy was improved.Moreover,the mean-shift method was a dopted to track the moving target in the fore ground region,thus extracting the posture feature of the moving target.Finally,it was input into the support vector machine decision function to complete the recognition of posture.The sim ulation results show that the proposed method has high tracking accuracy and efficiency,as well as high accuracy of attitude recognition.
作者
张清蓉
陈龙灿
刘庆
ZHANG Qing-rong;CHEN Long-can;LIU Qing(Chongqing College of Mobile Telecommunic ations,School of Intelligent Engineering,Chongqing 401520,China;Chongqing University of Posts and Telecom munications,Automation College,Chongqing 400065,China)
出处
《计算机仿真》
2024年第7期258-262,共5页
Computer Simulation
基金
重庆移通学院高等教育教学改革研究项目(22JG310)
重庆市教委科学技术研究计划项目(KJZD-K202202401)。
关键词
机器视觉
前景检测算法
目标跟踪方法
支持向量机
姿态识别
Machine vision
Foreground detection
Target tracking method
Support vector machine
Recognition of posture
