摘要
在进行多人视觉定位时,存在定位目标数量过多,使得存在大量的目标特征,传统的防丢失技术,多是在目标特征分类的基础上进行视觉定位防丢失的。当目标特征过多时,不能及时的对特征进行准确分类,出现特征遗漏的现象造成定位准确性差、丢失率大的问题。提出采用显著性加权最小二乘算法的多人视觉定位中的防丢失方法,通过计算图像中目标区域的像素聚类数量与初始聚类中心,并划分出跟踪图像初始帧中目标的显著性权值。利用灰色模型建立多人目标运动模型及预测多人目标的运动方向,并通过多人目标的多种特征建立多人目标图像匹配模板,提升多人目标定位的精确性。仿真结果表明,提出的新显著性加权最小二乘算法的多人视觉定位中的防丢失方法在出现多个定位目标区域时,能够提高运动目标定位的准确性,降低了多人视觉定位中的丢失率。
In the paper, a loss prevention method in multi - user visual positioning was proposed based on signifi- cance weighted least squares algorithm. By calculating the pixel of the target area in image, the clustering number and initial clustering center, the significant weight of the target in the initial frame of tracking image was marked off. By using the grey model, the motion model of multi - user target was established and the direction of movement was predicted. And through a variety of features of the multi - user target, the image matching template of multi - user target was established, to enhance the accuracy of multi - user target positioning. The simulation results show that the proposed loss prevention method in multi - user visual positioning based on significance weighted least squares algo- rithm, can improve the accuracy of moving target positioning and reduce the loss rate in multi - user visual positio- ning, when multiple positioning target area occurs.
出处
《计算机仿真》
CSCD
北大核心
2016年第6期433-436,共4页
Computer Simulation
基金
安徽省教育厅自然科学研究项目(KJ2012B143)
安徽省教育厅教学研究重点项目(2012jyxm820)
安徽省教育厅教学研究特色专业建设项目(2014tszy082)
关键词
图像目标
多人视觉定位
防丢失技术
Image target
Multi -user visual positioning
Loss prevention technology
