摘要
基于机器人视觉平台,对成像目标进行模板提取及边缘检测。针对目标在运动过程中的形变和部分遮挡问题,采用一种改进的Hausdorff距离进行相似度量。提出了一种边缘细化方法及自适应模板尺寸修正策略,减少了边缘特征点,大大降低Hausdorff距离的计算量。基于Hausdorff距离进行模板更新,避免固定模板误差累积问题。实验表明,实时跟踪效果良好,并有效解决了目标部分遮挡问题。
To detect the edge and template of moving object figure using a robotic visual platform, an improved Hausdorff distance algorithm is adopted to track the similar figures during deformation and part occlusion process. The strategies of edge thinning and templates self-adapting can decrease the calculating cost significantly to get the distance. And accumulated error of inaccuracy derived from fixed template can be avoided using Templates self-adapting based on Hausdorff distance. Experiment result showed that the efficiency of real-time tracking is good, and the part occlusion is effectively solved.
出处
《微计算机信息》
北大核心
2007年第03Z期237-238,161,共3页
Control & Automation
基金
国家自然科学基金(60374032)
北京市重点学科基金
