摘要
针对工业视觉检测系统中流水线多目标实时跟踪问题,提出一种目标中心点快速匹配方法与目标图像水平投影曲线投票校正技术相融合的多目标跟踪算法。其过程是:利用多个目标在连续帧图像中的整体分布的稳定性进行整体匹配,得到多点匹配的一一映射关系;针对局部目标的晃动、跳动问题,利用目标图像水平投影曲线,采用投票校正技术消除局部运动所带来的误差,对映射关系进行校正。该算法应用于某钢铁厂的棒材计数系统,有效地排除了复杂环境下多目标跟踪时的误识别、漏识别和位置交叉的问题,保证了计数的准确性与快速性,当目标数小于50时,跟踪时间小于1 ms。
For real-time tracking of multi-object on assembly line, a novel algorithm was presented based on fusion of object-center-fast-matching technique and image horizontal projection-curves voting-correction technique. Based on the stability of integral distribution of multi-objects don＇t change too much in consecutive frames, this algorithm makes integral matching and obtains the mapping relation for dots matching, and secondly projects image to horizontal axis and uses voting-correction technique to remove errors in mapping relation caused by objects＇ local movement. This algorithm was used in a certain steel-bar counting system. The results show that it can solve the problem of misrecognition, un-recognition and position crossing, and tracking time for each frame is less than 1 ms when objects are less than 50.
出处
《中南大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2007年第3期528-532,共5页
Journal of Central South University:Science and Technology
基金
国家自然科学基金资助项目(60573079)
关键词
多目标跟踪
实时算法
点匹配
校正
multi-object tracking
real-time algorithm
dots matching
correction
