SURF算法及其对运动目标的检测跟踪效果  

SURF Algorithm and Its Detection Effect on Object Tracking

作　　者：仝如强[1] 黄玉清[1] 田瑞娟[2] 

机构地区：[1]西南科技大学信息工程学院,四川绵阳621010 [2]西南自动化研究所,四川绵阳621000

出　　处：《西南科技大学学报》2011年第3期63-67,共5页Journal of Southwest University of Science and Technology

基　　金：国防基础研究项目(B3120110005)

摘　　要：视频图像的特征点提取和描述是智能交通系统中运动目标跟踪的一项关键技术。由于场景的复杂性、环境的变化以及目标运动的影响,Harris角点检测算法和尺度不变特征变换(SIFT)算法的精度和稳定性都不够,而加速鲁棒特性(SURF)算法具有很强的鲁棒性,运算速度比Harris角点检测算法和SIFT算法有明显提高。应用SURF算法对视频图像进行特征点提取和匹配,并结合聚类分析和卡尔曼滤波对匹配的目标进行跟踪。实验表明,SURF算法对亮度变化具有很强的鲁棒性,并且速度比较快。In intelligent transportation system, extracting and describing feature points of the video frame image is a key technique of the object tracking. Due to the influence of the scene complexity, the environ- ment changes and the object motion, Harris comer detection algorithm and scale invariant feature transform （SINF） algorithm can not meet the requirements of the precision and stability. SURF algorithm not only has good robustness, but also has faster speed than Harris corner detection algorithm and SIFT algo- rithm. Aiming at the above situation, the SURF algorithm was used to extract and match the feature points for the video image, cluster analysis and Kalman filter was combined to object tracking. Simulation results indicate that the algorithm based on SURF has good robustness and speed for the situation of the brightness changes.

关 键 词：目标跟踪 特征点 Harris角点检测算 尺度不变特征变换(SIFT)算法 加速鲁棒性(SURF)算法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]