CAJ | 학술논문

为解决 Sift 在目标跟踪中实时性不高的问题，提出了基于窗口的 Surf 目标跟踪方法。相比 Sift 目标跟踪，快速鲁棒尺度不变 Surf 在速度上有了一定的提高。为进一步提高算法的实时性，只对视频图像中包含目标的局部窗口提取 Surf 特征点。利用特征点之间的相对信息，采用仿射变换方法计算跟踪目标窗口的大小，并获得目标的大致位置信息，最后使用 Kalman 滤波器对窗口大小和位置信息平滑处理。实验结果表明，该方法在提升速度的同时，对目标发生尺寸变化及旋转等情况时能准确地跟踪物体。
위해결 Sift 재목표근종중실시성불고적문제，제출료기우창구적 Surf 목표근종방법。상비 Sift 목표근종，쾌속로봉척도불변 Surf 재속도상유료일정적제고。위진일보제고산법적실시성，지대시빈도상중포함목표적국부창구제취 Surf 특정점。이용특정점지간적상대신식，채용방사변환방법계산근종목표창구적대소，병획득목표적대치위치신식，최후사용 Kalman 려파기대창구대소화위치신식평활처리。실험결과표명，해방법재제승속도적동시，대목표발생척촌변화급선전등정황시능준학지근종물체。
In order to solve the problem that the performance of Sift was not high in real-time, fast Surf object tracking based on window was proposed in this paper. Compared with Sift, Surf had a more improvement in computational efficiency. Further more, in order to improve the efficiency, the Surf features were extracted from local window that only contains the tracking object. Through utilizing the relative information among feature points and the affine transform, the size and position of the tracking window could be calculated. Finally, kalman filter was used to smooth information about the size and location. Experimental results showed that the proposed method could adapt to changes of scale and orientation of the tracking object when improving the computational efficiency.