CAJ | 학술논문

针对摄像机运动下的目标跟踪问题,提出了一种适用于动态背景下的改进的Camshift算法.该算法用基于背景运动参数估计的二次差分法进行目标提取,跟踪过程则融合精简的主色调直方图和对遮挡与背景目标颜色相近情况具有鲁棒性的边缘方向直方图作为目标特征,以所得质心为初始搜索点,进行目标模板和候选目标相似性检测,搜索最佳匹配点得到目标对象在当前帧的准确位置.在动态背景下,该算法显著地提高了跟踪的准确性和鲁棒性.
침대섭상궤운동하적목표근종문제,제출료일충괄용우동태배경하적개진적Camshift산법.해산법용기우배경운동삼수고계적이차차분법진행목표제취,근종과정칙융합정간적주색조직방도화대차당여배경목표안색상근정황구유로봉성적변연방향직방도작위목표특정,이소득질심위초시수색점,진행목표모판화후선목표상사성검측,수색최가필배점득도목표대상재당전정적준학위치.재동태배경하,해산법현저지제고료근종적준학성화로봉성.
To solve the problem of object tracking in the condition of camera motion,an improved Camshift Algorithm was put forward.With this algorithm,objects are detected by a twice-frame-difference method which is based on the parameter estimation of background movement.And the tracking process uses the simple main hue histogram and spatial edge orientation histogram as the object feature,the centroid calculated with inter-frame difference as the original searching point,to detect the similarity between the target templet and candidate target and search the best matching point.The exact position of the target object in the current frame is thus determined.The experiment results indicated that the algorithm remarkably improved robustness and accuracy of the tracking in dynamic background.