CAJ | 학술논문

在基于视频的汽车跟踪系统中，运动汽车产生的阴影会导致无法对运动汽车进行准确跟踪，传统阴影去除算法通常具有一定的局限性和复杂性，无法直接应用于汽车跟踪系统。针对这一问题，提出一种基于梯度投影和小波变换的阴影去除算法。先对含有阴影的目标区域求梯度，将梯度值向水平方向投影，然后用较大尺度的Marr小波检测出投影曲线中的屋顶状曲线段的有效区间，最后在有效区间内，用较小尺度的Gauss小波检测出投影曲线的最大正峰值点和最大负峰值点，由2峰值点的位置即可确定车宽。实验表明，该算法简单快速，能有效去除运动汽车的阴影。
재기우시빈적기차근종계통중，운동기차산생적음영회도치무법대운동기차진행준학근종，전통음영거제산법통상구유일정적국한성화복잡성，무법직접응용우기차근종계통。침대저일문제，제출일충기우제도투영화소파변환적음영거제산법。선대함유음영적목표구역구제도，장제도치향수평방향투영，연후용교대척도적Marr소파검측출투영곡선중적옥정상곡선단적유효구간，최후재유효구간내，용교소척도적Gauss소파검측출투영곡선적최대정봉치점화최대부봉치점，유2봉치점적위치즉가학정차관。실험표명，해산법간단쾌속，능유효거제운동기차적음영。
The shadow by moving vehicle will lead inaccurate tracking in the video - based vehicle tracking system .Traditional shadow removal algorithm is with certain limitation and complexity ,and can not be applied to vehicle tracking system directly .In view of such limitation ,a shadow removal al‐gorithm based on gradient projection and Wavelet T ransform is raised .It firstly calculates the gradient value of target area with shadow ,then projects the gradient value horizontally ,and finally detects val‐id interval in the roof curve of projection curve by Marr wavelet .At last ,in the effective range ,the maximum positive peak and maximum negative peak of projective curve will be detected by Gauss wavelet ,and the vehicle width can be fixed by the position of two peak points .Experiment shows that such algorithm is proved to be simple and rapid ,and can efficiently remove the shadow of moving ve‐hicle .