CAJ | 학술논문

针对复杂背景下伪装色运动目标的跟踪问题，提出一种基于运动信息与梯度方向直方图的跟踪方法。通过金字塔光流法获取视频序列中目标运动矢量范围；利用目标区域的灰度特征以及Hog特征构建特征向量，沿着运动主方向在速度大小范围内进行二分法匹配，进而确定目标的准确位置。实验结果表明，针对平缓变化的视频序列，对于颜色色调范围较大或者边缘结构较为明显的伪装色运动目标具有良好的跟踪效果。
침대복잡배경하위장색운동목표적근종문제，제출일충기우운동신식여제도방향직방도적근종방법。통과금자탑광류법획취시빈서렬중목표운동시량범위；이용목표구역적회도특정이급Hog특정구건특정향량，연착운동주방향재속도대소범위내진행이분법필배，진이학정목표적준학위치。실험결과표명，침대평완변화적시빈서렬，대우안색색조범위교대혹자변연결구교위명현적위장색운동목표구유량호적근종효과。
In order to track the camouflaged moving object under complex background, a method based on multi-resolution optical flow and Histograms of Oriented Gradients(HOG)is proposed. The speed vector range of object is obtained by using the multi-resolution optical flow. The gray feature and HOG feature of object area are used to construct the feature vector, which is a standard of bipartite matching in the range of speed value along the main direction of object, and then the object area is obtained accurately. The experimental results demonstrate the proposed approach is effective to track the camou-flaged moving object in slowly changing background when the color of camouflaged object changes greatly or the moving object has obvious edge structure.