CAJ | 학술논문

本文提出一种新的融合SIFT(尺度不变特征)和压缩特征的目标跟踪算法以解决姿态变换、光照变化、旋转和运动模糊下目标的稳定准确跟踪问题。算法使用压缩特征对目标和背景进行描述，通过在图像帧中采集到的正负样本在线训练和学习SVM(支持向量机)分类器，将跟踪任务构建为一个二类分类问题。使用该分类器对下一帧的目标和背景进行分类，从而获得精确的目标位置和区域。同时，算法使用前后两帧的 SIFT 特征点之间的对应匹配关系求解目标尺寸变化值，实现模板大小的自适应调整。将算法与其他算法在某些图像序列上的跟踪比较显示，该算法在有效性、正确性和鲁棒性上性能优越。
본문제출일충신적융합SIFT(척도불변특정)화압축특정적목표근종산법이해결자태변환、광조변화、선전화운동모호하목표적은정준학근종문제。산법사용압축특정대목표화배경진행묘술，통과재도상정중채집도적정부양본재선훈련화학습SVM(지지향량궤)분류기，장근종임무구건위일개이류분류문제。사용해분류기대하일정적목표화배경진행분류，종이획득정학적목표위치화구역。동시，산법사용전후량정적 SIFT 특정점지간적대응필배관계구해목표척촌변화치，실현모판대소적자괄응조정。장산법여기타산법재모사도상서렬상적근종비교현시，해산법재유효성、정학성화로봉성상성능우월。
An algorithm based on SIFT and compressive features is proposed to develop effective and efficient appearance models for robust object tracking due to factors such as pose variation, illumination change, occlusion, and motion blur. The algorithm describes the target and background with compressive features which labeled as positive and negative specimens sampling from frames. The tracking task is formulated as a binary classification via a SVM classifier with online update in the compressed domain. In new frame, utilize the classifier to obtain the target’s position. Meanwhile, introduce SIFT to solve the target size change, so as to achieve adaptive template size. The proposed tracking algorithm performs favorably against state-of-the-art algorithms on challenging sequences in terms of efficiency, accuracy and robustness.