CAJ | 학술논문

SURF（Speeded Up Robust Feature）特征提取方法是SIFT（Scale Invariant Feature Transform）算法的改进，具有速度快和精度高等特点，但其对于较大尺寸图像的匹配速度仍然有待提高。文章提出了一种将基于SURF特征匹配算法与卡尔曼滤波相融合的目标跟踪算法，该算法用特征点的中心近似目标最佳位置；通过卡尔曼滤波预测出当前的目标位置，建立自适应匹配窗口；最后，应用SURF算法提取窗口内的特征向量进行匹配。实验表明，该算法在目标发生大尺度旋转和缩放、部分遮挡时能够稳定跟踪，其跟踪速度比SURF算法有很大的提高。
SURF（Speeded Up Robust Feature）특정제취방법시SIFT（Scale Invariant Feature Transform）산법적개진，구유속도쾌화정도고등특점，단기대우교대척촌도상적필배속도잉연유대제고。문장제출료일충장기우SURF특정필배산법여잡이만려파상융합적목표근종산법，해산법용특정점적중심근사목표최가위치；통과잡이만려파예측출당전적목표위치，건립자괄응필배창구；최후，응용SURF산법제취창구내적특정향량진행필배。실험표명，해산법재목표발생대척도선전화축방、부분차당시능구은정근종，기근종속도비SURF산법유흔대적제고。
In this paper, SURF (Speeded Up Robust Feature) algorithm is used for target tracking. SURF algo-rithm has properties of fast running and high precision, which is modified from SIFT (Scale Invariant Feature Transform) algorithm. However, it is difficult to deal with the pictures of large size. In this paper, a new meth-od for target tracking had was proposed, which is combine the SURF algorithm with kalman filter. In the pro-posed algorithm, the center is used to compute the target displacement, and then target position was doped out by kalman filter to build the adapted window. Finally, the SURF algorithm was used to extract target features and matching them. Experimental results showed that this method had strong robustness to zoomed, rotated and shielded of the target, and run faster than SIFT algorithm.