基于SIFT算法的目标匹配和识别

Object matching and recognition based on SIFT algorithm

针对目标跟踪过程易受噪声干扰导致跟踪效果不理想,甚至丢失跟踪目标的问题,利用了尺度不变特征变换(SIFT)方法对单帧图像进行了目标特征点的提取和匹配,并结合线性卡尔曼滤波和聚类分析,剔除误匹配,实现了目标位置的最小均方误差估计。仿真结果表明,当图像存在不同程度的噪声影响时,基于SIFT的卡尔曼滤波目标匹配算法能有效减小目标跟踪误差,精确识别目标位置,提高目标跟踪精度。

Aiming at the problem that it is non-ideal of the tracking effect, even losses object in the tracking process, which is easily affected by noise, scale invariant feature transform （SIFT） algorithm was used to extract and match the feature points, Kalman filter and cluster analysis was combined to eliminate false matches. The minimum-mean-square-error estimate of object location was obtained. The simulation results indicate that the algorithm based on SIFT can drop off the matching error, recognize the object location accurately and improve the accuracy of object tracking.