基于SIFT特征点的双目视觉定位

Object location technique for binocular stereo vision based on scale invariant feature transform feature points

提出了一种结合了SIFT特征点的双目立体视觉定位方法.介绍了对尺度、旋转、视角等变化具有良好鲁棒性的SIFT特征向量,利用SIFT特征向量匹配算法在双目视觉系统采集的左、右图片中分别检测目标、获取匹配的目标SIFT特征点.经过空间匹配点选择、标定点坐标计算等步骤获取左、右图片中具有空间位置一致性的目标标定点,并在摄像机坐标系中恢复目标标定点三维信息.实验结果表明,利用该方法进行目标定位具有较强的适应性,有一定的实用价值.

An object location method was developed based on scale invariant feature transform（SIFT） feature points that is useful for digital image based binocular stereo vision. First, the SIFT feature vector was introduced, as it has good robustness to changes such as scaling, rotation and visual angles. By the use of SIFT feature vector matching, objects which had been collected by a binocular stereo vision system were detected in both left and right images, and thus suitable SIFT feature points were found. Then, by choosing matching points, computing the calibrated point＇s coordinates, and so on, the calibration points of the object could be determined. These calibration points describe the same spatial locations of objects in the left and right images. Finally, the three-dimensional coordinates of the calibration points were rebuilt in the camera＇s coordinate system. The results show that the method discussed has good robustness and practicability.