针对双目视觉定位与测量中某些被测物体角点不明显导致检测精度不高的问题, 提出一种基于亚像素边缘拟合的双目视觉定位与测量方法。利用双目系统标定的结果对拍摄的图像进行去畸变和立体校正处理; 使用Zernike矩方法对预处理的图像进行亚像素边缘检测, 对获得的亚像素点进行聚类和拟合, 计算拟合曲线的交点; 根据对极几何原理来完成左右图像中的交点的立体匹配, 利用视差及三角测距原理获得被测物体的位置信息及其尺寸。实验结果表明, 新的方法能较好地解决角点不明显导致双目视觉立体匹配和定位精度问题, 并能提高检测效率。

In order to solve the problem that the measurement accuracy is not high due to the inconspicuous corner points of some measured objects in binocular vision localization and measurement, a binocular vision location and measurement method based on sub-pixel edge fitting is proposed. The calibration results of binocular system were used to remove distortion and execute stereo rectification of the images, and then the sub-pixel edge detection of the preprocessed images is performed by using the Zernike moment method. After clustering and fitting the edge sub-pixel points, the intersection points of the edge curve are solved. Finally, the stereo correspondence of intersection points in the left and right images was accomplished according to the principle of polar geometry, and the parallax and triangulation principles were used to obtain the position information and sizes of object. The experimental results show that the new method can commendably solve the matching and accuracy problems due to to the inconspicuous corner points of some measured objects ,and also can improve the detection efficiency.