Abstract: In order to improve the flexibility and convenience of the measurement of structure surface cracks, a measurement method based on binocular stereo vision with single moving camera was proposed. Firstly, when the precise internal parameters of the camera were known, the images of the cracks were collected from two perspectives respectively. Then, the motion estimation of the camera was performed according to the acquired images, so as to solve the motion parameters of the two positions before and after the camera. Furtherly, crack edge detection based on Canny algorithm was carried out. Finally, the binocular stereo vision model for crack measurement was established according to the internal parameters of the camera, the motion parameters and the edge information of the crack image. The measurement test results show that within the measuring distance of 2m, the relative measuring error of the crack is less than 5%, which can effectively verify the measuring accuracy of the proposed measuring method.

Key words: road and railway engineering; crack measurement; single camera; stereo vision; motion estimation

摘要： 为进一步提高结构表面裂缝测量的灵活性与便捷性，提出了一种基于运动式单摄像机双目立体视觉（binocular stereo vision，BSV）的测量方法。首先，在已知摄像机精确内参的情况下，分别从两个视角采集裂缝的图像；然后，根据所采集到的图像对摄像机进行运动估计，从而求解摄像机前后两个位置的运动参数；进一步地，基于Canny算法进行裂缝的边缘检测；最后，根据摄像机内参数、运动参数以及裂缝图像边缘信息，构建裂缝测量的立体视觉模型。测量试验结果表明：在2 m的测量距以内，对裂缝的相对测量误差不超过5%，可有效验证所提出的测量方法的测量精度。

关键词: 道路与铁道工程；裂缝测量；单摄像机；立体视觉；运动估计