板材折弯件全长弯曲角度的双目视觉检测方法研究

Research on binocular vision method for full-length bending angle of bending plate

为满足大尺寸板材折弯件弯曲角度的在线检测,提出一种基于双目视觉的全长角度检测方法。首先利用格雷码-相移技术采集折弯件与下模同侧点云,对单幅点云进行数据精简、分割,通过改进特征值法分别对折弯件和下模外表面进行平面拟合;确定公共垂面,并用该公共垂面对点云进行等距切片,求得切片处折弯角度,并计算单幅点云数据对应的全长角度,根据每幅点云采集时对应的位置坐标进行全长拼接,即可得折弯件全长角度。研究表明:该方法能瞬间获得大量点云数据,相当于采用数十万个传感器同时对平面进行采集,有效的提高了数据采集效率和角度的测量精度。

In order to measure the full-length bending angle for large metal plate online, a new angle detection method based on binocular vision was presented. Firstly, point cloud data of the workpiece was obtained using the Gray code- phase shift technology. After the point clouds were simplified and segmented, an improved feature value method was used to fit the outer surface of workpiece and lower mold respectively. Then, the public vertical surface was confirmed, with which the isometric slice for the point clouds was conducted, the bending angle in the section place was obtained and the full-length bending angle corresponding to the single point colud data was calculated. Finally, according to the position coordinates of each point cloud data, the full length bending angle could be calculated completely. The results show that a great quantity of point cloud data can be retrieved by the above method instantly, which is equal to arrange tens of thousands of sensors on the workpiece, and the efficiency of the data acquisition and angle measurement accura- cy are greatly improved.