复杂纹理工件的位姿检测方法研究

针对工业检测中一类具有平面特征的复杂纹理金属工件的位姿检测问题，使用双目视觉和DLP投影仪，将工件的局部特征作为检测重点，提出一种投影差分多线结构光的方法，使差分结构光的中心线相互补偿，减少提取光条中心线出现异常点的概率。运用极线约束特征点行对齐的特性，设计无编码双目结构光系统特征点匹配方法，提高提取平面特征点的稳定性。利用Tukey权重函数改进拟合算法，进一步提升工件被测平面拟合精度和提取工件边缘和角点的准确度，解决了传统拟合模型算法权重分配不合理的问题。经三轴精密滑台实验，并与高精度三坐标仪对比检测结果，验证该方法可以满足高精度三维测量的需求。

Research on Pose Detection of Complex Texture Workpiece

To address the pose detection problem of complex texture metal workpieces with planar features in industrial inspection,using binocular vision and DLP projector,taking the local features of the workpiece as the focus of detection,a method of projecting differential multiline structured light was proposed,which could make the centerlines of the differential structured light compensate each other and reduce the probability of abnormal points on the new centerline.Based on the line alignment properties of epipolar constraint feature points,a stereo matching method for non-coding binocular structured light system was designed,which improved the stability of extracting planar feature points.Then,in order to avoid the defect of irrational weight distribution of traditional fitting model algorithms,Tukey weight function was used to modify those algorithms,which could improve the accuracy of the fitted plane and extracted edges and corners of the workpiece.Finally,the three-axis precision sliding table experiment was performed,the measurement results were compared with the 3D measuring machine.It shows that this method meets the measurement requirements.