冷轧带钢缺陷表面形状恢复新算法

New algorithm of shape from shading for cold rolling steel surface defect

摘要:

为改善冷轧带钢表面缺陷分类识别性能，提出基于从阴影恢复形状原理的表面缺陷三维重构算法． 针对表面缺陷检测系统的光路设计，提出一种改进的Oren-Nayar漫反射模型，结合透视投影模型推导出相应 的反射图方程，利用基于高阶Lax-Friedrichs汉密尔顿函数和牛顿迭代相结合的快速扫描算法完成了该方程 的求解，实现表面缺陷三维信息的提取．利用合成图像和表面缺陷图像进行三维重构实验，结果表明该算法 重构精度高，验证了改进光照模型的正确性．该三维重构算法能够有效地提取冷轧带钢表面缺陷三维信息， 有助于提高表面缺陷分类识别性能

Abstract:

To improve the classification performance of cold rolling steel surface defect, a 3D reconstruction algorithm of surface defect is proposed based on shape from shading theory. A modified Oren-Nayar diffuse re- flection model is presented according to the design of optical path for surface defect detection system, then cor- responding reflection map equation is deduced with perspective projection model. The equation is solved by fast sweeping method based on high-order Lax-Friedrichs Hamilton function and Newton iteration, and finally the extraction of surface defect 3D information is realized. The 3D reconstruction experiments with synthetic image and surface defect image are carried out, and the results show that the reconstruction precision is high, the accuracy of modified illumination model is confirmed. The 3D information of cold rolling steel surface de- fect could be extracted effectively by this algorithm, which is helpful for improving the classification perform- ance of surface defect