【摘要】 针对传统的工件缺陷提取方法存在鲁棒性差的问题，提出一种基于同态滤波的改进低秩矩阵恢复算法。首先使用同态滤波方法增强光照分量、抑制工件反射分量，减小光照不均和工件强反光产生伪缺陷的影响；然后应用鲁棒主成分分析模型将工件表面缺陷提取问题转换为低秩背景矩阵和稀疏缺陷矩阵分离的低秩矩阵恢复问题；最后使用非精确拉格朗日乘子法对由鲁棒主成分分析模型转化的凸优化模型进行求解。以带有凹坑、划痕缺陷的轴类工件为样本，通过计算F-measure值完成方法验证，实验结果表明：在不同光照强度的实验条件下，离散傅里叶算法提取凹坑缺陷和划痕缺陷的平均F值分别为0.435 7和0.381 9；本文提出算法提取凹坑缺陷和划痕缺陷的平均F值分别为0.726 0和0.716 9，结果验证了所提算法的有效性和较高鲁棒性。更多还原

【Abstract】 An improved low rank matrix recovery algorithm based on homomorphic filtering was proposed to solve the problem of poor robustness in the traditional methods of extracting workpiece defects. Firstly, the homomorphic filtering method was used to enhance the illumination component and suppress the reflection component of the workpiece to reduce the influence of illumination unevenness and strong reflection of the workpiece to produce false defects. Then the robust principal component analysis model was used to transform the surface defect extraction problem into the low-rank matrix recovery problem separated from the low-rank background matrix and the sparse defect matrix. Finally, an imprecise Lagrangian multiplier method was used to solve the convex optimization model transformed from the robust principal component analysis model. With indentation, scratch defects such as axial workpiece as sample, by calculating the Fmeasure values completion methods validation, the experimental results show that under the condition of different light intensity experiment, the average F value of pit defect and scratch defect extracted by Discrete fourier transform is 0.435 7and 0.381 9; in this paper, the average F value of pit defect and scratch defect extracted by the algorithm is 0.726 0 and0.716 9, the results show that the proposed algorithm is efficient and robust.更多还原