基于机器视觉的钢丝绳表面缺陷检测

Surface Defect Detection of Wire Rope Based on Feature Fusion and IWOA-SVM

针对钢丝绳表面缺陷检测困难、精度低、表面特征利用不充分、识别分类器缺乏优化的问题,提出一种基于图像特征融合和改进鲸鱼算法(IWOA)优化支持向量机(SVM)的钢丝绳表面缺陷检测方法。该方法包括特征提取和缺陷类型识别两个部分。特征提取阶段,使用中心对称局部二值模式(CS-LBP)和梯度方向直方图(HOG)提取不同缺陷的钢丝绳表面纹理和梯度特征,并将两特征串联融合,实现特征互补。缺陷识别阶段,首先通过改进WOA控制因子和引入惯性权值提高WOA的提的搜索能力、避免陷入局部最优,利用改进WOA对SVM参数优化,提高SVM的泛化能力并对不同缺陷钢丝绳识别分类。实验表明,该方法能有效识别出不同缺陷的钢丝绳,且稳定性较高。

In order to solve the problems of wire rope surface defect detection difficulty,low detection accuracy,inadequate use of surface features and lack of optimization of recognition classifier,a wire rope surface defect detection method based on image feature fusion and improved whale algorithm(IWOA)optimized support vector machine(SVM)was proposed.The method includes feature extraction and defect type recognition.In the feature extraction stage,CS-LBP and HOG were used to extract the surface texture and gradient features of wire rope with different defects,and the two features were fused in series to achieve complementary features.In the defect identification stage,firstly,WOA control factor and inertia weight were optimized to improve the search ability of WOA and avoid falling into local optimum.Then,SVM parameter optimization was improved using WOA to improve the generalization ability of SVM and identify and classify steel wire ropes with different defects.Experimental results show that the method can effectively identify different defects of steel wire rope and has high stability.