基于改进粒子群优化的K-means聚类的焊接缺陷图像识别

Welding Defect Image Recognition Based on Improved Particle SwarmOptimization K-means Clustering

针对传统检测方法对焊接缺陷图像识别的缺点,提出基于改进粒子群算法优化的K-means聚类的焊接缺陷图像识别方法。运用HOG算法提取焊接缺陷图像的特征,利用IPSO算法对K-means聚类模型的聚类点数K进行参数寻优,从而实现对焊接缺陷的检测识别,实验验证表明,该方法能够有效提高焊接缺陷图像的识别效果,总体识别准确度达到94%。

Aiming at the shortcomings of traditional detection methods for welding defect image recognition,a welding defect image recognition method based on K-means clustering optimized by improved particle swarm optimization is proposed.The characteristics of welding defect image are extracted by hog algorithm,and the parameters of K-means clustering points K of K-means clustering model are optimized by IPSOalgorithm,so as to realize the detection and recognition of welding defects.The experimental verification shows that this method can effectively improve the recognition effect of welding defect image,and the overall recognition accuracy reaches 94%.