基于改进支持向量机的皮革划痕检测方法  

Leather Scratch Detection Method Based on an Improved Support Vector Machine

作　　者：马静[1] MA Jing(Shaanxi Institute of Technology,Xi'an 710300,China)

机构地区：[1]陕西国防工业职业技术学院,陕西西安710300

出　　处：《中国皮革》2024年第11期22-28,共7页China Leather

基　　金：陕西省“十四五”教育科学规划2023年度课题(SGH23Y3055);陕西国防工业职业技术学院2022年度重点课题(Gfy22-13)。

摘　　要：针对传统皮革划痕检测方法存在检测准确率低、检测效率不高的问题,提出一种基于K-means聚类算法改进支持向量机SVM的皮革划痕检测方法。首先,对支持向量机基本原理进行分析;然后采用K-means聚类算法解决支持向量机SVM的二分类问题;最后搭建一个K-means-SVM皮革划痕检测模型,通过此模型实现皮革划痕快速准确检测。试验结果表明,本模型的检测精度为96.74%,相较于传统的YOLOv5模型、CRNN模型和SVM-DS模型分别高出了18.85%、20.17%、13.06%,且本模型进行皮革划痕检测的所用时长仅为11.52 s,均低于另外3种模型。由此说明,本模型的检测精度更高,检测速度更快,满足真皮表面划痕检测的实时性和准确性需求。Aiming at the problems of low detection accuracy and low detection efficiency,a method of support vec-tor machine(SVM)was proposed.First,the basic principle of SVM was analyzed.Then K-means clustering algo-rithm was used to solve the two classification problem of SVM.Finally,a K-means-SVM leather scratch detection model was constructed,which enables rapid and accurate detection of leather scratches.The experimental results show that the detection accuracy of this model is 96.74%,18.85%,20.17%and 13.06%higher than that of the tra-ditional YOLOv5 model,CRNN model and SVM-DS model,respectively,and the time used by this model is only 11.52 s,which is lower than the other three models.This shows that the present model has higher detection accuracy and faster detection speed,which meets the real-time and accuracy requirements of dermal surface scratches detection.

关 键 词：皮革 划痕检测 K-MEANS聚类 支持向量机 二分类问题 

分 类 号：TS56[轻工技术与工程—皮革化学与工程]