基于图像处理的金属腐蚀等级评价方法  

Evaluation Method of Metal Corrosion Grade Based on Image Processing

作　　者：郝亚东 万博 张钟庆 徐如远 张生鹏[2] HAO Yadong;WAN Bo;ZHANG Zhongqing;XU Ruyuan;ZHANG Shengpeng(School of Reliability and Systems Engineering,Beihang University,Beijing 100191,China;Defense Technology Research and Test Center of China Aerospace Science&Industry Corporation,Beijing 100854,China)

机构地区：[1]北京航空航天大学可靠性与系统工程学院,北京100191 [2]航天科工防御技术研究实验中心,北京100854

出　　处：《腐蚀与防护》2024年第6期103-110,共8页Corrosion & Protection

摘　　要：金属在服役过程中不可避免会产生锈蚀,目前工程中的金属腐蚀等级评价以人工评价为主,存在效率低和准确性差的问题。根据金属腐蚀前后像素点差异化特征,利用卷积神经网络结合滑动窗口法实现腐蚀特征分类及腐蚀区域定位,提出了一种颜色聚类结合标准色图谱信息表,实现了金属腐蚀等级的计算机评价。结果表明:该方法评价准确率达到96%,具有检测速度快、客观性强和准确性高的优点,解决了基于多指标对金属腐蚀等级进行快速评价的问题。Metals inevitably undergo corrosion on its surface during service.The current metal corrosion grade evaluation in the engineering is mainly based on manual evaluation,which has the problems of low efficiency and poor accuracy.According to the differentiation characteristics of pixels before and after metal corrosion,convolutional neural network combined with sliding window method was used to achieve corrosion feature classification and corrosion area location,and a method of color clustering combined with standard color map information table was proposed to achieve metal corrosion grade evaluation by computer.The results show that the evaluation accuracy rate of this method reached 96%,which had the advantages of fast detection speed,strong objectivity and high accuracy,and solved the problem of rapid evaluation of metal corrosion grades based on multiple indexes.

关 键 词：卷积神经网络 滑动窗口 颜色聚类 标准色图谱信息表 腐蚀等级评价 

分 类 号：TB304[一般工业技术—材料科学与工程]