基于计算机视觉的钢构件防腐涂层缺陷检测

Defect Identification of Anti-corrosive Coating of Steel Members Based on Computer Vision

钢结构在现代建筑中有着广泛应用,但钢材的耐腐蚀性较差,锈蚀将降低钢构件的承载力甚至影响结构安全。对钢构件防腐涂层进行定期检测与维护是保证钢结构耐腐蚀性能的主要方法之一。为实现钢构件防腐涂层表面缺陷的自动识别,结合图像处理技术与支持向量机分类算法进行钢构件防腐涂层空鼓、裂纹、剥落3种缺陷的检测。在图像处理阶段,采用自适应中值滤波去除图像噪声,并基于Otsu阈值和改进的Canny算子对增强图像进行分割;在特征提取阶段,提取出包含简单几何特征、不变矩特征、投影特征及纹理特征的65维缺陷图像特征,随后采用Fisher判别准则将特征向量降至37维;将所得特征向量作为输入,基于支持向量机算法搭建多分类模型,识别率达95.83%。所提方法有效实现了钢结构构件防腐涂层表面缺陷的识别分类。

S:teel structures are widely used in building constructions,but the corrosion resistance of steel is not satisfying.Corrosion will reduce the load-bearing capacity of the steel members,even affecting the safety of the structure.Periodical inspection and maintenance of the anti-corrosive coating of steel members is one of the most important methods to ensure the coating's effectiveness.In order to automatically detect and classify surface defects in the anticorrosive coating of steel members,a computer vision detection method based on image processing technology and the Support Vector Machine(SVM)algorithm was proposed.In the image processing stage,adaptive median filtering and image enhancement are applied to the gray-scale defect image.Then the Otsu thresholding method and Canny operator were used to segment the enhanced images.In the feature extraction stage,65-dimensional features including simple geometric features,invariant moment features,projection features,and texture features of the images were extracted.The Fisher criterion was used to select the 37 most contributing features.The eigenvectors were used as input to develop a multi-classification model based on the SVM algorithm,achieving a recognition rate of 95.83%.The proposed method can effectively detect and identify the surface defects in the anti-corrosive coating of steel structural members.