针对射线焊缝图像缺陷识别精确度不高问题,提出一种基纹理特征的焊缝缺陷识别方法.由于缺陷图像存在噪声,用中值滤波对焊缝图像进行降噪;通过分析焊缝图像纹理的全局与局部特征,构造焊缝图像的全局与局部直方图特征;把这两部特征进行联合用于表示焊缝图像;用最近邻分类器对缺陷图像进行识别.同时,使用多组焊缝缺陷图像数据集测试该方法.另外,为了模拟实际生产中缺陷种类的复杂性,还对缺陷图像进行9个不同角度的旋转.综合实验表明该方法具有较高的识别率且具有旋转不变性.实验中对复杂缺陷图像识别率超过93.96%,优于现有方法,能够满足实际需要.%Aiming at the problem of low accuracy of radiographic weld image defect recognition, a seam defect recognition method based on texture feature was proposed.Due to the noise in the defect image, the median filter was used to reduce the noise of the weld image.By analysing the global and local features of the weld image texture, the global and local histogram features of the weld image were constructed.The two features were combined to represent the weld image.The nearest neighbour classifier was used to identify the defect image.At the same time, the method was tested using multiple sets of weld defect image data sets.In addition, in order to simulate the complexity of the actual production of the type of defect,the defect image was also rotated nine different angles.The comprehensive experiment shows that this method has higher recognition rate and rotation invariance.In the experiment,the recognition rate of the image of complex defects exceeds 93.96%,which is better than the existing methods and can meet the actual needs.
展开▼