Abstract: The analysis of texture direction feature of strip surface defects was reasonable or not affected the correctness of defect image recognition. To strip surface defects and texture direction feature abstraction which described defect image, the relationship between detection distance and detective direction angle and the basic methods of texture direction feature abstraction of defect image were given. Based on the case studies of crack defect image to strip surface with both noise and noise free, the texture feature should take the maximum value in texture direction and the minimum value in the direction perpendicular to the texture were proposed. The detection distance should be selected according to the specific situation, not too large or too small and the threshold value should be small when analysis defect image with noise.

Key words: Strip, Surface defects, Texture, Direction feature, Gray scale

摘要： 带钢表面缺陷图像纹理方向特征分析是否合理,关乎缺陷图像识别的正确率。针对带钢表面缺陷, 提炼描述缺陷图像的纹理特征量,给出检测距离和检测方向角的关系以及缺陷图像纹理特征提取的基本方法。通过带钢表面无噪声和有噪声的裂纹缺陷图像实例分析,提出纹理方向的特征量取最小值,与纹理方向垂直的特征量取最大值;检测距离根据具体情况选取,不易过大或过小;夹杂噪声的缺陷图像分析时阈值取较小值。

关键词: 带钢, 表面缺陷, 纹理, 方向特征, 灰度级