基于小波多尺度分解的竹条缺陷检测算法  

Bamboo Defect Inspection Algorithm Based on Wavelet Multi-scale Transform

作　　者：贺峰[1] 秦现生[1] 刘琼[1] 宋昕[1] 

机构地区：[1]西北工业大学机电学院,西安710072

出　　处：《机电一体化》2010年第1期46-49,共4页Mechatronics

摘　　要：在竹条缺陷的图像检测过程中,针对缺陷种类多,尺寸和形状变化大且竹纹理干扰大,提出一种竹条缺陷检测算法。首先设计了竹条边界检测算法,以判断竹条的三角条缺陷;其次对竹条图像的小波分解子图像进行32灰度级量化处理,经过分析图像共生矩阵的逆差矩差分值来获取最佳的小波分解层,实现了对竹纹理的滤除;最后结合Otsu自动阈值检测方法,完成了竹条常见的6种缺陷的检测。经过实验,该算法准确稳定,能有效滤除竹纹理的干扰,竹条缺陷检测效果较好。Due to the interference of bamboo texture and various bamboo defects, this paper proposed a bamboo defect inspection algorithm in the image defect inspection process. Firstly, the bamboo border detection algorithm was designed to detect the defect of triangular strip. Secondly, processed the wavelet decomposition sub-image by using 32 gray-scale image quantization algorithm, and analyzed the co-occurrence matrix of the image in order to obtain the best wavelet decomposition level and filter the bamboo texture. Finally, six kinds of common defects were detected combined with Otsu threshold segmentation. The algorithm was proved to be accurate and stable by experiment, it could filter the bamboo texture effectively, and the result of defect inspection was good.

关 键 词：缺陷检测 图像量化 小波分解 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置] TP391.41[自动化与计算机技术—控制科学与工程]