摘要
针对曲面工件加工过程中出现的表面缺陷问题,使用Gabor函数能够提取出图像在不同位置、频率和方向上特征的优点,构建5个方向8个尺度的Gabor滤波器组.对采集的图像进行Gabor滤波,提取滤波后图像的灰度均值和方差作为Gabor纹理特征向量,使用主成分分析法进行特征的降维处理,最后建立马氏距离最近邻分类器,实现了对工件表面缺陷的识别和分类.
Aiming at the surface defects that emerged in the process of cambered workpiece matching, the Gabor function was used to extract characteristics on different frequencies and orientations. A Gabor filter bank was built with five directions and eight dimensions. The collected images were filtrated by the Gabor filter. Then, the mean gray level and the variance of filtered images were detected as the Gabor texture feature vector. The principal component analysis was used for feature reduction. A nearest neighbor classifier of Markov was established to realize the identification and classification of surface defects.
出处
《中国计量学院学报》
2015年第1期46-49,59,共5页
Journal of China Jiliang University
基金
浙江省重大科技专项计划项目(No.2011C12025)
关键词
GABOR滤波器组
特征提取
识别分类
Gabor filter bank
defects detection, identification and classification