基于灰度共生矩阵参数的柳杉材表面节疤缺陷识别

Detection of sound knots and dead knots on Sugi timber using grey level co-occurrence matrix parameter

本研究利用木材表面纹理特征和色彩统计特征开发了柳杉(Cryptomeria japonica L.f.)锯材表面活节和死节的机器视觉自动识别系统。该系统由3部分组成:CCD工业摄像图像采集硬件系统、缺陷自动检出的图像处理模块和基于识别规则的缺陷识别模块。通过对空间灰度共生矩阵参数Contrast运用大津自动阈值分割算法检出潜在缺陷区域,结果表明提案的缺陷检出算法可有效提取柳杉材表面的缺陷区域。根据活节和死节的表面色彩统计直方图确立了区分活节和死节的阈值并构建了活节和死节的识别规则,结果表明基于色彩统计特征的识别规则可有效地识别活节和死节。为了验证系统的识别精度,随机从工厂现场抽取并检测了含有单个和/或多个缺陷的试件156块(共含有94个活节和86个死节),结果表明,活节和死节的正确检出率分别为94.7%和97.6%,活节和死节的正确识别率分别为96.6%和98.8%,整个系统的准确识别率为93.9%。系统的识别精度表明,基于表面纹理特征可实现对柳杉锯材表面节疤缺陷的有效识别,胜任生产线对于缺陷检测精度的要求。

The parameters of contrast calculated from the GLCM were used to locate the potential defects of sound knots and dry knots on Sugi.The rule-based approach was used to identify sound knots and dry knots.The rules to identify sound knots and dry knots were built according to the color feature histograms.A series of samples containing single or multiple sound and/or dry knots were selected at random to verify the efficiency and accuracy of the proposed system.There were 94 sound knots and 86 dry knots on the surfaces of these samples.The accuracies of locating sound knots and dry knots were 94.7% and 97.6% respectively.The identifying accuracies of sound knots and dry knots were 96.6% and 98.8% respectively.The total detecting accuracy of the system was 93.9%.The results indicated that the proposed vision system is an efficient means of detecting sound knots and dry knots.