针对实木地板的活节、死节、虫眼、裂纹缺陷在线检测速度慢及准确率低的问题,提出在提取R分量图像基础上,利用数学形态学优选缺陷种子点,设计缺陷图像快速增长策略,完成图像缺陷的快速分割;给出了以缺陷面积、边缘灰度均值、内部灰度均值、长宽比为特征的特征提取方法与步骤;构建了基于自组织映射(SOM)神经网络的缺陷辨识模型.实验结果表明:R分量可以准确表征缺陷信息;基于形态学的缺陷分割与SOM网络的缺陷辨识具有快速性和准确性,缺陷分割用时18.8 ms,缺陷分类用时4 ms,平均辨识准确率在89％以上,可以满足木地板缺陷在线检测的速度和准确率要求.%Sound knots, dead knots, pin knots and cracks are the most ordinary wood floor defects. With solving the problems of low speed and low accuracy of online detection against these defects, a wood floor defect detection method is proposed. It extracted R component image at first, and then used mathematical morphological methods to optimize the process of seed selection. With the help of defect image rapid growth strategy, the rapid segmentation of image defect was realized. At last a defect identification model based on self-organizing feature map (SOM) network and with defect area, edge gray-scale average value, inner gray-scale average value and length-width ratio as characteristics was constructed. The result of the experiment show that R component can give accurate information about the defects, and the defect identi fication method based on SOM network is rapid and accurate. Defect segmentation and classification take 18. 8 ms and 4ms respectively, and the average accuracy rate of defect identification is above 89% ,which can meet the speed and accuracy requirements of online detection.
展开▼
