基于图像融合分割的实木地板表面缺陷检测方法

Defects segmentation for wood floor based on image fusion method

针对实木地板表面缺陷检测速度慢、精确度低的问题,设计了实木地板视觉检测分选系统,并提出一种基于图像融合的区域生长分割方法。分割方法首先提取缺陷的R分量图像并进行图像缩小,在低维图像空间内运用区域生长方法完成缺陷的快速定位;利用梯度信息插值对缩小图像进行放大复原,并对缺陷进行标记生成参考图像;应用小波变换检索标记参考图像的边缘,以边缘像素点为种子在原图像进行禁忌快速搜索,实现缺陷区域的快速、精准分割。对20幅含有活节、死节、裂纹的样本图像进行了缺陷在线测试,平均分割时间为13.21 ms,缺陷分割区域的准确率达到96.8%。

The surface defects of wood floor directly influence its quality and sorting levels. To solve the problem of slow speed and low accuracy of defects segmentation methods, a fast visual sorting system was designed and a novel segmenting method based on image fusion was proposed. R component image was extracted first and scaling methods were applied to the image. Defects were rapidly located through region growing algorithms in low-dimensional space. Then gradient interpolation method was used to restore the image, and defects were marked to generate the reference image. The wavelet transform was used to identify the margin of the reference image. Finally, dual-threshold growth criterions and taboo table of rapidly located defects were set up to complete the taboo search from the margin of rapidly located region to the outside. The result of the experiment made on 20 sample images with sound knots, dead knots and cracks revealed that the average segmentation time of this method is 13. 21ms, and the accuracy of defect segmentation is 96. 8%.