基于形态学重构的实木地板缺陷分割方法研究

针对传统区域生长方法中,由于噪声种子存在及种子点单步邻域搜索所导致的分割时间长、检测精度低的问题,提出基于形态学重构的实木地板在线缺陷分割方法。方法首先定义不同阈值下的两幅模版图像,其中低阈值图像用于种子优化,高阈值模版用作种子膨胀生长;通过定义腐蚀终止准则,完成低阈值图像下的缺陷骨架提取;运用"去毛刺"操作,最终实现缺陷骨架内的种子点优选;然后,运用测地膨胀,结合高阈值模版,完成板材缺陷区域的快速生长;最后,应用"孔洞填充"、"去毛刺"优化边缘,实现缺陷目标的提取。实验分别在像素512*512、256*256和128*128下进行,通过与传统区域生长方法的比较,表明方法实现了缺陷区域的准确分割,分割速度与精度能够满足地板在线分选要求。

Research on Wood Floor Defects Segmentation based on Morphological Reconstruction

Due to the problems of noise existed, time consuming, and lower detection precision in region growth, this paper proposed a novel method using morphological reconstruction technique to conduct on-line defects detection for wood floors. Firstly, this method used two gray thresholds to get two gray images. The lower threshold image was used for seeds optimization, while the higher threshold image was used for seed expansion growth. Skeleton extraction of defects for the lower threshold image was accom- plished by defining corrosion stop criteria. The burr operation was then employed on the lower threshold image to achieve seeds optimization. Secondly, the expansion, holes filling and burr optimization were exerted separately on the high threshold image to accomplish rapid growth of lumber defect area defects extraction. The experiments were carried out in the resolutions of 512 × 512, 256 × 256, and 128× 128, respectively. By contrasting with the traditional region growth method, the tests showed that the proposed method could guarantee the accuracy and the speed for wood floor on-line defects detection.