高效消除表面缺陷检测盲区的机器视觉在线检测系统

Machine vision on-line detecting system for effectively eliminating non-detection zone of surface defects

  • 摘要: 针对皮革表面类缺陷用机器视觉在线检测时,某些缺陷成像质量对照明方向呈现各向异性甚至无法显现而形成检测盲区的问题,研究设计了一套可消除该类检测盲区的在线检测系统。基于缺陷处的散射和信息采集机理得出缺陷特征光强随照明方向的变化规律,并以压印和油墨类缺陷为例进行实验研究,发现2类缺陷特征光强分布均与照明方向相关,缺陷的检测盲区与有效区呈正交态。据此,采用自主设计的侧入式均匀照射离轴面光源,设计出线阵和面阵正交复合的在线检测系统,采用垂直目标面采集信息、45°角暗场正交照明,线阵相机和线光源组成主检测系统,面阵相机和离轴面光源辅检测系统,对随机出现的各向异性缺陷进行上百个样本检测,实验证明,缺陷检测盲区被有效消除,缺陷识别率提升了近22%,检测效率也得到大幅度提高。

     

    Abstract: For leather surface defects with the machine vision on-line detection, the image quality of some leather surface defects presented aeolotropism to the lighting direction, and the defects even failed to be shown, therefore the non-detection zone was formed. Considering this problem, an on-line detecting system which could eliminate the non-detection zone was designed. Based on the scattering and the information acquisition mechanism of the defect, the change law of the defect feature light intensity with the lighting direction was obtained. An experimental study on the defects of imprinting and ink was carried out to show that these two types of defect feature light intensity distributions were related to the lighting direction, and the non-detection zone as well as the effective zone of defects presented a orthogonal state. According to this, a self-designed off-axis plane light source with side-entering uniform irradiation was adapted to design the on-line detecting system for the orthogonal composition with linear array and planar array. The vertical target plane was used to collect the information and the dark field orthogonal lighting was carried out at 45°. The main detecting system was composed of the linear-array camera and the linear light source, while the auxiliary detecting system was composed of the planar camera and the off-axis plane light source to detect hundreds of samples of random anisotropic defects. The experimental results show that the defects non-detection zone is effectively eliminated, the defects recognition rate is increased by 22%, and the detection effciency is also greatly improved.

     

/

返回文章
返回