基于机器视觉与图像分割的工件表面残胶识别算法

Target recognition algorithm of residual glue on workpiece surface based on machine vision and image segmentation

为了解决当前工件残胶因胶特征不规则和分布随机而导致人工及机器检查均存在漏检的问题,提出了基于机器视觉与图像分割的工件表面残胶识别算法。首先,融合最大类间阈值分割和形态学处理,设计一个残胶目标分割模型,得到若干包含残胶的区域。然后,基于连通区域与轮廓查找,进一步缩小目标区域范围,根据几何与面积特征精确定位残胶位置。最后,根据视觉打光原理,有机组合工业相机、镜头、光源、支架和平台,完成视觉选型,编码实现系统的功能测试。实验验证:与当前主流缺陷识别技术相比,算法拥有更高的检出能力与稳定性。

In order to solve the problem of the omission of manual and machine inspection induced by the irregular and random distribution of residual glue features,the recognition algorithm of residual glue on workpiece surface based on machine vision and image segmentation is proposed.Firstly,a residual object segmentation model is designed by fusing the maximum inter-class threshold segmentation and morphological processing methods to obtain several regions containing residual rubber.Then,the connected area and contour search are used to reduce the target area,and the residue location is accurately located according to the geometric and area characteristics.Finally,according to the principle of visual lighting,the visual hardware system is composed of industrial cameras,lenses,light sources,brackets and platforms,and the functional testing of this system is tested by coding.The experimental results show that compared with the current mainstream defect recognition technology,the proposed algorithm has higher detection ability and stability.