基于机器视觉的泡罩药板穿泡缺陷检测方法

Defect Detection Method of Drugs Blister Board Based on Machine Visions

为提高工业生产线上对药板穿泡缺陷的检测速度与准确度,提出一种实用性算法。该算法采用级联检测方法,提出一种基于搜索子图像灰度差算法来实现缺陷粗定位,同时提出基于Gabor滤波的纹理抑制改进算法进行纹理抑制,通过对频谱能量分析,设计带有方向性的高效带通滤波器,实现缺陷图像重构,最终采用选择分块迭代阈值法实现阈值分割,成功定位残缺位置。该方法通过MATLAB仿真,试验结果表明,缺陷检测准确率达95.29%。通过与传统方法对比,该方法具有更高的准确度和可行性。

A practical algorithm was proposed in order to improve the detection speed and accuracy of the bubble defect detection on the industrial production line. Firstly,based on the cascade detection method,a sub-image gray-scale difference algorithm was proposed to locate defects roughly.Then,an improved texture suppression algorithm based on Gabor filtering was proposed. Then, a high-efficiency bandpass filter with directionality was designed through analysis of the spectral energy to realize reconstruction of defect image. Finally,the selective block iterative threshold method was used to realize threshold segmentation,so that the defective position was located successfully. Through image test of a few tablets by MATLAB,the detection accuracy of defects reached 95.29%.Compared with the traditional method,this method has higher accuracy and feasibility.