基于机器视觉编织袋缝合缺陷识别与检测

Recognition and detection of stitching defects of woven bags based on machine vision

针对包装生产线编织袋密封检测技术的缺乏,设计一种基于机器视觉的编织袋缝合质量智能在线检测系统。分析编织袋缝合的6种主要缺陷类型及其对应特征信息。通过二值化、去噪、边缘检测算法计算出编织袋倾斜角度进行图像校正。校正后二值图像进行列求和,得到留线和中间缝线裁剪位置,缝线二值图像按裁剪位置进行裁剪得到3部分图像。提取对应数字特征,使用机器学习算法对特征进行类型判别。结果表明,使用集成模型中RUSBoostedTrees算法的缺陷分类准确率为97.8%,平均检测时间约为0.652 s,检测效率超过缝合机生产效率。智能检测系统设计合理,满足编织袋袋口缝合质量智能检测的需求,为机器视觉缺陷检测技术在包装行业的智能化进一步应用提供技术支持。

Considering the lack of sealing detection technology for woven bags in packaging production lines,an intelligent online detection system was designed for sewing quality of woven bags based on machine vision.6 main defect types and their corresponding characteristic information of woven bag sewing were analyzed.The inclination angle of the woven bag was calculated by binarization,denoising and edge detection algorithms for image correction.After the correction,the binary image was summed in columns to obtain the cutting position of the remaining line and the intermediate suture,and the suture binary image was cut according to the cutting position to obtain three parts of the image.The corresponding digital features were extracted,and machine learning algorithms were used to perform type discrimination of the features.The results show that the defect classification accuracy using the RUSBoostedTrees algorithm in the ensemble model is 97.8%,the average detection time is about 0.652 s,and the detection efficiency exceeds the production efficiency of the sewing machine.The intelligent inspection system is reasonably designed and can meet the needs of intelligent detection of the stitching quality of the woven bag mouth,which provides technical support for the further intelligent application of machine vision defect detection technology in the packaging industry.