基于机器视觉的包装袋缺陷检测算法研究与应用 下载: 1536次
Machine-Vision Based Defect Detection Algorithm for Packaging Bags
沈阳城市建设学院信息与控制工程系, 辽宁 沈阳 110167
图 & 表
图 1. 机器视觉检测系统架构图
Fig. 1. Architectural diagram of machine-vision based detection system
下载图片 查看原文
图 2. 缺陷检测算法流程图
Fig. 2. Flow chart of defect detection algorithm
下载图片 查看原文
图 3. 二值化图像。(a) T1阈值图像;(b) T2阈值图像
Fig. 3. Binary images. (a) T1 threshold image; (b) T2 threshold image
下载图片 查看原文
图 4. 包装袋实物图
Fig. 4. Picture of packing bag
下载图片 查看原文
图 5. 实验测试平台
Fig. 5. Platform for experimental testing
下载图片 查看原文
图 6. 标准设置模块
Fig. 6. Standard setting module
下载图片 查看原文
图 7. 定位设置模块
Fig. 7. Location setting module
下载图片 查看原文
图 8. 部分缺陷分类检测结果。(a)合格图像;(b)连袋(超长);(c)连袋(超宽);(d)包装版面移动;(e)外形尺寸错误;(f)包装上有异物
Fig. 8. Partial detection results of defect classification. (a) Qualified image; (b) continuous bag (over length); (c) continuous bag (over width); (d) motion of packaging layout; (e) dimension error; (f) foreign matter on packages
下载图片 查看原文
表 1特征与缺陷匹配表
Table1. Feature and defect matching
No. | Condition | Defect classification |
---|
1 | L>Lup or W>Wup | Defect 1: continuous bag | 2 | L>Lup or L<Llow or W>Wup or W<Wlow | Defect 2: dimension error(over length or over width) | 3 | θ<θT | Defect 3: foreign matteron packages | 4 | O⊄M | Defect 4: motion ofpackaging layout |
|
查看原文
表 2混合矩阵
Table2. Confusion matrix
Classification | Detection |
---|
Defectnumber | Qualifiednumber |
---|
Actual | Defect number | PT | NF | Qualified number | PF | NT |
|
查看原文
表 3不同检测方法的混合矩阵
Table3. Confusion matrices for different detection methods
Category | Proposedmethod | Templatematching | Manualdetection |
---|
Defect | Qualified | Defect | Qualified | Defect | Qualified |
---|
Defect | 195 | 5 | 178 | 22 | 186 | 14 | Qualified | 6 | 294 | 25 | 275 | 16 | 284 |
|
查看原文
表 4不同检测方法的真正率、真负率和准确率
Table4. True positive rates, true negative rates, and accuracy of different detection methods
Method | True positiverate /% | True negativerate /% | Accuracy /% |
---|
Proposed method | 97.5 | 98 | 97.8 | Template matching | 89 | 91.7 | 90.6 | Manual detection | 93 | 94.7 | 94 |
|
查看原文
表 5测试分类结果
Table5. Test of classification results
No. | Defecttype | Sample | Successnumber | Missingnumber | Wrongnumber | Missingrate /% | Errorrate /% | Positiverate /% |
---|
1 | Continuous bag | 200 | 199 | 0 | 1 | 0 | 0.5 | 99.5 | 2 | Dimension error | 200 | 199 | 0 | 1 | 0 | 0.5 | 99.5 | 3 | Foreign matter on packages | 200 | 195 | 1 | 4 | 0.5 | 2.0 | 97.5 | 4 | Motion of packaging layout | 200 | 197 | 0 | 3 | 0 | 1.5 | 98.5 | | Total | 800 | 790 | 1 | 9 | 0.125 | 1.125 | 98.75 |
|
查看原文
李丹, 白国君, 金媛媛, 童艳. 基于机器视觉的包装袋缺陷检测算法研究与应用[J]. 激光与光电子学进展, 2019, 56(9): 091501. Dan Li, Guojun Bai, Yuanyuan Jin, Yan Tong. Machine-Vision Based Defect Detection Algorithm for Packaging Bags[J]. Laser & Optoelectronics Progress, 2019, 56(9): 091501.