合肥工业大学学报自科版

合肥工业大学学报(自然科学版)

JOURNAL OF HEFEI UNIVERSITY OF TECHNOLOGY (NATURAL SCIENCE)

主管：中华人民共和国教育部

主办：合肥工业大学

ISSN 1003-5060, CN 34-1083/N

基于改进 YOLOv3 的输送带纵向撕裂多视角检测方法

下载全文

Multi-view detection method for longitudinal tear of conveyor belt based on improved YOLOv3

期刊信息

合肥工业大学（自然科学版），2023年1月，第46卷第1期：28-35，80

DOI: 10.3969/j.issn.1003-5060.2023.01.005

作者信息

王文善 $ ^{1,2} $，郭永存 $ ^{1,2,3,4} $，刘普壮 $ ^{1,2} $，杨豚 $ ^{1,2} $，童佳乐 $ ^{1,2} $

（1. 安徽理工大学深部煤矿采动响应与灾害防控国家重点实验室，安徽淮南 232001；2. 安徽理工大学机械工程学院，安徽淮南 232001；3. 矿山智能技术与装备省部共建协同创新中心，安徽淮南 232001；4. 安徽理工大学矿山智能装备与技术安徽省重点实验室，安徽淮南 232001）

摘要和关键词

摘要: 针对输送带纵向撕裂检测中存在的检测视角单一、速度慢、精度低等问题，文章提出一种基于改进YOLOv3算法的输送带纵向撕裂多视角检测方法。首先对原始YOLOv3网络结构进行优化设计，采用29层网络模型(Darknet-29)作为特征提取网络，将原有的3种不同尺度锚点改用为2种不同尺度 $ (26\times26,52\times52) $锚点；将位于多视角检测点的工业相机所采集的纵向撕裂图像制作成数据集，使用K-means算法对输送带纵向撕裂标签进行维度聚类分析，确定先验框参数；最后将改进的YOLOv3算法在数据集上进行测试与训练，并与其他几种算法进行比较。实验结果表明：该检测方法不仅可以较好地检测出输送带纵向撕裂，还可以分类识别出大裂纹或完全撕裂情形；相较于原始YOLOv3算法，改进后的YOLOv3算法平均检测精度均值提高0.4%，达到98.7%，检测速度提高60.6%，达到53帧/s，模型占用内存减少93Mb，仅为141Mb，优于YOLOv2和YOLOv3-Tiny算法。该文提出的输送带纵向撕裂检测方法具有模型占用内存低、检测精度高及速度快等优点，为输送带纵向撕裂提供了一种新的检测方案。

关键词: YOLOv3算法；纵向撕裂；多视角；Darknet-29网络模型；K-means聚类

Authors

WANG Wenshan $ ^{1,2} $, GUO Yongcun $ ^{1,2,3,4} $, LIU Puzhuang $ ^{1,2} $, YANG Tun $ ^{1,2} $, TONG Jiale $ ^{1,2} $

(1, State Key Laboratory of Mining Response and Disaster Prevention and Control in Deep Coal Mines, Anhui University of Science and Technology, Huainan 232001, China; 2. School of Mechanical Engineering, Anhui University of Science and Technology, Huainan 232001, China; 3. Collaborative Innovation Center for Mining Intelligent Technology and Equipment, Huainan 232001, China; 4. Anhui Key Laboratory of Mine Intelligent Equipment and Technology, Anhui University of Science and Technology, Huainan 232001, China)

Abstract and Keywords

Abstract: In order to solve the problems of single view, slow speed and low accuracy in longitudinal tear detection of conveyor belt, multi-view detection for longitudinal tear of conveyor belt based on improved YOLOv3 algorithm is proposed. Firstly, the original YOLOv3 network structure was optimized and the Darknet-29 network model was used as the feature extraction network. The original three anchor points with different scales were replaced with two anchor points with different scales $ (26 \times 26, 52 \times 52) $. Secondly, the longitudinal tear image collected by the industrial camera located at the multi-view detection point was made into a data set. K-means algorithm was used to conduct dimensional clustering analysis on the longitudinal tear label of conveyor belt, and the prior box parameters were determined. Finally, the improved YOLOv3 algorithm was tested and trained on the data set, and compared with other algorithms. The experimental results show that this method can not on-

Keywords: YOLOv3 algorithm; longitudinal tear; multi-view; Darknet-29 network model; K-means clustering

基金信息

国家重点研发计划资助项目(2020YFB1314203); 国家自然科学基金资助项目(51874004; 51904007)和安徽省自然科学基金资助项目(1908085QE227)

前一篇后一篇

返回期刊目录