| 基于深度学习算法的叉车危险操作行为检测 |
| |
| 引用本文： | 李千登,王廷春,崔靖文,穆波.基于深度学习算法的叉车危险操作行为检测[J].中国安全生产科学技术,2020,16(5):155-159. |
| |
| 作者姓名： | 李千登 王廷春 崔靖文 穆波 |
| |
| 作者单位： | （1.中国石油化工股份有限公司青岛安全工程研究院，山东 青岛 266104；2.应急管理部化学品登记中心，山东 青岛 266104） |
| |
| 基金项目： | * 基金项目: 国家重点研发计划项目(2018YFC0809300)；山东省重大科技创新工程项目(2018YFJH0802) |
| |
| 摘 要： | 随着物流仓储行业快速发展及叉车数量的不断增多，针对叉车作业过程中存在的人员碰撞、挤压、坠落等潜在风险，迫切需要对危险行为进行及时检测和预警。为解决人员值守易漏报误报及传统方法检测精度低的问题，建立基于图像特征识别的叉车检测深度学习模型和算法。通过采集、处理现场视频图像素材，完成模型的训练及性能评价，建立相应的报警规则和报警阈值，搭建测试环境并进行仿真测试，开发相应的软件系统。结果表明:模型检测速率为130 ms/帧，人员靠近叉车准确率为85.6%，叉车举升人员准确率为83.7%，达到良好的实践效果。
|
| 关 键 词： | 叉车 危险行为 深度学习算法 图像检测 系统开发 |
Detection on dangerous operation behavior of forklift based on deep learning algorithm |
| |
| LI Qiandeng,WANG Tingchun,CUI Jingwen,MU Bo.Detection on dangerous operation behavior of forklift based on deep learning algorithm[J].Journal of Safety Science and Technology,2020,16(5):155-159. |
| |
| Authors: | LI Qiandeng WANG Tingchun CUI Jingwen MU Bo |
| |
| Affiliation: | (1.SINOPEC Research Institute of Safety Engineering,Qingdao Shandong 266104,China;2.Registration Center for Chemicals of Ministry of Emergency Management,Qingdao Shandong 266104,China) |
| |
| Abstract: | With the rapid development of logistics and warehousing industry,the number of forklifts increases rapidly.Aiming at the potential risk of personnel collision,extrusion and downfall during the operation process of forklifts,the timely detection and early warning of dangerous behavior is urgently required.In order to solve the problem of prone under reporting and false alarm in the personnel on duty and the low detection accuracy of traditional methods,the deep learning model and algorithm of forklift detection based on the image characteristics recognition were established.Through collecting and processing the field video image material,the training and performance evaluation of the model were completed,and the corresponding alarm rules and threshold values were determined.The testing environment was established,then the simulation tests were carried out,and the corresponding software system was developed.The results showed that the detection rate of the model was 130 milliseconds per frame,the accuracy of personnel close to forklift was 85.6%,and the accuracy of forklift lifting personnel was 83.7%,which achieved good practice effect. |
| |
| Keywords: | forklift dangerous behavior deep learning algorithm image detection system development |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《中国安全生产科学技术》浏览原始摘要信息 |
|
点击此处可从《中国安全生产科学技术》下载全文 |
