摘要
近年来,快递行业内“暴力分拣”现象层出不穷,为解决传统依靠人工监控耗时耗力且作用不明显的弊端,提出一种基于计算机视觉的快递暴力分拣行为识别方法,将识别过程分为两个阶段:第一个阶段采用卷积神经网络与长短期记忆网络相结合形成的网络进行分拣人员分拣动作的识别及分类,对于一些常见暴力分拣的动作直接判定;第二阶段,采用三帧差法和Canny边缘检测相结合进行运动包裹的检测,从而获取其运动轨迹,判断运动包裹的移动轨迹是否大于阈值,大于则判定为暴力分拣。文章所提出的方法,能有效提高快递暴力分拣行为识别的实时性及准确率,从而提升快递行业服务质量。
In recent years,the phenomenon of“violent sorting”is constantly seen in the express delivery industry.In order to solve the disadvantages of traditional manual monitoring analysis,which is time-consuming and ineffective,this paper proposes a method based on computer vision to identify the violent sorting behavior.This identification process is mainly divided into two stages:the first stage uses the LRCN network which combines CNN and LSTM to identify and classify the sorter s sorting behaviors.If there is kicking and trampling behavior,it is directly judged as violent sorting.If there is no kicking or trampling behavior,it enters the second stage,and uses the combination of three-frame difference method and Canny edge detection method to detect the moving package,so as to obtain its moving track.If the horizontal or vertical distance of the moving package is greater than the threshold value,it will be determined as violent sorting.The method proposed in this paper can effectively improve the real-timeness and accuracy rate of violent sorting behavior identification,so as to improve the service quality of the express delivery industry.
作者
邓秀琴
何鹏志
倪卫红
赵成国
DENG Xiu-qin;HE Peng-zhi;NI Wei-hong;ZHAO Cheng-guo(School of Economics and Management,Nanjing Tech University,Nanjing,Jiangsu 211816)
出处
《供应链管理》
2021年第6期109-116,共8页
SUPPLY CHAIN MANAGEMENT
基金
江苏省大学生创新创业训练计划项目“视觉识别技术在智慧物流系统中的应用研究--配送分拣领域”(2020DC0656)。
