复杂场景下传送带实时偏移检测  

Real-Time Detection Algorithm for Conveyor Belt Offset in Complex Scenarios

作　　者：宫法明[1] 兰光诚 牛博 GONG Faming;LAN Guangcheng;NIU Bo(College of Computer Science and Technology,China University of Petroleum(East China),Qingdao,Shandong 266580,China;iFLYTEK Co.,Ltd.,Qingdao,Shandong 266400,China)

机构地区：[1]中国石油大学(华东)计算机科学与技术学院,山东青岛266580 [2]科大讯飞股份有限公司,山东青岛266400

出　　处：《计算机工程与应用》2025年第5期269-278,共10页Computer Engineering and Applications

基　　金：山东省自然科学基金(ZR2020MF136,ZR2020MF140)。

摘　　要：传送带在工业生产线和物流系统中扮演着重要的角色,然而传送带的偏移可能导致生产效率下降、质量损失以及生产线堵塞等问题。针对传统传送带偏移检测算法存在特征提取困难、缺乏通用性、实时性差等问题,提出了一种基于REO的传送带实时偏移检测算法,该算法主要由三部分组成:传送带区域提取、边缘线检测、偏移量计算。通过提出的UNet-CRFs网络从视频图像中精确分割出传送带区域,极大地降低了背景干扰对检测结果的影响,同时增加特征增强与区域提取模块细化分割结果;相较于传统方法只检测传送带外侧两条边缘线,提出的边缘线检测算法(CH-LaneNet)分别对传送带及物料区域的边缘进行提取,同时识别并消除冗余的边缘线;设计了一种偏移判定策略,通过几何方法计算偏移量并构建数学模型,实现传送带偏移检测,避免了传统方法中易发生的图像偏移失真等情况。实验结果表明,所提出的方法在传送带区域提取的像素精度达到了98.21%,边缘线检测任务的像素精度达到96.49%,F1分数为80.40%,偏移量算法的AUC为85%,证明提出的方法对传送带偏移检测具有有效性。Conveyor belts play an important role in industrial production lines and logistics systems,however,conveyor belt offset may lead to productivity loss,quality loss,and production line blockage.Traditional conveyor belt offset detection algorithms have problems such as difficult feature extraction,lack of generality,and poor real-time performance.This paper proposes a REO-based real-time conveyor offset detection algorithm,which consists of three main components:region extraction,edge line detection,and offset calculation.The proposed UNet-CRFs network accurately segments the conveyor belt region from the video image,which greatly reduces the influence of background interference on the detection results.Additionally,it adds feature enhancement and region extraction modules to refine the segmentation results.Compared to the traditional method that detects only two edge lines on the outside of the conveyor belt,the proposed edge line detection algorithm(CH-LaneNet)extracts the edges of the conveyor belt and the material region respectively,and identifies and eliminates the redundant edge lines;an offset determination strategy is designed to calculate the offset amount by geometric methods and construct a mathematical model by geometric methods to realize conveyor belt offset detection,which avoids the distortion of image offset that is prone to occur in the traditional methods.The experimental results show that the proposed method achieves 98.21%pixel accuracy in conveyor belt region extraction,96.49%pixel accuracy in edge line detection task,80.40%F1 score,and 85%AUC of offset algorithm,which proves the effectiveness of the proposed method for conveyor belt offset.

关 键 词：传送带 语义分割 特征提取 偏移检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]