复杂场景下传送带实时偏移检测

doi:10.3778/j.issn.1002-8331.2310-0371

摘要/Abstract

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。

关键词: 传送带, 语义分割, 特征提取, 偏移检测

Abstract: 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.

Key words: conveyor belt, semantic segmentation, feature extraction, offset inference

宫法明, 兰光诚, 牛博. 复杂场景下传送带实时偏移检测[J]. 计算机工程与应用, 2025, 61(5): 269-278.

GONG Faming , LAN Guangcheng, NIU Bo. Real-Time Detection Algorithm for Conveyor Belt Offset in Complex Scenarios[J]. Computer Engineering and Applications, 2025, 61(5): 269-278.

参考文献

[1] HOU C, QIAO T, QIAO M, et al. Research on audio visual detection method for conveyor belt longitudinal tear[J]. IEEE Access, 2019, 7: 120202-120213.
[2] LI Z, ZHU X, ZHOU J. Intelligent monitoring system of coal conveyor belt based on computer vision technology[C]//Proceedings of the 6th International Conference on Dependable Systems and Their Applications, 2019: 359-364.
[3] MIAO D, WANG Y, YANG L, et al. Foreign object detection method of conveyor belt based on improved nanodet[J]. IEEE Access, 2023, 11: 23046-23052.
[4] GUO X, LIU X, SULOWICZ M, et al. Damage detection for convey or belt surface based on conditional cycle generative adversarial network[J]. Sensors: 2022, 22(9): 3485.
[5] 谭恒, 张红娟, 靳宝全, 等. 基于机器视觉的煤矿带式输送机跑偏检测方法[J]. 煤炭技术, 2021, 40(5):152-156.
TAN H, ZHANG H J, JIN B Q, et al. Method for detecting deviation of coal mine belt conveyor based on machine visiona[J]. Coal Technology, 2021, 40(5): 152-156.
[6] CAO Y, WU D, DUAN Y. A new image edge detection algorithm based on improved canny[J]. Journal of Computational Methods in Sciences and Engineering, 2020, 20(2): 629-642.
[7] ILLINGWORTH J, KITTLER J. A survey of the hough transform[J]. Computer Vision, Graphics, and Image Processing, 1988, 44(1): 87-116.
[8] LIU Y, MIAO C, LI X, XU G. Research on deviation detection of belt conveyor based on inspection robot and deep learning[J]. Complex, 2021, 2021: 1-15.
[9] YANG Y, MIAO C, LI X, MEI X. On-line conveyor belts inspection based on machine vision[J]. Optik, 2014, 125(19): 5803-5807.
[10] KLIPPEL E, OLIVEIRA R A R, MASLOV D, et al. Embedded edge artificial intelligence for longitudinal rip detection in conveyor belt applied at the industrial mining environment[J]. SN Computer Science, 2022, 3(4): 280.
[11] ZENG C, ZHENG J, LI J. Real-time conveyor belt deviation detection algorithm based on multi-scale feature fusion network[J]. Algorithms, 2019, 12(10): 205.
[12] LI X, SU J, YUE Z, et al. Adaptive multiroi agricultural robot navigation line extraction based on image semantic segmentation[J]. Sensors, 2022, 22(20): 7707.
[13] LIU Y, WANG Y, ZENG C, et al. Edge detection for conveyor belt based on the deep convolutional network[C]//Proceedings of the Chinese Intelligent Systems Conference, 2019: 275-283.
[14] RONNEBERGER O, FISCHER P, BROX T. U-Net: convolutional networks for biomedical image segmentation[J]. arXiv:1505.04597, 2015.
[15] WANG Y, MA X, CHEN Z, et al. Symmetric cross entropy for robust l-earning with noisy labels[J]. arXiv:1908.06112,2019.
[16] LAFFERTY J D, MCCALLUM A, PEREIRA F C N. Conditional random fields: probabilistic models for segmenting and labeling sequence data[C]//Proceedings of the 21st International Conference on Machine Learning, 2001: 282-289.
[17] KR?HENBüHL P, KOLTUN V. Efficient inference in fully connected CRFs with Gaussian edge potentials[J]. arXiv:1210. 5644, 2012.
[18] DENG Q, LI X, NI P, et al. ENet-CRF-lidar: lidar and camera fusion for multi-scale object recognition[J]. IEEE Access, 2019, 7: 174335-174344.
[19] HU J, GU X, WU Y, et al. RoI uncertainty for RGB-thermal image segmentation[C]//Proceedings of the 5th International Conference on Machine Learning and Machine Intelligence, 2022: 132-139.
[20] NEVEN D, BRABANDERE B D, GEORGOULIS S, et al. Towards end-to-end lane detection: an instance segmentation approach[J]. arXiv:1802.05591, 2018.
[21] DENG F, FENG H, LIANG M, et al. FEANet: feature-enhanced attention network for RGB-thermal real-time semantic segmentation[J]. arXiv:2110.08988, 2021.
[22] GAO X, BAI H, XIONG Y, et al. Robust lane line segmentation based on group feature enhancement[J]. Engineering Applications of Artificial Intelligence, 2023, 117: 105568.
[23] WOO S, PARK J, LEE J, et al. CBAM: convolutional block attention module[J]. arXiv:1807.06521, 2018.
[24] LIN T, GOYAL P, GIRSHICK R B, et al. Focal loss for dense object detection[J]. arXiv:1708.02002, 2017.
[25] TEUTSCH M, TRANTELLE P, BEYERER J. Adaptive real-time image smoothing using local binary patterns and Gaussian filters[C]//Proceedings of the IEEE International Conference on Image Processing, 2013: 1120-1124.
[26] ZHANG J, ZHANG L, TENG Y, ZHANG X, et al. Interactive binary image segmentation with edge preservation[J]. arXiv:1809.03334, 2018.
[27] JASSIM F A, ALTAANI F H. Hybridization of Otsu method and median filter for color image segmentation[J]. arXiv:1305.1052, 2013.
[28] KIO O G, SHARK L. Geometric distortion correction for projections on non-planar and deformable surfaces based on displacement of peripheral image points[J]. International Journal Computer Application Technology, 2022, 68(2): 101-113.
[29] QIN Z, WANG H, LI X. Ultra fast structure-aware deep lane detection[J]. arXiv:2004.11757, 2020.
[30] JI X, GONG F, YUAN X, et al N. A high-performance framework for personal protective equipment detection on the offshore drilling platform[J]. Complex & Intelligent Systems, 2023, 9: 5637-5652.
[31] PASZKE A, CHAURASIA A, KIM S, et al. ENet: a deep neural network architecture for real-time semantic segmentation[J]. arXiv:1606.02147, 2016.
[32] CHEN L, ZHU Y, PAPANDREOU G, et al. Encoder-decoder with atrous separable convolution for semantic image segmentation[J]. arXiv:1802.02611, 2018.