输送带跑偏故障检测是煤矿传输设备运维的关键环节。深度学习凭借其强大的特征提取能力，显著提高了故障检测的效率，但是在实际的跑偏检测应用中，现有的方法仅限于定性的根据输送带边缘位置判断跑偏故障。基于此，论文致力于解决输送带跑偏方向判断及跑偏程度量化的问题，为纠偏提供数据支撑。主要研究工作如下：

（1）针对煤矿复杂环境导致图像细节模糊、特征信息丢失的问题，提出采用图像增强技术DehazeNet构建高质量的带式输送机图像。首先，选取暗通道和颜色衰减等先验知识与DehazeNet网络相结合，更有针对性地捕捉图像煤尘的相关特征，有效地估计图像透射率，然后经过大气模型处理恢复清晰的图像，完成数据集的构建。

（2）针对托辊中心特征不明显的问题，提出一种强化中心感知的托辊检测模型。首先，用全局信息调节机制整合全局上下文和局部角区域的信息，加强对托辊特征的提取；然后通过扩展感受野，增强对中心区域的感知能力，有助于获取托辊的中心关键点位置；最后，利用多组托辊的中心点位置信息计算出跑偏检测点和输送机中心线。实验结果表明，改进模型与对照模型相比MAP提升了3.42%。

（3）针对输送带边缘形态弱纹理导致分割精度不高的问题，提出三分支语义分割模型，在双分支网络的基础上，添加专注于边缘细化的检测分支，精确地捕捉输送带的边缘信息，并通过并行聚合模块以边缘引导三分支特征的有效融合；采用优化的单像素跟踪策略对分割后的输送带边缘线进行精细化提取矫正；最后通过中轴变换提取输送带中轴线。实验结果表明，三分支分割模型与改进之前相比mIoU提升了5.21%。

根据用户需求，研发了一套带式输送机跑偏智能检测系统，通过计算以输送带中轴线相对于输送机中心线上跑偏检测点的位置关系，实现对跑偏方向的准确判断，以及用像素比例尺对跑偏量大小进行精确计算，最后对输送带运行状态进行可视化展示，提升了输送带跑偏故障检测的智能化水平。

The detection of conveyor belt runout faults is a key link in the operation and maintenance of coal mine transmission equipment. Deep learning has significantly improved the efficiency of fault detection by virtue of its robust feature extraction capability, however, in the actual runout detection application, the existing methods are limited to qualitatively judging the runout fault based on the position of the conveyor belt edge. Based on this, the thesis is dedicated to solving the problems of determining the direction of conveyor belt runout and quantifying the degree of runout, which provides data support for deviation correction. The main research work is as follows:

Aiming at the problem of blurred image details and loss of feature information due to the complex environment of coal mines, it is proposed to construct high-quality belt conveyor images using the image enhancement technology DehazeNet. Firstly, the a priori knowledge such as dark channel and color attenuation is selected and combined with DehazeNet network to capture the relevant features of image coal dust in a more targeted way and estimate the image transmittance efficiently, and then clear images are recovered after atmospheric model processing to complete the construction of the dataset.

Aiming at the problem that the center features of the rollers are not clear, a roller detection model with enhanced center perception is proposed. Firstly, the global information adjustment mechanism is used to integrate the information of global context and local corner region to strengthen the extraction of the features of the rollers; then the perception of the center region is enhanced by expanding the sensing field, which helps to obtain the location of the center key point of the rollers; finally, the runout detection point and the centerline of the conveyor were calculated by using the information of the center point location of multiple groups of rollers. The experimental results show that the improved model improves the MAP by 3.42% compared with the control model.

Aiming at the problem of low segmentation accuracy due to the weak texture of conveyor belt edge morphology, a three-branch semantic segmentation model is proposed, where a detection branch focusing on edge refinement is added based on a two-branch network that accurately captures the edge information of the conveyor belt, and the effective fusion of the three-branch features is guided by edges through the parallel aggregation module; an optimized single-pixel tracking strategy is adopted to correct the segmented conveyor belt edge line by The optimized single-pixel tracking strategy is used to extract and correct the segmented conveyor belt edge lines; finally, the conveyor belt center axis is extracted by the center-axis transformation. The experimental results show that the mIoU of the three-branch segmentation model is improved by 5.21% compared with that before the improvement.

According to the user's demand, a set of belt conveyor runout intelligent detection system is developed, which improves the intelligent level of conveyor runout fault detection by calculating the positional relationship of the runout detection point with the central axis of the conveyor belt relative to the centerline of the conveyor, realizing the accurate judgment of the direction of runout, as well as accurately calculating the size of runout with a pixel scale, and finally displaying the conveyor belt running status visually.

[1]卜令超.长距离带式输送机运行过程建模与优化控制[D].中国矿业大学,2022. 000251..

[2]陈晓晶. 井工煤矿运输系统智能化技术现状及发展趋势[J]. 工矿自动化, 2022, 48(06): 6-14+35.

[3]武林海. 带式输送机常见故障检测及防治系统研究[J]. 煤矿机械, 2019, 40(02): 145-147.

[4]王宏武. 煤矿带式输送机常见故障分析及防范措施[J]. 矿业装备, 2023, (02): 162-164.

[5]刘峰, 曹文君, 张建明. 持续推进煤矿智能化促进我国煤炭工业高质量发展[J]. 中国煤炭, 2019, 45(12): 32-36.

[6]张宇飞. 矿用带式输送机纠偏装置优化改进[J]. 矿业装备, 2023, (12): 158-160.

[7]李伟. 矿用带式输送机检测与纠偏装置设计及应用[J]. 机械管理开发, 2023, 38(09): 98-99+125.

[8]石斌. 矿用带式输送机机械联动纠偏装置设计及应用[J]. 机械管理开发, 2023, 38(08): 151-152.

[9]张飞, 朱锋, 田圣彬, 等. 基于图像处理技术的输送带跑偏移动监测系统设计[J]. 矿山机械, 2021, 49(11):16-19.

[10]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(05): 152-156.

[11]Dąbek P, Wróblewski A, Wodecki J, et al. Application of the methods of monitoring and detecting the belt mistracking in laboratory conditions[J]. Applied Sciences,2023,13(04): 2111.

[12]Liu Y, Miao C, Li X, et al. Research on deviation detection of belt conveyor based on inspection robot and deep learning[J]. Complexity, 2021: 1-15.

[13]Xu C, Zeng X, Zhang R, et al. Detection method of edge position of belt conveyor based on complex environment[C]//2021 4th International Conference on Robotics, Control and Automation Engineering (RCAE). IEEE, 2021: 417-422.

[14]Wang J, Liu Q, Dai M. Belt vision localization algorithm based on machine vision and belt conveyor deviation detection[C]//2019 34rd Youth Academic Annual Conference of Chinese Association of Automation (YAC). IEEE, 2019: 269-273.

[15]Kim J, Lee M. Robust lane detection based on convolutional neural network and random sample consensus[C]//Neural Information Processing: 21st International Conference, ICONIP 2014, Kuching, Malaysia, November 3-6, 2014. Proceedings, Part I 21. Springer International Publishing, 2014: 454-461.

[16]Wu L, Li J, Zhang H, et al. Research on a system for the diagnosis and localization of co-nveyor belt deviations in belt conveyors[J]. Measurement Science and Technology, 2023, 35(03): 035110.

[17]刘吉乔, 李敬兆, 石晴, 等. 基于机器视觉的带式输煤机智能纠偏系统设计[J]. 科技与创新, 2023, (17): 48-50.

[18]张飞, 朱锋, 田圣彬, 等. 基于图像处理技术的输送带跑偏移动监测系统设计[J]. 矿山机械, 2021, 49(11): 16-19.

[19]陈修虎. 皮带运输机安装调试常见故障分析与处理[J]. 中国科技博览, 2011(35): 2.

[20]贾东. 皮带机用防皮带跑偏装置的分析与应用[J]. 机械管理开发, 2021, 36(02): 76-77.

[21]杨彦利, 苗长云, 亢伉, 等. 输送带跑偏故障的机器视觉检测技术[J]. 中北大学学报(自然科学版), 2012, 33(06): 667－671.

[22]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.

[23]Zhang M, Jiang K, Cao Y, et al. A deep learning-based method for deviation status detect-ion in intelligent conveyor belt system[J]. Journal of Cleaner Production, 2022, 363: 132575.

[24]Liu Y, Wang Y, Zeng C, et al. Edge detection for conveyor belt based on the deep convol-utional network[C]//Proceedings of 2018 Chinese Intelligent Systems Conference: Volu-me II. Springer Singapore, 2019: 275-283.

[25]韩涛, 黄友锐, 张立志, 等. 基于图像识别的带式输送机输煤量和跑偏检测方法[J]. 工矿自动化, 2020, 46(04): 17-22.

[26]曾飞, 陶玉衡, 苏俊彬, 等. 融合ResNet18和Deconvolution的输送带横向跑偏检测方法[J]. 现代制造工程, 2023, (08): 121-126.

[27]赵光辉, 赵鹏, 胡金良, 等. 语义分割的传送带跑偏视频检测算法[J]. 中国安全科学学报, 2023, 33(S1): 81-84.

[28]杨志方, 张立亚, 郝博南, 等. 基于双流融合网络的运输机皮带跑偏检测方法[J/OL]. 煤炭科学技术, 1-10[2024-03-27].

[29]Viola P, Jones M. Rapid object detection using a boosted cascade of simple features[C]// Proceedings of the 2001 IEEE computer society conference on computer vision and pattern recognition. CVPR 2001. IEEE, 2001, 1: I-I.

[30]Dalal N, Triggs B. Histograms of oriented gradients for human detection[C]//2005 IEEE computer society conference on computer vision and pattern recognition (CVPR'05). IEEE, 2005, 1: 886-893.

[31]Lowe D G. Distinctive image features from scale-invariant keypoints[J]. International journal of computer vision, 2004, 60: 91-110.

[32]Ren S, He K, Girshick R, et al. Faster R-CNN: Towards real-time object detection with region proposal networks[J]. IEEE transactions on pattern analysis and machine intelligence, 2016, 39(06): 1137-1149.

[33]Redmon J, Divvala S, Girshick R, et al. You only look once: Unified, real-time object detection[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2016: 779-788.

[34]Huang L, Yang Y, Deng Y, et al. Densebox: Unifying landmark localization with end to end object detection[J]. arXiv preprint arXiv:1509.04874, 2015.

[35]Wang J, Chen K, Yang S, et al. Region proposal by guided anchoring[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2019: 2965-2974.

[36]Kong T, Sun F, Liu H, et al. Foveabox: Beyound anchor-based object detection[J]. IEEE Transactions on Image Processing, 2020, 29: 7389-7398.

[37]Law H, Deng J. Cornernet: Detecting objects as paired keypoints[C]//Proceedings of the European conference on computer vision (ECCV). 2018: 734-750.

[38]Duan K, Bai S, Xie L, et al. Centernet: Keypoint triplets for object detection[C]//Proceed-ings of the IEEE/CVF international conference on computer vision. 2019: 6569-6578.

[39]Sobel I, Feldman G. A 3x3 isotropic gradient operator for image processing[J]. a talk at the Stanford Artificial Project in, 1968: 271-272.

[40]Prewitt J M S. Object enhancement and extraction[J]. Picture processing and Psychopictorics, 1970, 10(01): 15-19.

[41]Marr D, Hildreth E. Theory of edge detection[J]. Proceedings of the Royal Society of London. Series B, Biological sciences, 1980, 207(1167): 187-217.

[42]Canny J .A Computational Approach to Edge Detection[J].IEEE Transactions on Pattern Analysis and Machine Intelligence, 1986, PAMI-8(06):679-698.

[43]Long J, Shelhamer E, Darrell T. Fully convolutional networks for semantic segmentation[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2015: 3431-3440.

[44]Badrinarayanan V, Kendall A, Cipolla R. Segnet: A deep convolutional encoder-decoder architecture for image segmentation[J]. IEEE transactions on pattern analysis and machine intelligence, 2017, 39(12): 2481-2495.

[45]Yu C, Gao C, Wang J, et al. Bisenet v2: Bilateral network with guided aggregation for real-time semantic segmentation[J]. International Journal of Computer Vision, 2021, 129: 3051-3068.

[46]宋亮, 谷玉海, 石文天, 等. 基于改进BiSeNet的非结构化道路分割算法研究[J]. 应用光学, 2023, 44(03): 556-564.

[47]Land E H, McCann J J. Lightness and retinex theory[J]. Josa, 1971, 61(01): 1-11.

[48]Chen Z, Wang Y, Yang Y, et al. PSD: Principled synthetic-to-real dehazing guided by phy-sical priors[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2021: 7180-7189.

[49]Zhu Q, Mai J, Shao L. A Fast Single Image Haze Removal Algorithm Using Color Atten-uation Prior. IEEE Trans Image Process. 2015 Nov;24(11):3522-33.

[50]He K, Sun J, Tang X. Single image haze removal using dark channel prior[J]. IEEE trans-actions on pattern analysis and machine intelligence, 2010, 33(12): 2341-2353.

[51]Cai B, Xu X, Jia K, et al. Dehazenet: An end-to-end system for single image haze remov-al[J]. IEEE transactions on image processing, 2016, 25(11): 5187-5198.

[52]熊超, 周海峰, 郑东强, 等. 结合空洞编码器和特征金字塔的中心点船舶检测[J]. 船舶工程, 2023, 45(02): 154-161.

[53]张硕, 刘禹, 熊坤, 等. 基于特征工程的大田作物行中心线识别方法[J]. 农业机械学报, 2023, 54(S1): 18-26.

[54]Hong Y, Pan H, Sun W, et al. Deep dual-resolution networks for real-time and accurate semantic segmentation of road scenes[J]. arXiv preprint arXiv:2101.06085, 2021.

[55]刘浩, 任宏, 赵丁选, 等. 基于亚像素定位的图像边缘检测策略研究[J/OL]. 农业机械学报, 1-8[2024-03-10].

[56]朱厚盛, 朱春元, 鲍宪帅, 等. 基于中轴变换的参数化图形构造方法[J]. 计算机应用与软件, 2021, 38(10): 234-241.

[57]王锴, 曾祥进, 黎新, 等. 输送带跑偏检测方法研究[J]. 工矿自动化, 2023, 49(03): 23-30+52.