矿用电子封条是矿业安全开采“互联网+”监管的关键技术手段。为严厉打击非法组织生产行为，加快推进“互联网+”监管应用，国家矿山安全监察局全面开展煤矿“电子封条”推广建设。在煤矿主井口、副井口、调度室和货运车辆出入口等关键地点安装摄像机、图像分析终端等设备，利用智能化视频识别等技术，实现全天候远程监管。针对矿用电子封条的关键技术，论文主要研究工作为：

（1）针对井下摄像头容易被遮挡或角度偏移导致无法正常拍摄，提出一种基于机器视觉的摄像头遮挡和角度偏移检测算法。对于摄像头遮挡检测，首先计算待检图像的梯度，再根据图像梯度的范数判断摄像头是否有遮挡事件；对于摄像头角度偏移检测，分别计算正常图像和待检图像的梯度，利用梯度的范数计算摄像头角度偏移系数，实现摄像头遮挡或角度偏移检测。

（2）针对井下运动目标相互遮挡以及运动模糊导致无法准确检测目标，提出一种基于光流场补偿的视频目标检测算法。首先通过浅层特征提取模块获取相邻帧图像的特征，同时估计相邻帧之间的光流场；然后根据光流场实现相邻帧特征的对齐，并进行相邻帧特征融合；最后将融合的特征送入深层特征提取模块，进行精准目标检测。经过实验验证，基于光流场补偿的视频目标检测算法在矿用电子封条数据集上具有较好的性能表现，相对于改进前网络mAP提升了0.4%。

（3）针对深度学习模型部署到嵌入式平台难以达到在线实时推理，提出一种轻量化深度学习模型部署方案。首先根据模型通道的侧重对通道进行裁剪，再对模型进行优化训练得到调优模型，最后对模型参数进行量化处理，从而实现模型轻量化；将轻量化模型部署到HUAWEI Atlas 200嵌入式平台上，实现在线推理，处理速度达到19.8帧/s。

依据应用需求，研发一套矿用电子封条系统。实现对摄像头遮挡和角度偏移、主井口输送带作业状态、副井口人员以及货场出口车辆实时监测，异常状况时输出报警信号。

Mining electronic seal is the mining safety mining "Internet +" supervision of the key technical means. In order to crack down on the illegal organization of production behavior, accelerate the "Internet +" supervision and application, the National Mine Safety Supervision Bureau to carry out a comprehensive coal mine "electronic seal" to promote the construction. In the coal mine main entrance, vice entrance, scheduling room and freight vehicles entrance and exit and other key places to install cameras, image analysis terminals and other equipment, the use of intelligent video recognition and other technologies to achieve all-weather remote supervision. Aiming at the key technologies of mining electronic seals, the main research works of the thesis are:

(1) Aiming at the underground cameras which are easily blocked or angle shifted resulting in the failure to shoot normally, a machine vision based camera blocking and angle shifting detection algorithm is proposed. For camera occlusion detection, the gradient of the image to be examined is calculated first, and then the camera is judged to have an occlusion event or not according to the parametric number of the image gradient; for camera angle offset detection, the gradient of the normal image and the image to be examined are calculated separately, and the camera angle offset coefficient is calculated using the parametric number of the gradient to realize camera occlusion or angle offset detection.

(2) A video target detection algorithm based on optical flow field compensation is proposed for downhole motion targets that cannot be detected accurately due to mutual occlusion and motion blur. Firstly, the shallow feature extraction module obtains the features of adjacent frames and estimates the optical flow field between adjacent frames; then, it realizes the alignment of adjacent frame features according to the optical flow field and performs adjacent frame feature fusion; finally, the fused features are sent to the deep feature extraction module for accurate target detection. After experimental validation, the video target detection algorithm based on optical flow field compensation has a better performance on the mining electronic seal dataset, with a 0.4% improvement in mAP relative to the network before improvement.

(3) A lightweight deep learning model deployment scheme is proposed for deep learning models deployed to embedded platforms that are difficult to achieve online real-time inference. Firstly, the channels are trimmed according to the focus of the model channels, then the model is optimized and trained to get the tuned model, and finally the model parameters are quantized to realize the lightweight model; the lightweight model is deployed to the HUAWEI Atlas 200 embedded platform to realize online inference, and the processing speed reaches 19.8 frames/s.

Develop a mining electronic sealing system based on application requirements. Realize real-time monitoring of camera blocking and angle offset, conveyor belt operation status at the main shaft entrance, personnel at the secondary shaft entrance and vehicles at the exit of the cargo yard, and output alarm signals in case of abnormal conditions.

[1] 郭建生. 煤矿“电子封条”安装与应用分析[J]. 电视技术, 2022, 46(08):185-187.

[2] 杨传印, 王春素. 煤矿“电子封条”智能监管技术研究[J]. 采矿技术, 2021, 21(S1):140-142.

[3] 郭建生. 煤矿“电子封条”安装与应用分析[J]. 电视技术, 2022, 46(08):185-187.

[4] 梁秀龙. “电子封条”智能监管技术在煤矿安全监管中的应用探析[J]. 江西煤炭科技, 2022(04):236-238.

[5] 本刊讯. 王祥喜在国家矿山安全监察局调研时强调：切实担起矿山安全监管监察的重大责任坚决遏制矿山重特大事故[J]. 中国安全生产, 2022, 17(08):4.

[6] 刘峰. 煤炭行业科技创新对煤矿安全的影响研究[J]. 煤矿安全, 2020, 61(10):6-9.

[7] Liu Q, Dou F, Meng X. Building risk precontrol management systems for safety in China's underground coal mines[J]. Resources Policy, 2021, 74: 101631.

[8] 吴兵, 邹向炜, 周瑶, 等. “三要素”煤矿安全管理体系研究[J]. 煤炭工程, 2014, 46(10):232-234.

[9] 张长鲁. 煤矿事故隐患大数据处理与知识发现分析方法研究[J]. 中国安全生产科学技术, 2016, 12(09):176-181.

[10] 诸利一, 吕文生, 杨鹏等. 2007-2016年全国煤矿事故统计及发生规律研究[J]. 煤矿安全, 2018, 49(07):237-240.

[11] Min Y, Yexiang F, Weilin T, et al. Study on safety behavior planning theory and control strategies for coal chemical workers[J]. Safety science, 2020, 128: 104726.

[12] Miao C, Duan M, Sun X, et al. Safety management efficiency of China’s coal enterprises and its influencing factors-Based on the DEA-Tobit two-stage model[J]. Process Safety and Environmental Protection, 2020, 140:79-86.

[13] 蒋星星, 李春香. 2013-2017年全国煤矿事故统计分析及对策[J]. 煤炭工程, 2019, 51(01):101-105.

[14] Choudhry R M. Implementation of BBS and the Impact of Site-Level Commitment[J]. Journal of Professional Issues in Engineering Education & Practice, 2012, 138(4):296-304.

[15] 李霞. 基于危险源的煤矿员工不安全行为管理模型研究[J]. 煤矿安全, 2019, 60(10):248-262.

[16] Qiao W, Liu Q, Li X, et al. Using data mining techniques to analyze the influencing factor of unsafe behaviors in Chinese underground coal mines[J]. Resources Policy, 2018, 69:210-216.

[17] Aliyachen A S, Yadav B P, Bhakshi S. Enhancing safety culture in cement industry using behavior-based safety technique[C]// Advances in Fire and Process Safety: Select Proceedings of HSFEA 2016. Springer Singapore, 2018: 103-114.

[18] Zhang J, Fu J, Hao H, et al. Root causes of coal mine accidents: Characteristics of safety culture deficiencies based on accident statistics[J]. Process Safety and Environmental Protection,2020, 136:78-91.

[19] 钱敏, 穆丹丹. 煤矿安全管理评价指标体系[J]. 采矿与安全工程学报, 2008(03):376-378.

[20] 丁振, 张驎. 浅析大数据技术助力煤矿安全管理[J]. 中国煤炭, 2016, 41(10):121-123.

[21] 王海军, 武先利. “互联网+”时代煤矿大数据应用分析[J]. 煤炭科学技术, 2016, 44(02):139-143.

[22] Krizhevsky A, Sutskever I, Hinton G E. Imagenet Classification with Deep Convolutional Neural Networks[J]. Communications of the ACM, 2017, 60(06): 84-90.

[23] Bharati P, Pramanik A. Deep learning techniques—R-CNN to Mask R-CNN: A Survey[J]. Computational Intelligence in Pattern Recognition, 2020, 18(08): 657-668.

[24] Li J, Liang X, Shen S M, et al. Scale-aware Fast R-CNN for Pedestrian Detection[J]. IEEE transactions on Multimedia, 2017, 20(04): 985-996.

[25] 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.

[26] Nguyen D T, Nguyen T N, Kim H, et al. A High-throughput and Power-efficient FPGA Implementation of YOLO CNN for Object Detection[J]. IEEE Transactions on Very Large Scale Integration (VLSI) Systems, 2019, 27(8): 1861-1873.

[27] Kumar A, Srivastava S. Object Detection System Based on Convolution Neural Networks Using Single Shot Multi-box Detector[J]. Procedia Computer Science, 2020, 171: 2610-2617.

[28] Wang L, Yang S, Yang S, et al. Automatic Thyroid Nodule Recognition and Diagnosis in Ultrasound Imaging with The YOLOv2 Neural Network[J]. World journal of surgical oncology, 2019, 17(1): 1-9.

[29] Lin T Y, Goyal P, Girshick R, et al. Focal loss for dense object detection[C]// Proceedings of the IEEE international conference on computer vision. 2017: 2980-2988.

[30] Zhao L, Li S. Object Detection Algorithm Based on Improved YOLOv3[J]. Electronics, 2020, 9(3): 537.

[31] Tan M, Pang R, Le Q V. Efficientdet: Scalable and efficient object detection[C]// Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2020: 10781-10790.

[32] Abdurahman F, Fante K A, Aliy M. Malaria Parasite Detection in Thick Blood Smear Microscopic Images Using Modified YOLOV3 and YOLOV4 Models[J]. BMC Bioinformatics, 2021, 22(01): 1-17.

[33] 沈科,季亮,张袁浩,邹盛.基于改进 YOLOv5s 模型的煤矸目标检测[J].工矿自动化, 2021, 47(11): 107-111+118.

[34] 张翼翔, 林松, 李雪. 基于CenterNet-GhostNet的选煤厂危险区域人员检测[J]. 工矿自动化, 2022, 48(04): 66-71.

[35] Zhang Q Q, Zhu Z J, Ge Z F, et al. Key Components Detection and Identification of Transmission Lines Based on An Improved CornerNet Network[J]. Journal of Computers, 2021, 32(2): 124-136.

[36] Chen Y, Zhang Z, Cao Y, et al. Reppoints v2: Verification Meets Regression for Object Detection[J]. Advances in Neural Information Processing Systems, 2020, 33: 5621-5631.

[37] Song J, Zhao Y, Song W, et al. Fisheye Image Detection of Trees Using Improved YOLOX for Tree Height Estimation[J]. Sensors, 2022, 22(10): 3636.

[38] Zhu X, Xiong Y, Dai J, et al. Deep Feature Flow for Video Recognition[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2017: 2349-2358.

[39] Zhu X, Wang Y, Dai J, et al. Flow-Guided Feature Aggregation for Video Object Detection[C]// Proceedings of the IEEE International Conference on Computer Vision. 2017: 408-417.

[40] Hetang C, Qin H, Liu S, et al. Impression Network for Video Object Detection[J]. arXiv preprint arXiv:1712.05896, 2017.

[41] Lu Y, Lu C, Tang C K. Online Video Object Detection Using Association LSTM[C]// Proceedings of the IEEE International Conference on Computer Vision. 2017: 2344-2352.

[42] Xiao F, Lee Y J. Video Object Detection with An Aligned Spatial-Temporal Memory[C]// Proceedings of the European Conference on Computer Vision (ECCV). 2018: 485-501.

[43] Kang K, Ouyang W, Li H, et al. Object Detection from Video Tubelets with Convolutional Neural Networks[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2016: 817-825.

[44] Kang K, Li H, Yan J, et al. T-CNN: Tubelets with Convolutional Neural Networks for Object Detection from Videos[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2017, 28(10): 2896-2907.

[45] Han W, Khorrami P, Paine T L, et al. Seq-NMS for Video Object Detection[J]. arXiv preprint arXiv:1602.08465, 2016.

[46] Belhassen H, Zhang H, Fresse V, et al. Improving Video Object Detection by Seq-Bbox Matching[C]// VISIGRAPP (5: VISAPP). 2019: 226-233.

[47] Tan M, Pang R, Le Q V. Efficientdet: Scalable and Efficient Object Detection[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2020: 10781-10790.

[48] Chen K, Wang J, Yang S, et al. Optimizing Video Object Detection via A Scale-Time Lattice[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2018: 7814-7823.

[49] LEcun Y. Optimal brain damage [J]. Neural Information Proceeding Systems, 1990, 2(279):598-605.

[50] Han S, Pool J, Tran J, et al. Learning both Weights and Connections for Efficient Neural Networks[J]. Advances in neural information processing systems. 2015, 13(1): 1135-1143.

[51] Hassibi B. Second order derivatives for network pruning: Optimal brain surgeon[J]. Advances in Neural Information Processing Systems, 1993, 74(3): 164-171.

[52] Srinivas S, Babu R V. Data-free parameter pruning for Deep Neural Networks[J]. computer science, 2015, 14(2): 2830-2838.

[53] Jin X, Yuan X, Feng J, et al. Training Skinny Deep Neural Networks with Iterative Hard Thresholding Methods[J]. Advances in neural information processing systems. 2016, 23(2): 987-988.

[54] 舒军, 蒋明威, 杨莉等. DenseNet模型轻量化改进研究[J]. 华中师范大学学报(自然科学版), 2020, 54(02):187-193.

[55] Kim W, Jung W S, Choi H K. Lightweight Driver Monitoring System Based on Multi-task Mobilenets[J]. Sensors, 2019, 19(14): 3200.

[56] Paoletti M E, Haut J M, Pereira N S, et al. GhostNet for Hyperspectral Image Classification[J]. IEEE Transactions on Geoscience and Remote Sensing, 2021, 59(12): 10378-10393.

[57] Hsiao T Y, Chang Y C, Chou H H, et al. Filter-based Deep-compression with Global Average Pooling for Convolutional Networks[J]. Journal of Systems Architecture, 2019, 95: 9-18.

[58] 赵旭剑, 李杭霖. 基于混合机制的深度神经网络压缩算法[J/OL]. 计算机应用:1-8[2023-04-10].

[59] Wang Z, Li F, Shi G, et al. Network Pruning Using Sparse Learning and Genetic Algorithm[J]. Neurocomputing, 2020, 404: 247-256.