College of Mechanical Engineering, Xi'an University of Science and Technology
Shaanxi Key Laboratory of Mine Electromechanical Equipment Intelligent Detection and Control
Xi'an Heavy Industry Hancheng Coal Mining Machinery Co., Ltd.

传统的带式输送机异常状态识别采用人工巡检或机械综合保护系统进行检测，人工巡检劳动强度大、效率低、难以准确发现故障等，机械综合保护系统易造成误判，识别效果不佳，已无法满足煤炭行业智能化需求。随着机器视觉、深度学习和工业以太网技术发展，视频AI技术成为煤矿带式输送机异常状态智能识别的研究热点。分析了采用视频AI技术识别煤矿带式输送机输送带跑偏、托辊故障、人员入侵、人员不安全行为、堆煤及异物等异常状态的研究现状，指出目前煤矿带式输送机异常状态视频AI识别技术存在视频图像数据集构建耗时长、异常状态识别精度不高、视频信息传输延时大3个主要问题。针对视频图像数据集构建耗时长问题，提出加强基于半监督、无监督及小样本学习的视频AI识别算法研究、基于生成模型等方式扩充数据集的解决思路；针对异常状态识别精度不高问题，提出加强数据去模糊方法研究、利用生成对抗网络等算法均衡正负样本和改进AI识别算法的解决思路；针对视频信息传输延时大问题，提出构建“云−边−端”协同的带式输送机异常状态视频AI识别系统架构，合理部署高带宽、低延时的网络通信系统的解决思路。从高性能视频AI识别算法，高带宽、低延时视频通信技术，“云−边−端”高效协同的视频AI识别系统和健全视频AI识别技术标准4个方面展望了带式输送机异常状态视频AI识别技术的发展趋势。

Traditional belt conveyor abnormal state recognition uses manual inspection or mechanical comprehensive protection system for detection. The manual inspection is labor-intensive, inefficient, and difficult to accurately detect faults. Mechanical comprehensive protection system is prone to misjudgment and poor recognition effect. The above methods can no longer meet the needs of coal industry intelligence. With the development of machine vision, deep learning, and industrial Ethernet technology, video AI technology has become a research hotspot for intelligent recognition of abnormal states of coal mine belt conveyors. This paper analyzes the current research status of using video AI technology to identify abnormal states of coal mine belt conveyors, such as belt deviation, idler failure, personnel invasion, unsafe behavior of personnel, coal stacking, and foreign objects. It is pointed out that there are three main problems in the current video AI recognition technology for abnormal states of coal mine belt conveyors: long construction time-consumption of video image datasets, low precision of abnormal state recognition, and large time delay in video information transmission. To address the issue of long construction time-consumption of video image datasets, a solution is proposed to strengthen the research on video AI recognition algorithms based on semi supervised, unsupervised, and small sample learning, and to expand the dataset based on generative models. To address the issue of low precision of abnormal state recognition, a solution is proposed to strengthen research on data deblurring methods, and to utilize algorithms such as generative adversarial networks to balance positive and negative samples, and improve AI recognition algorithms. To address the issue of large time delay in video information transmission, a solution is proposed to build a "cloud-edge-end" collaborative video AI recognition system architecture for abnormal states of belt conveyors, and to deploy a high bandwidth and low time delay network communication system. This article looks forward to the development trend of video AI recognition technology for abnormal states of belt conveyors from four aspects: high-performance video AI recognition algorithms, high bandwidth and low time delay video communication technology, "cloud-edge-end" efficient collaborative video AI recognition system, and sound video AI recognition technology standards.

陕西省煤矿带式输送机智能测控技术研究与应用“科学家+工程师”队伍项目（2023KXJ-238）。

毛清华，郭文瑾，翟姣，等. 煤矿带式输送机异常状态视频AI识别技术研究[J]. 工矿自动化，2023，49（9）：36-46.

MAO Qinghua, GUO Wenjin, ZHAI Jiao, et al. Research on video AI recognition technology for abnormal state of coal mine belt conveyors[J]. Journal of Mine Automation，2023，49（9）：36-46.