摘要
针对当前煤矿视频AI分析系统开发场景种类繁多且不能自适应调整算法模型的问题,研制了一种基于改进VGG自适应多模型的煤矿视频AI分析系统。通过建立视频AI分析、学习和识别3种基础算法模型,采用自适应模型镜像技术,该系统可设置不同场景,通过学习训练可完成某一种固定场景的模型发布,完成由单一模型向多模型算法的克隆。该系统为矿井多场景视频AI识别提供了一种新的、可靠的技术手段,简化了应用场景算法的开发流程,实现了视频AI快速部署和应用,从而保障煤矿安全生产。
In view of the problem that there are many kinds of development scenarios for the current coal mine video AI analysis system and the algorithm model cannot be adaptively adjusted,a coal mine video AI analysis system based on improved VGG adaptive multi model was developed.By establishing three basic algorithm models for video AI analysis,learning and recognition,and using adaptive model mirroring technology,this system can set different scenes.Through learning and training,the model release of a fixed scene can be completed,and the cloning from a single model to a multi model algorithm was completed.This system provides a new and reliable technical means for the coal mine multi scene video AI recognition,simplifies the development process of application scene algorithms,and realizes the rapid deployment and application of video AI,so as to ensure the safe production of coal mine.
作者
张诺
李向东
赵诺善
Zhang Nuo;Li Xiangdong;Zhao Nuoshan(Chengzhuang Mine,Jinneng Holding Equipment Manufacturing Group,Jincheng 048006,China)
出处
《煤矿机械》
2025年第9期211-217,共7页
Coal Mine Machinery
基金
晋能控股装备制造集团科技项目(LYCZ-JSKF-2022-0004)。
