[{"data":1,"prerenderedAt":215},["ShallowReactive",2],{"content-query-jzi2BlNQVo":3},{"_path":4,"_dir":5,"_draft":6,"_partial":6,"_locale":7,"title":8,"description":9,"date":10,"cover":11,"type":12,"category":13,"body":14,"_type":209,"_id":210,"_source":211,"_file":212,"_stem":213,"_extension":214},"/technology-blogs/zh/3541","zh",false,"","基于昇思打造智慧矿山发明专利：旷“识”之瞳——小样本条件下矿石粒度检测系统","本产品聚焦矿石粒度检测中的棘手问题，分析矿石图像的特征，利用小样本来解决矿石粒度检测中面临的瓶颈问题，在多种矿石筛选产线上实现多平台协同的矿石图像粒度高精度实时检测识别，以提升检测效率与智能化水平。","2024-12-17","https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2024/12/20/6e1ac15998834ff99f55e21a46fa7a3f.png","technology-blogs","实践",{"type":15,"children":16,"toc":206},"root",[17,25,35,40,48,53,58,62,70,78,83,90,98,103,110,118,123,130,135,140,147,152,157,165,170,184,192,199],{"type":18,"tag":19,"props":20,"children":22},"element","h1",{"id":21},"基于昇思打造智慧矿山发明专利旷识之瞳小样本条件下矿石粒度检测系统",[23],{"type":24,"value":8},"text",{"type":18,"tag":26,"props":27,"children":28},"p",{},[29],{"type":18,"tag":30,"props":31,"children":32},"strong",{},[33],{"type":24,"value":34},"团队介绍",{"type":18,"tag":26,"props":36,"children":37},{},[38],{"type":24,"value":39},"“矿”识之瞳团队深耕绿色矿冶行业领域，拥有强大的科研支撑平台，依托矿冶过程智能优化制造全国重点实验室、武汉人工智能计算中心等平台，与上海云统科技、矿冶科技有限公司等公司展开校企合作，开发小样本条件下矿石粒度检测系统。团队指导老师张杨为湖北工业大学副教授，昇思MindSpore布道师，华为布道师，武汉人工智能科研创新联合体发起人，从事机器视觉，图像处理，深度学习与多媒体信息处理等方面的研究。主要团队成员是5位研究生，分别是李明晶，刘瑾彧、梁启航、刘顶杰、李明应。团队成员分工协作，拥有丰富的视觉系统开发经验。",{"type":18,"tag":26,"props":41,"children":42},{},[43],{"type":18,"tag":30,"props":44,"children":45},{},[46],{"type":24,"value":47},"项目背景",{"type":18,"tag":26,"props":49,"children":50},{},[51],{"type":24,"value":52},"采矿业是现代经济发展的基础，是工业生产的能源与原材料来源。但由于采矿工作环境恶劣、劳动力短缺、企业追求降本等多重因素现状挑战，急需找寻新的破局路径。智慧矿山是未来矿山建设的关键途径，有望妥善解决当下采矿行业面临的各种难题。矿石的粒度信息能够反映破碎过程中各级破碎机的工作状况，是破碎效果的重要数据指标。",{"type":18,"tag":26,"props":54,"children":55},{},[56],{"type":24,"value":57},"高效而精确的矿石粒度分布的检测是对矿石破碎优化的基础，直接影响到整个选矿过程的生产效率。但是，在当前传统的选矿生产中，对于矿石粒度的检测主要通过人工筛分和沉降法检测等方法来实现，这些检测方法存在检测时间长，反馈矿石粒度的分布信息滞后，人为干扰影响大，易造成检测的粒度的精确度比较低，效率不高。随着深度学习的发展，各种目标检测系统也被应用于工业检测领域，然而这些检测网络需要大量图片样本进行训练，要得到高质量的图片样本需要人工标注真值框，这十分消耗人力物力。",{"type":18,"tag":26,"props":59,"children":60},{},[61],{"type":24,"value":9},{"type":18,"tag":26,"props":63,"children":64},{},[65],{"type":18,"tag":66,"props":67,"children":69},"img",{"alt":7,"src":68},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2024/12/20/dd195bf3b9094398a705c29af917984c.png",[],{"type":18,"tag":26,"props":71,"children":72},{},[73],{"type":18,"tag":30,"props":74,"children":75},{},[76],{"type":24,"value":77},"项目简介",{"type":18,"tag":26,"props":79,"children":80},{},[81],{"type":24,"value":82},"在我国“双碳”战略目标的指引下，促进大数据与人工智能等技术与矿冶产业的深入融合，建设智慧矿山以提高能效并提升厂矿安全水平具有重大意义。针对矿石筛选过程中矿石图像标注难、堆叠粘粘严重导致模型计算复杂度高、选矿时硬件资源受限等多种矿石检测应用约束，设计了基于小样本矿石粒度实时在线检测系统。构建小样本学习算法，利用图像边缘与设备运行参数等先验知识，设计特征提取网络结合移位操作，仅需少量训练样本即可实现矿石粒度信息高精度实时检测。在端-云-边协同的发展趋势下，建立知识联动机制,设计有效的压缩模型算法，实现低时延推理,实现多平台协同的矿石图像粒度高精度实时检测识别。提升检测效率与智能化水平，提高我国矿产企业的竞争力，助力国家早日实现碳达峰、碳中和目标。",{"type":18,"tag":26,"props":84,"children":85},{},[86],{"type":18,"tag":66,"props":87,"children":89},{"alt":7,"src":88},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2024/12/20/2c7da675a2cb4643bb2be08442e2039e.png",[],{"type":18,"tag":26,"props":91,"children":92},{},[93],{"type":18,"tag":30,"props":94,"children":95},{},[96],{"type":24,"value":97},"项目实现思路",{"type":18,"tag":26,"props":99,"children":100},{},[101],{"type":24,"value":102},"在基于深度学习的矿石图像小样本识别网络构建与部署的过程中，团队先对行业进行分析：市场需求方面，我国大型煤炭企业的自动化率连年上升，同时在智能化采掘各方面也在稳步上升，但当前行业存在精度低，效率低，隐患多等问题，而市场所需的是高准确率，高实时性，高可靠性的高性能系统。市场需求与行业现状严重不匹配，存在市场空白矿石检测应用约束多。针对矿石粒度检测方面主要存在难点：1标注难、人工标注成本高；2矿石堆叠粘连严重，图像对比度低，边界模糊；3井下环境硬件资源受限，这些难点进行设计算法并调优。随后分析部署环境，选择合适的硬件及预测库，最后接入流程，并提供可视化工具。",{"type":18,"tag":26,"props":104,"children":105},{},[106],{"type":18,"tag":66,"props":107,"children":109},{"alt":7,"src":108},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2024/12/20/9186c359345040c79ab6f4f0314ff4be.png",[],{"type":18,"tag":26,"props":111,"children":112},{},[113],{"type":18,"tag":30,"props":114,"children":115},{},[116],{"type":24,"value":117},"项目技术与效果",{"type":18,"tag":26,"props":119,"children":120},{},[121],{"type":24,"value":122},"基于昇思MindSpore AI框架的全场景开发优势，团队构建了一套从云端训练到边缘部署的完整解决方案。MindSpore框架支持NPU、CPU、GPU等多种计算平台，采用统一的API设计，使得模型能够在不同场景下灵活切换和部署。针对矿石数据标注难，标注成本高，现有高质量的矿石标记样本少，团队设计了基于小样本学习的矿石粒度检测网络结构，借助迁移学习等技巧，仅使用十张样本训练网络，就能得到与目前传统目标检测算法相媲美的精度和速度，团队使用自主设计的轻量级主干使模型大小减少93.6%。在人工智能领域发表了1篇高水平论文。得益于MindSpore与昇腾AI处理器，模型在边缘设备上也能实现高效推理。团队目前正在探索利用基于MindSpore的Diffusion等多模态大模型来进行数据增强和扩充，以进一步提升系统性能，相关工作还在进行中。",{"type":18,"tag":26,"props":124,"children":125},{},[126],{"type":18,"tag":66,"props":127,"children":129},{"alt":7,"src":128},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2024/12/20/7028ebf67261433599163ca4845b6764.png",[],{"type":18,"tag":26,"props":131,"children":132},{},[133],{"type":24,"value":134},"轻量级主干（OreFSDet）",{"type":18,"tag":26,"props":136,"children":137},{},[138],{"type":24,"value":139},"针对矿石堆叠粘连严重，图像对比度低，边界模糊的问题，团队改进MLP分割网络，充分发挥MindSpore自动并行和自动微分特性，实现了局部和全局信息的高效平衡。同时利用框架提供的算子开发套件，创新性地引入移位操作来提取多轴向的局部特征，提升了实例边缘点的匹配精度。基于MindSpore的新一代图神经网络开发工具，团队优化了特征提取模块，在保持较小模型体积的同时显著提升了边界检测精度。",{"type":18,"tag":26,"props":141,"children":142},{},[143],{"type":18,"tag":66,"props":144,"children":146},{"alt":7,"src":145},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2024/12/20/08750bba7da6492c836ba6c5cdf7c7a4.png",[],{"type":18,"tag":26,"props":148,"children":149},{},[150],{"type":24,"value":151},"改进MLP分割网络",{"type":18,"tag":26,"props":153,"children":154},{},[155],{"type":24,"value":156},"针对井下场景硬件资源受限，团队提供云平台给予用户算力支持。团队的云平台系统通用型强，可移植性强，野外环境下也能拥有满足行业标准的检测速度及精度。对于采集的图像平均每张25毫秒（40fps），模型大小仅26MB。系统目前已得到昇腾技术认证，后续将与多模块化机械拣选等技术有机高效结合，部署在多种矿石筛选产线上，实现端边云协同的矿石图像粒度高精度实时检测。",{"type":18,"tag":26,"props":158,"children":159},{},[160],{"type":18,"tag":30,"props":161,"children":162},{},[163],{"type":24,"value":164},"项目总结",{"type":18,"tag":26,"props":166,"children":167},{},[168],{"type":24,"value":169},"小样本条件下矿石粒度检测系统在保证识别精度的前提下实现品质控制自动化，为矿冶全生命周期的健康管理提供了数据支撑。通过快速准确提取并识别矿石粒度，增加了矿产资源的利用率，减少了冶炼的废料，提高了废弃资源的利用率，助力打造智慧、安全、环保的新型矿山。同时还能引领教育，做到产教结合，在创新学术人才和高端产业人才培养方面齐头并进。",{"type":18,"tag":26,"props":171,"children":172},{},[173,175],{"type":24,"value":174},"项目链接：",{"type":18,"tag":176,"props":177,"children":181},"a",{"href":178,"rel":179},"https://git.openi.org.cn/MindSpore%5C_OpenI/kuangshizhitong",[180],"nofollow",[182],{"type":24,"value":183},"https://git.openi.org.cn/MindSpore\\_OpenI/kuangshizhitong",{"type":18,"tag":26,"props":185,"children":186},{},[187],{"type":18,"tag":30,"props":188,"children":189},{},[190],{"type":24,"value":191},"项目证书与专利",{"type":18,"tag":26,"props":193,"children":194},{},[195],{"type":18,"tag":66,"props":196,"children":198},{"alt":7,"src":197},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2024/12/20/12686c4311d849bc808c0eed41487d9e.png",[],{"type":18,"tag":26,"props":200,"children":201},{},[202],{"type":18,"tag":66,"props":203,"children":205},{"alt":7,"src":204},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2024/12/20/227ceac33d05472697eda379f93ac8d7.png",[],{"title":7,"searchDepth":207,"depth":207,"links":208},4,[],"markdown","content:technology-blogs:zh:3541.md","content","technology-blogs/zh/3541.md","technology-blogs/zh/3541","md",1776506130710]