T.J. Project 02 of 05项目 02 / 05 / Selected Work · 2026精选项目 · 2026 / EN · 中文 Index ↩返回首页 ↩
Back to index返回首页
Project · 02 / 05项目 · 02 / 05

Deep-Ground Smart-Eye
Mining Robot

探地 Smart-Eye
矿用巡检机器人

01 · Overview01 · 概述 Why为什么

First-of-its-kind vehicle-mounted rail inspection system for an underground gold mine. End-to-end ownership of the perception stack — KBA18(F) explosion-proof camera integration, Livox Mid-360 LiDAR for 3D mapping, on-board inference on Jetson Orin NX 16GB, and a full data plus simulation architecture for iteration.

面向井下黄金矿的首创车载轨道巡检系统。端到端负责整套感知栈 —— KBA18(F) 防爆相机集成、Livox Mid-360 LiDAR 三维建图、Jetson Orin NX 16GB 板载推理,以及一套为迭代搭建的数据 + 仿真架构。

The system replaces human patrol on one of the most hazardous legs of underground gold mining — the haulage rail. Every detection and every frame is a step toward removing people from harm, while giving mine operators a continuous, queryable record of tunnel health.

系统替代人工巡检井下黄金矿最危险的一段 —— 运输轨道。每一次检测、每一帧画面,都是在把人从危险里拿出来,同时为矿方留下连续、可查询的隧道健康档案。

Jetson Orin NX/ Livox Mid-360/ KBA18(F)/ Industrial CV工业 CV/ Edge AI边端 AI
03 · Details03 · 细节 How怎么做

Perception感知

  • CameraKBA18(F) explosion-proof
  • LiDARLivox Mid-360
  • DetectorsYOLOv8 · OpenCV
  • MappingSLAM · point cloud registration
  • 相机KBA18(F) 防爆
  • LiDARLivox Mid-360
  • 检测YOLOv8 · OpenCV
  • 建图SLAM · 点云配准

Compute算力

  • EdgeJetson Orin NX 16GB
  • RuntimeONNX · TensorRT
  • SimUnreal + CARLA-style harness
  • DataCaptured + synthetic hybrid
  • 边端Jetson Orin NX 16GB
  • 运行时ONNX · TensorRT
  • 仿真Unreal + 类 CARLA 框架
  • 数据实采 + 合成混合

Delivery交付

  • OwnershipPerception stack end-to-end
  • TeamSolo on AI perception stack
  • OutputInspection reports · alerts
  • StateIn field trial
  • 负责范围整套感知栈端到端
  • 团队AI 部分独立完成
  • 产出巡检报告 · 告警
  • 状态现场试运行