Development of a 3D Image-Based Loading Assistance System (3차원 영상 기반의 적재 보조 시스템 구축 용역)

RGB-D loading assistance with goods, worker and free-space perception.

Visual Intelligence & Logistics

2018-07-23 – 2018-12-21 · Completed

Korea Railroad Research Institute (KRRI)

Results

  • Estimated loading positions and orientations from product geometry and free space.
  • Integrated worker recognition and localization with ROS-based logistics-robot control.

Tests and measurements reported in 2018.

Task Result Conditions
Goods recognition 97.5% accuracy RGB-D ensemble model
Worker recognition 98.3% success Approximately five-minute video · 4 failures in 238 frames
Worker localization Approximately 3–4 cm error Reported position-measurement examples
Processing rate RGB 30 fps · Depth 90 fps ROS integration environment