视觉引导反光圆柱棒料上料-封面图

Random Cylindrical Bar Stock Bin Picking

Published: 2026.09.08Industry: Automotive ManufacturingDeployment Location: Shanghai, China

Project Background

Cylindrical bar stock is a common blank in automotive parts machining, and its loading and unloading directly affects machine tool utilization and machining quality. Bar stock comes in many types and sizes, arrives randomly stacked in material bins, and production lines require greater flexibility and automation. Traditional centering loaders and 2D vision can only handle neatly arranged materials, forcing line stoppages and manual sorting whenever randomly stacked parts arrive.

Case Background and Customer Needs

Recognition and Grasping Challenges Posed by Flexible Production

  • Cylindrical bar stock surfaces are highly reflective, and bars are often tightly pressed together.
  • Bar stock comes in many types and sizes, making recognition and segmentation difficult.
  • Manual loading requires constant bending and lifting, and positioning deviations can easily cause tool crashes and workpiece damage.

Workflow

Built on DexForce Technology's embodied AI architecture, the KINGFISHER-W-3003 binocular stereo intelligent camera performs 3D reconstruction of the bar stock in the material bin and rapidly outputs high-precision point clouds. The DexVerse® embodied AI engine accurately segments tightly packed individual bars, identifies graspable end faces, and outputs 6D poses in real time. The system guides the robot to grasp bars stably from deep bins, completing automated loading at the machining station.

Application Advantages

  • Clear imaging of highly reflective bars: China's only binocular stereo intelligent camera generates high-precision point clouds through deep learning models, solving the imaging challenges of highly reflective metal bars.
  • Precise segmentation of tightly packed bars: advanced vision algorithms accurately recognize cylindrical bars of all sizes packed tightly together, with no dedicated fixtures or manual teaching required.
  • Rapid changeover across dozens of specifications: the generative AI engine drives synthetic data training from CAD models, completing vision configuration in 5 minutes.
  • Stable grasping from deep bins: mature collision-avoidance algorithms plan collision-free grasping paths, keeping grasp success rates and bin-clearing rates stable.

Application Results

The DexForce Technology DexVerse® Spatial and Embodied Intelligence Suite enabled the customer's bar stock loading station to achieve fully automated grasping of randomly stacked parts, replacing the manual sorting step. Bottlenecks in multi-specification mixed production were effectively eliminated, lifting line capacity and machining precision together.