As industrial automation continues to advance, automated depalletizing and loading/unloading of material bins remains a major challenge for many manufacturers—especially when the production line involves bins of different specifications, deformed after long use, reflective under varying light conditions, or stacked in complex patterns. Traditional manual handling or pre-programmed robots often struggle to achieve stable and efficient picking.
Recently, a leading manufacturing enterprise adopted an AI + 3D vision–based intelligent handling system, achieving high-precision recognition and automated picking of three different bin types. The system delivers recognition accuracy up to ±0.5 mm and picking accuracy of ±2 mm, significantly improving both efficiency and reliability of the production line.
01 Project Challenges
There are three types of bins (E, M, and L), made of either iron or plastic on the production line needed to process, each with different sizes, weights, and stacking patterns. The system faced several challenges:
Diverse specifications: The three bin types had very different structures—E bins featured detachable, combined designs, while M/L bins were plastic crates. The vision system had to support multiple models simultaneously.
Deformation and gap control: After long-term use, many bins were deformed, requiring the system to dynamically adapt during picking.
Reflection and lighting interference: Plastic bins were prone to surface reflections, and variable on-site lighting demanded strong imaging robustness.
Complex stacking: Incoming bins could be stacked in multiple patterns, including single-row, nested-square, or crisscross arrangements, requiring strong anti-interference and stack recognition capabilities.
02 The Solution
The project integrated a 3D vision system with a gantry robot and a six-axis robot arm to enable fully automated loading and unloading. Key system components included:
High-precision 3D camera
Equipped with a Laser L structured light camera, supporting a 1200–3000 mm working distance. Strong resistance to ambient light ensures complete point cloud output even in bright environments.
Intelligent recognition algorithms
Built-in AI vision algorithms automatically recognize bin type, poses, and picking points. An error-proofing mechanism ensures that only correctly identified bins are picked.
Multi-scenario adaptability
Supports both palletizing and depalletizing operations. Recognition automatically adjusts to different stacking patterns, with each capture recognizing a full bin layer or a single bin.
03 Technical Highlights
Recognition accuracy: ±0.5 mm.
Picking accuracy: ±2 mm.
System recognition rate: 99.9%, eliminating mis-picks.
Strong resistance to reflections and ambient light via laser source and HDR imaging algorithms.
Supports deep bins and narrow-edge grasping, with suction and gripper collaboration for stable, collision-free picking
With the continuous integration of 3D vision technology and AI algorithms, traditional manufacturing is entering a new era of intelligent upgrades. This project not only validates the feasibility of vision systems in handling complex bin handling scenarios, but also provides a replicable technical pathway for flexible, multi-variety, small-batch production. Transfer Technology will continue to advance industrial vision solutions, helping more enterprises achieve unmanned, intelligent transformation.
Return to List
