Bin picking with 3D vision camera

Random or automated bin picking is typically at the input stage of a manufacturing system. It is widely used in automotive, white goods, mechanical assembly, or machine tending solutions. Instead of traditional static fixtures or pre-filled stacking patterns, a robot empties a bin bulk-filled with parts directly into a machine or into a feeder, conveyor, or sorter for further processing in the plant.

Bin-picking often contains only one type of part with a specific purpose rather than general goods. Bin-picking objects are highly varied in size, color, and material. Developing a bin-picking solution requires the precise placement of the objects for measurement, stamping, or other item manipulations. The ability to properly handle an object can determine if the next step in processing the part or materials will be executed correctly.

The choice of the vision system is critical to fulfilling the bin-picking requirements since the 3D camera impacts the robot’s ability to successfully detect, pick, and place all types of parts. On this page, you will learn more about how to increase the number of pickable parts using our Zivid 3D machine vision cameras.

Bin-picking is still a challenge in robotics. Problematic objects are typically complex-shaped, hard to manipulate, entangled, and/or highly reflective machined or polished metals. The difficulty stems from the inherent uncertainty in physics, perception, and control. Poor detection may lead to collisions, damaged parts, reduced productivity, and poor path planning. To get optimal detection, complete crisp point cloud quality gives the best possible detection for CAD matching. These point clouds must also be accurate and scale with reality to get the best pick pose for the part.

Example of bin with metal parts from laser sensor

Benchmark 1. Point clouds from a laser structured light camera

Metal parts in a bin captured with a high-end stereo sensors produce point clouds

Benchmark 2. Point clouds from a high-end stereo camera

Zivid 3D camera captures high-quality point clouds of metallic cylinders in a bin for pick and place applications

Benchmark 3. Point clouds from Zivid Two

Point cloud example with Zivid Two - Various metallic objects in a bin, typical for industrial bin-picking scenarios. View in 3D HERE

Point cloud example with Zivid Two - Highly shiny and reflective, chrome plated, machined and polished parts. View in 3D HERE.

Occlusion by bin edges; left image captured with laser system (large baseline), right image is with Zivid Two. View in 3D HERE.

After applying Zivid's Stripe Engine. Much more data on scene, far more data on shiny parts.

After applying Zivid's refection filters. Aimed at shiny and reflective objects. Better at handling the small floating objects, like wires and edges with no close background.

Contrast Distortion before vs. after (1)

After applying Contrast Distortion. Useful for correcting cylinders that are aligned with the patterns.

Stationary mounting with the Zivid 3D camera

Stationary is the traditional way we have use cameras in applications until recently. It is tried and true and is able to cover and service a single area.