Bin-picking challenges - Occlusion

What is Occlusion?

Occlusion occurs due to the relative placement of projector/camera and camera, creating "blind spots" (shadow areas) in our target scene. This causes a loss of scene details and hence negatively affecting 3D image quality.

Occlusion is directly proportional to the camera's baseline and becomes very prevalent on smaller objects with relatively large features in the Z-axis (towards the camera). A larger baseline camera reduces the overlapping region that both cameras or camera/projector can detect, while a small baseline camera increases it.

Three types of occlusion 

• Inter-occlusion: From a certain viewpoint, one part of an object is occluded by another part. 

Point cloud examples: 3D laser scanner with missing data (left) vs. Zivid 3D camera (right)  

• Object-occlusion: Occlusion on a single object due to large baseline. Results in less data.

Point cloud examples: 3D laser sensor (left) vs. Zivid 3D camera (right)

• Bin-occlusion: Occlusion from a bin wall shadowing objects.

Point cloud examples: Laser 3D camera (left) vs. Zivid 3D camera (right) 

How to solve it with Zivid 3D

Zivid One+ and Zivid 2 3D cameras are designed for reducing challenges in bin-picking and pick and place applications. The small baseline gives higher pixel coverage, which is especially relevant for small parts. With Zivid cameras, robots can better see smaller objects, such as shiny cylinders with smaller diameters than other 3D cameras with a larger baseline.

A large baseline - loss of scene details, creating holes, and missing data

Zivid 3D with a smaller baseline - Fewer occlusion outliers