One of the challenges of 3D point cloud processing is the localization of 3D CAD models. With our PartFinder function in the EnsensoSDK we offer a solution to this task, which is optimized to process Ensenso camera point clouds and is integrated into EnsensoSDK GUI tools.

Simple Parametrization

After loading your CAD model PartFinder can be parametrized and used directly from within NxView. The point clouds of all connected Ensenso cameras can be used for CAD model search, and the quality and accuracy of the results can be examined in the 3D view. Recognition of different part positions or effects of specific search settings can be verified interactively. A successful and tested configuration can then be easily transferred into your application, which can execute the PartFinder function via the EnsensoSDK API.

Robust Detection

The combination of depth, surface normals and texture data enables PartFinder to robustly solve even challenging localization problems. The internal multi-hypothesis processing evaluates multiple probable poses for each part in order to correctly distinguish similar poses and yields high rates of correct detections. By preprocessing the CAD models on the GPU and optimizations for rotation symmetries PartFinder achieves low setup and search times, even for complex geometries and high resolution point clouds.