3D Scanner System for Stair Reconstruction
The industry partner Weitzer Parkett produces parquet and stair claddings. Each stair cladding can be made from different types of parquet. They have professional customers – mainly floorers who are also installing wooden stairs at private and public premises. With their application “3D Measurement for complex constructions” as specified in the Vision+ application document, it is intended to develop a 3D measurement system for stairs in Vision+. The 3D measurement system is able to capture raw stairs – before application of the wooden cover – in order to model the shapes of the covering elements (boards / planks) for automatic production.
The scene to be measured is captured by a portable active depth sensor. Hereby, a sensor system was built consisting of COTS industrial computer vision components (three industrial cameras, synchronization box, infrared band-pass filter, lenses, mechanical set-up, power supply) combined with the projection unit of a commercial 3D sensor (infrared speckle pattern) and control software that is able to capture synchronized videos with about 15 frames-per-second (fps) in full resolution of 1380 * 1024 pixels.
A workflow for dense distance measurement integrates the raw data to a coherent 3D model of the scene. The achievable accuracy lies in the mm-range and is mainly dependent on the resolution of the cameras, respectively the applied sensors.
The workflow includes:
- Lens distortion correction, stereo matching, artefacts removal and distance image generation.
- Dense Stereo Matching: SGM (semi global matching) was used, together with the necessary epipolar rectification.
- Depth map fusion that is optimized for the noise and outlier characteristics of the new depth sensor.
- Assessment: The 3D model is segmented into geometric primitives (tread and riser for stairs) and the data is prepared for further processing in CAD programs.
The 3D-scanning system is able to automatically combine individual measurements and can thus measure a larger volume by generating a dense net of measurement points.
The system offers high precision and is ideal for any application that requires an accurate 3D capturing of the environment.