| Abstract [eng] |
The goal of this study was to implement the 3D scanner HandyScan 700 on the robot arm IRB 1200 for human limb scanning and evaluate the effect scanning and automation parameters have on recorded meshes. Scanning of a 3D printed replica arm was evaluated using industry standard metrics of surface area covered by the point cloud and its standard deviation, which shows how far away on average are the captured points from the reference. To ensure consistent test results, a static hand replica was used instead of a real one, which had to be scanned in a top and bottom pass and then merged due to insufficient robot reach. Four main experiments were conducted; one assessed the effects of laser exposure time for two different surface finishes (white-glossy, grey-matte), where it was found that higher times are better suited for a glossy surface, lower times – matte surface. Another experiment tested scan repeatability by comparing 7 scans under the same conditions to each other and to a reference mesh, finding that scanner noise was very low at 0.0277 mm for the matte finish. Grey-matte surface finish had 50% lower noise and on average more stable standard deviation than the white-glossy one, demonstrating the negative effects of a reflective surface. Robot speed was also evaluated, with highest accuracy scans being reached while using 100 mm/s speed (preset v100) and 0.2 mm resolution, where standard deviation was as low as 0.039 mm. A recommended compromise between speed and quality is a resolution of 0.6-1 mm and speeds of 500-800 mm/s, which resulted in 0.05-0.1 mm standard deviation and average scan time of 23 seconds (one side). Another test was aimed to quantify how the auto setup compared to human operators, where it was found that there is no difference in accuracy of results, but the auto setup is ~3 times faster. The automated setup was able to match results of similar studies that were using manual scanning methods, as well as of those that aimed to automate scanning. Major limiting factors for scanning of real hands were motion blur and angle change between top and bottom scans, further research is needed regarding limb support during scanning, as well as different robot arm use to support 360° movement around the subject. Rough cost calculations showed that implementing this auto 3D scanning system would cost €65,000 and if used instead of plaster casting for lost limb measurement would result in cost savings of €87 per patient, which would also prevent >1 kg of plaster waste per measurement. |
