AbstractView references

Small and start-up companies that need product quality control can usually only afford low-cost systems. The main goal of this investigation was to estimate the influence of high dynamic range images as input for the low-cost photogrammetric structure from motion 3D digitization. Various industrial products made of metal or polymer suffer from poor visual texture. To overcome the lack of visual texture and ensure appropriate 3D reconstruction, stochastic image in the form of the light pattern was projected on the product surface. During stochastic pattern projection, a set of low dynamic range and sets of high dynamic range images were captured and processed. In this investigation digital single lens reflex camera that supports five different tone-mapping operators to create high dynamic range images were used. Also, high precision measurements on a coordinate measuring machine are performed in order to verify real product geometry. The obtained results showed that reconstructed polygonal 3D models generated from high dynamic range images in this case study don’t have a dominant influence on the accuracy when compared to the polygonal 3D model generated from low dynamic range images. In order to estimate 3D models dimensional accuracy, they were compared using computer-aided inspection analysis. The best achieved standard deviation distance was +0.025 mm for 3D model generated based on high dynamic range images compared to the nominal CAD model. © 2019 CPE, University of Maribor. All rights reserved.