One-shot pattern projection for dense and accurate 3D acquisition in structured light

This thesis focuses on the problem of 3D acquisition using coded structured light (CSL). CSL aims to retrieve the 3D information of an object shape by means of a camera or set of cameras and a coded active device that projects a pattern onto the scene. This pattern imposes the illusion of texture onto an object, increasing the number of correspondences even in presence of textureless surfaces. Afterwards, the camera(s) images the scene, and 3D acquisition is pursued using the same triangulation principle used in stereovision. The vast amount of literature on 3D acquisition using structured light gives an idea of the relevance of the topic. Nowadays, an active research is being done in CSL techniques that are able to work in moving scenarios. This implies the use of few or just one projected pattern, from which 3D information must be extracted. In this thesis, a review of the main CSL approaches is presented. The main features of the proposed CSL algorithms present in the literature are studied. We propose a first approach for one-shot dense acquisition using Wavelet Transform (WT) analysis and color multiplexing of different fringe patterns. This algorithm performs well for smooth surfaces, but fails under presence of discontinuities as the slopes are not detected optimally by the WT algorithm. Therefore, a deep study of the two most used frequency-based techniques is performed, and a new proposal for automatic selection of the window width using Windowed Fourier Transform (WFT) is made. Using this analysis, we implemented a new technique for one-shot dense acquisition. The technique is based on adaptive WFT and DeBruijn coding. The experimental results show that the proposed method obtains accuracy levels comparable to DeBruijn algorithm, but providing absolute dense acquisition. Finally, the last part of the thesis focuses on the problem of registration, as many applications need to register more than one single scan into a big 3D acquisition of a large shape. With this we finish the work of this thesis. The thesis concludes with an analysis of the pros and cons of the technique. The proposed algorithm sets a new trend in CSL as it merges the density of WFT frequency coding with the accuracy of DeBruijn spatial coding, which had been separated approaches until now.