Research On Key Technologies Of 3D Point Clouds Data Processing For Virtual Reality

Virtual reality technology is a cross-discipline developed on the basis of computer graphics, multimedia technology, sensor technology, computer simulation and human-computer interaction technology. It has become a hot topic in computer research field. 3D modeling technology is the foundation of the whole virtual reality system and is the key technology in virtual reality technology. Point cloud data processing technology in 3D modeling technology has become an important research field in virtual reality technology.During the past few decades, considerable work has been done in this research area, such as data pre-processing techniques, analysis of geometric attributes and geometric modeling methods. Data pre-processing techniques mainly involve data acquisition, de-noising smooth, hole filling, data reduction and segmentation. Analysis of geometric attributes mainly includes the calculation of the geometrical characteristics, feature extraction and model retrieval. Geometric modeling methods involve surface reconstruction and model deformation. In this paper, we conducted in-depth research on key technical problem of 3D point clouds based on computer graphics. The main research contents and creative achievements are as follows:(1) The research fields of 3D point clouds are introduced. A detailed exposition and summarization in these areas are provided, such as the consolidation of point clouds, data sampling, feature extraction, data segmentation and model deformation, which are used in the following chapters.(2) A supposedly valid tool in producing a set of de-noised, outlier-free and evenly distributed particles over the original point clouds, Weighted Locally Optimal Projection(WLOP) algorithm has been used in the consolidation of unorganized 3D point clouds by many researchers. However, the algorithm is considered relatively ineffective, due to the large amount of the point clouds data and the iteration calculation. In this paper, a resampling method applied to the point set of 3D model, which significantly improves the computing speed of the WLOP algorithm. In order to measure the impact of error, which will increase with the improvement of calculation efficiency, on the accuracy of the algorithm, we define two quantitative indicators, i.e. the projection error and uniformity of distribution. The performance of our method will be evaluated by using both quantitative and qualitative analyses. Our experimental validation demonstrates that this method greatly improves calculating efficiency, notwithstanding the slightly reduced projection accuracy in comparison to WLOP.(3) A novel approach of 3D human model segmentation is proposed. The heat kernel signature is computed for each feature point which is associated to its semantic group by employing a learning-by-example procedure exploiting manual point labeling. The advantage is that heat diffusion distance is intrinsic and thus deformation-invariant, which makes it available in deformable shape analysis. Finally, the hierarchical segmentation result of the input model can be easily obtained by using detected semantic landmarks and geodesic distance. Experimental tests carried out on 3D human models with different pose variation shows the proposed algorithm is very robust. Furthmore it can avoid the inconsistent segmentation problem, which often occurs in the existing algorithm.(4) The deformation method using the energy-minimum optimization is based on the physically-based deformable models. The curves with the point clouds can be conducted to the curves of a series of polylines. So we convert the curves into the polylines. The polylines are deformed with the arc-length constraint and the multi-points position constraints. The test results show that the proposed method has good performance. Compared to the other method, shape preserving of the curve is better. We use Resampling WLOP algorithm for the consolidation of the human body model. And then the model is divided into parts using the segmentation method which is introduced in chapter 4. The part of the model is transformed to the wireframe model and the multilayer curves are deformed by the method of the energy-minimum optimization. Finally, local deformation of human body model is achived.(5) A method of non-homogeneous mesh resizing using cage-based deformation is proposed. The initial cage is automatically constructed, which is associated with model features or content definitions. Then the minimum elastic energy method was adopted to adjust the scale of each edge of the cage at the different direction. The deformed model which meets the scaling requirements is calculated by using the mean value coordinates of the model. We demonstrate the potential of our algorithm in various applications. The results show that our technique resizes models while suppressing undesirable distortion, creating models that preserve the structure and features of the original ones.Our research work in this paper provides some new methods concerned on 3D point clouds data processing. Finally, we discuss some directions for future work in the last chapter.