| Surveying, modeling and 3D visualization of the surface features of a mine are essential for digital mine construction. Compared to traditional surveying and mapping technologies, three-dimensional laser measurement is a newly developed technology to survey faster with wider range and higher accuracy and less time-consuming. The spatial morphology of mine surface features is complex, which includes natural and man-made features. It is a quite professional work to use three-dimensional laser measurement to acquire point cloud data, and it is important to choose appropriate methods for point cloud data de-noising, feature point extraction,3D modeling and visualization of natural features and man-made objects of mine surface.3D reconstruction needs to extract feature points of man-made and natural features. Because spatial morphology of man-made object is regular, its characteristic points are easier to extract from point cloud data. However, due to irregular forms of natural features, the extraction of characteristic points of natural features is more complex, and there still exist many problems to solve in characteristic point extraction process. Focused on faster algorithm and efficient approaches to extract characteristic points of natural feature’s irregular point cloud data, experiments have been done in this study.Natural feature point cloud data obtained from a three-dimensional laser measurements is the data source in this study, and software, such as HD3LSSCENE and HD-Modeling are used to extract the feature points of point cloud data of man-made and natural objects, and surface mesh model construction approach have been given. First, in the process to extract feature points of natural objects, different extraction method have been compared, and experimental results has been analyzed by selecting different morphological characteristics of natural landscapes to figure out the relationship between extraction parameter and natural landscape features. Scale parameter setting approach has been adopted in experiment. According to different natural landscape features, the parameter has been adjusted and compared to select the best scale parameter. Secondly, according to the characteristics of the natural landscaped features, scale parameter in mesh surface modeling process has been studied by adjusting grid density to create mesh surface, and the best surface mesh model has been selected to build natural landscape features model. Finally, the point cloud data model has been reconstructed.Conclusions obtained from experiments of this study can be summarized as follows:1. In simplification and de-noising process of point clouds, random sampling and bilateral filtering approach is focused on for point cloud de-noising. For different types of point cloud data, both of the two approaches are of advantage. By comparing the results of the two approaches, it can be conclude that the bilateral filtering has remarkable advantage to deal with irregular point cloud data.2. In the case of natural features characteristic points description and extraction, for the reason that there are variety of algorithms of description and extraction, K-tree near algorithm and Descriptive method are proposed to describe point cloud feature points. With the aid of related PCL characterization and extraction module, PCL extraction of feature points has been illustrated in the study to achieve a faster extraction of natural landscape feature point, which can reduce irrelevant data point in subsequent train gulation and surface reconstruction.3. Experiment and analysis have been conducted for cloud point data of natural landscape features, and model reconstruction has been performed by selecting best model reconstruction extracting feature points. It has been shown that in the case of K- tree Neighbor feature point extraction, selection of appropriate parameters is important for model reconstruction. It has been also shown that point cloud after the parameter setting has a significant effect on surface reconstruction.
|
PDF Full Text Request