Hao Deng

A mathematical morphology-based multi-level filter of LiDAR data for generating DTMs

Light detecting and ranging (LiDAR) technology has become an effective way to generate high-resolution digital terrain models (DTMs). To generate DTMs, point measurements from non-ground features, such as buildings, vegetation and vehicles, have to be identified and removed while preserving the terrain points. This paper proposes an efficient mathematical morphology-based multi-level filter to generate DTMs from airborne LiDAR data. Preliminary non-ground points are first identified with the characteristics of the multi-echo airborne LiDAR data. The localized mathematical morphology opening operations are then immediately applied to the remaining points. By gradually increasing the window size of the filter and using a dynamic critical gradient threshold, the non-ground points are removed, while the ground points are preserved. Eight samples were chosen from eight sites provided with the ISPRS Commission III, Working Group 3, to evaluate the accuracy of our algorithm. Both the qualitative and quantitative experiment analyses show that our morphology-based multi-level filter method achieves promising results, not only in flat urban areas but also in rural areas, especially in preserving complex terrain details, while non-ground spatial objects are removed.

Interactive Urban Context-Aware Visualization via Multiple Disocclusion Operators

In 3D urban environments, features of interest (FOIs) are often occluded by clusters of buildings, which prevent a clear overview of important spatial features. State-of-the-art disocclusion methods for urban environments fall short of preserving cityscape appearance or require time-consuming computation. These methods use only one or two operators for disocclusion and might not strike a good balance between disocclusion and distortion control. We present a novel, automatic method enabling interactive context-aware visualization of urban features of interest, which combines four effective disocclusion operators including viewpoint elevation, road shifting, building scaling, and building displacement to disocclude the features of interest. Our method provides an optimum compromise among the disocclusion operators via an efficient constrained optimization and the post-polishing phrases, which minimizes the distortions while enforcing the visibility of the FOIs. The 3D views generated at interactive frame rates ensure a resemblance in the cityscape appearance to its original ones and provide a good overview of the FOIs. The experiments with real data demonstrate that our method can greatly facilitate tasks such as navigation, wayfinding, and information overlay.

Efficient occlusion-free visualization for navigation in mountainous areas

In three-dimensional (3D) navigation, if mountainous terrain is displayed based on ordinary perspective projection, viewers often find that the features of interest are occluded, which prevents an overview of the features. This paper presents an approach for the automatic generation of consecutive non-perspective views of mountainous terrain. The proposed method can generate views without occlusions of important features, and allows viewers to navigate the landscape. The ray-tracing technique is employed to detect occlusions. The local elevations that occlude important features are transformed, while the resemblance and realism of the 3D landscape are maintained by solving global optimization problems. The approach maximizes the visibility of the features of interest on the deformed terrain. It also maintains a good balance between the elimination of occlusion and the preservation of resemblance. The occlusion-free visualization framework satisfies the demand for navigation and tour guidance in mountainous areas at interactive frame rates.

Three-dimensional morphological analysis method for geologic bodies and its parallel implementation

It has been found that the spatial locations and distributions of orebodies, especially for certain hydrothermal mineral deposits, are closely related to the shape of intrusive geologic bodies. For complex and large-scale geologic bodies, however, it is challenging to achieve rigorous and quantitative morphological analysis by standard geological surface reconstruction and trend-surface analysis methods. This paper presents a novel, quantitative morphological analysis method for general geologic bodies of closed 2-manifold surface based on mathematical morphology. Through the processes of morphological filtering, set operations and three-dimensional Euclidean distance transform (3D-EDT), the global trend shape, local convex and concave zones as well as degree of surface undulation of a geologic body are extracted respectively. All of the three analysis phases are speeded up via parallel algorithms implemented by using the message passing interface (MPI) standard. The proposed method is tested with a case study of the Xinwuli intrusion with complex shape in Fenghuangshan deposit of the Tongling district, China. The results demonstrate that the method is an effective and efficient way to achieve quantitative morphological analysis, thereby decreasing the time necessary to find the association between morphological parameters of geologic bodies and mineralization. A 3D morphological analysis method for general closed 2-manifold geologic bodies.Parallel algorithms that allows the morphological analysis to be more scalable.A case study discovering the association between geologic shape and mineralization.

Interactive panoramic map-like views for 3D mountain navigation

In 3D terrain navigation applications, the views based on general perspective projection often find features of interest (FOIs) being occluded. As an alternative, panorama-like views preserve the similarity between 3D scenes before and after the deformations while ensuring the visibility of interested features. In this paper, an automatic method for generating panoramic map-like views is proposed in mountainous areas. The created panorama-like views by moving up the view position as well as the terrain deformation can successfully avoid occlusions of the FOIs. The final views also ensure the resemblance in appearance for the FOIs and landscapes, and thus satisfy the demand for interactive occlusion-free navigation in 3D complex terrain environments.

Learning a Discriminative Distance Metric With Label Consistency for Scene Classification

To achieve high scene classification performance of high spatial resolution remote sensing images (HSR-RSIs), it is important to learn a discriminative space in which the distance metric can precisely measure both similarity and dissimilarity of features and labels between images. While the traditional metric learning methods focus on preserving interclass separability, label consistency (LC) is less involved, and this might degrade scene images classification accuracy. Aiming at considering intraclass compactness in HSR-RSIs, we propose a discriminative distance metric learning method with LC (DDML-LC). The DDML-LC starts from the dense scale invariant feature transformation features extracted from HSR-RSIs, and then uses spatial pyramid maximum pooling with sparse coding to encode the features. In the learning process, the intraclass compactness and interclass separability are enforced while the global and local LC after the feature transformation is constrained, leading to a joint optimization of feature manifold, distance metric, and label distribution. The learned metric space can scale to discriminate out-of-sample HSR-RSIs that do not appear in the metric learning process. Experimental results on three data sets demonstrate the superior performance of the DDML-LC over state-of-the-art techniques in HSR-RSI classification.

3D Quantitative Predictivity of Concealed Ore Bodies in Fenghuangshan Copper Deposit, Tongling District, China

For some old mines, their available mineral reservoirs are rapidly decreasing and even exhausting after decades of mining. However, there may be great mineral resource potentials in their depth and margin. Thus, techniques are required to prospect new mineral recourses in deep and marginal parts of the crisis mines. With the maturing of geographical information system (GIS) techniques, the GIS-based mineral prediction and appraisal have becomes the mainstreams of the mineral resource appraisal regional resource prospectivity (Zhou et al., 2007; Carranza et al., 2008; Cassard et al., 2008). However, this line of researches only rely on 2D and 2.5D GIS, which are limited to the meet the requirement to prospect the mineral resources in 3D space, say, the deep and margin parts. On the other hand, since 1990s the techniques of 3D geological modeling (Houlding, 1994) have become more practical, which lays a solid foundation for generalizing the quantitative prediction of mineral resources to the 3D space. This paper presents a novel method that adopts 3D geological modeling, 3D spatial analysis and visualization to achieve 3D quantitative mineral prospectivity. A case study from Fenghuangshan copper deposit, China was conducted, aiming at prospecting of mineral resources in the deep parts of the old and crisis mines.

Automatic Generation of 2.5D Terrain Models without Occluding Routes of Interest

This article discusses how, when a car drives in mountainous regions, the views based on conventional perspective protections suffer from features of interest being occluded. The article proposes a method for generating disocclusion views in mountainous regions. The terrain is segmented in order to build a potential set of occluders and then an optimized viewpoint is determined and elevations are arranged. In order to obtain a smooth deformed terrain, a smooth displacement function is introduced to deform the level-of-detail terrain models. Compared with previous methods, the benefits lie in automatically generating disocclusion views with the temporal coherence, while keeping the details of the deformed terrain the same as the original terrain. Moreover, the shapes of the features of interest on the driving route without occlusion and the spatial configuration of geographical landmarks in the neighborhood can be easily recognized.