Xiaojuan Ning

Segmentation of architecture shape information from 3D point cloud

Object Segmentation is an important step in object reconstruction from point cloud data of complex urban scenes and in applications to virtual environment. This paper focuses on strategies to extract objects in 3D urban scenes for further object recognition and object reconstruction. Segmentation strategies are proposed according to object shape features. Rough segmentation is first adopted for objects classification, and further detailed segmentation is implemented for object components. Normal directions are adopted to segment each planar region, so that architectures and the ground can be extracted from other objects. Architectural components are further extracted through an analysis of planar residuals, and the residuals are used to choose seed points for region growing. And meanwhile, the size of segmental regions is used to determine whether or not it includes sparse noisy points. Experimental results on the scene scan data demonstrate that the proposed approach is effective in object segmentation, so that more details and more concise models can be obtained corresponding to real outdoor objects.

Asymmetric double-image encryption method by using iterative phase retrieval algorithm in fractional Fourier transform domain

Abstract. A double-image encryption scheme is proposed based on an asymmetric technique, in which the encryption and decryption processes are different and the encryption keys are not identical to the decryption ones. First, a phase-only function (POF) of each plain image is retrieved by using an iterative process and then encoded into an interim matrix. Two interim matrices are directly modulated into a complex image by using the convolution operation in the fractional Fourier transform (FrFT) domain. Second, the complex image is encrypted into the gray scale ciphertext with stationary white-noise distribution by using the FrFT. In the encryption process, three random phase functions are used as encryption keys to retrieve the POFs of plain images. Simultaneously, two decryption keys are generated in the encryption process, which make the optical implementation of the decryption process convenient and efficient. The proposed encryption scheme has high robustness to various attacks, such as brute-force attack, known plaintext attack, cipher-only attack, and specific attack. Numerical simulations demonstrate the validity and security of the proposed method.

Tree Segmentation from Scanned Scene Data

Tree segmentation is an important step in tree reconstruction from scanned data. A new method is presented for automatic extraction of single objects from a complex scene. The proposed method can be used as a solution for tree segmentation from 3D point cloud data with few restrictions, where many complex objects are included in the scene, like trees, building, cars, and so on. The scene data is initially segmented into several small regions according to the distances between points. A weighted combination is constructed on distances and normal angles in each small region for further segmentation. The minimization of the function will be used to determine whether these regions will be merged or not. This method is tested on several data sets. Effective segmental results demonstrated that this approach could be applied to nondestructive measurements in forestry.

Object shape classification and scene shape representation for three-dimensional laser scanned outdoor data

Abstract. Shape analysis of a three-dimensional (3-D) scene is an important issue and could be widely used for various applications: city planning, robot navigation, virtual tourism, etc. We introduce an approach for understanding the primitive shape of the scene to reveal the semantic scene shape structure and represent the scene using shape elements. The scene objects are labeled and recognized using the geometric and semantic features for each cluster, which is based on the knowledge of scene. Furthermore, the object in scene with a different primitive shape could also be classified and fitted using the Gaussian map of the segmented scene. We demonstrate the presented approach on several complex scenes from laser scanning. According to the experimental result, the proposed method can accurately represent the geometric structure of the 3-D scene.

Shape decomposition and understanding of point cloud objects based on perceptual information

Decomposition and segmentation of the objects represented by point cloud data become increasingly important for purposes like shape analysis and object recognition. In this paper, we propose a perception based approach to segment point cloud into distinct parts, and the decomposition is made possible of spatially close but geodetically distant parts. Curvature is a critical factor for shape representation, reflecting the convex and concave characteristics of an object, which is obtained by cubic surface fitting in our approach. To determine the number of patches, we calculate and select the critical feature points based on perception information to represent each patch. Taking the critical marker sets as a guide, each marker is spread to a meaningful region by curvature-based decomposition, and also further constraints are provided by the variation of normals. Then a skeleton extraction method is proposed and a label-driven skeleton simplification process is implemented. Further, a semantic graph is constructed according to the decomposed model and the skeletal structure. We illustrate the framework and demonstrate our approach on point cloud data to evaluate its function to decompose object shape based on human perceptions. Meanwhile, the result of decomposition is demonstrated with extracted skeletons. Performance of this algorithm is exhibited with experimental results, which proves its robustness to noise.

Automatic architecture model generation based on object hierarchy

Terrestrial laser scanner (TLS) can be used to acquire 3D facade information of modern architectures, represented as point cloud data (PCD). Basic shape elements of an architecture, like windows and doors, should be recovered in reconstruction; and the model should be represented corresponding to the information of architectural design, such as lines and polygons. Most recent approaches could not reconstruct models automatically with designed shape details. Either users interactions are needed [Zheng et al. 2010; Nan et al. 2010]; or the reconstructed model is coarse without information of shape details. Therefore, it is necessary to develop new algorithms to generate geometric models automatically, fitting well the design information of architectural PCD. A novel framework is proposed to generate explicitly an architectural model from scanned points of an existing architecture. An automatic, hierarchical and fast facade reconstruction framework is presented based on a novel combination of facade structures, detailed windows propagation, hierarchical model consolidation and contextual semantic representations. As a result, a high-quality geometric model of an architecture ia generated. Figure 1 shows the procedure of this work, from building detection, to planar region decomposition, to boundary point extraction, and to the consolidated hierarchal model.

Hierarchical model generation for architecture reconstruction using laser-scanned point clouds

Abstract. Architecture reconstruction using terrestrial laser scanner is a prevalent and challenging research topic. We introduce an automatic, hierarchical architecture generation framework to produce full geometry of architecture based on a novel combination of facade structures detection, detailed windows propagation, and hierarchical model consolidation. Our method highlights the generation of geometric models automatically fitting the design information of the architecture from sparse, incomplete, and noisy point clouds. First, the planar regions detected in raw point clouds are interpreted as three-dimensional clusters. Then, the boundary of each region extracted by projecting the points into its corresponding two-dimensional plane is classified to obtain detailed shape structure elements (e.g., windows and doors). Finally, a polyhedron model is generated by calculating the proposed local structure model, consolidated structure model, and detailed window model. Experiments on modeling the scanned real-life buildings demonstrate the advantages of our method, in which the reconstructed models not only correspond to the information of architectural design accurately, but also satisfy the requirements for visualization and analysis.

Automatic Building Extraction from Terrestrial Laser Scanning Data

of point clouds with different local densities, especially in the presence of random noisy points, is still a formidable challenge. In this paper, we present a complete strategy for building extraction from terrestrial laser scanning data. First, a novel segmentation method is proposed to facilitate the task of building extraction. The points are grouped based on the normals and the adjacency relationships. Second, the planar surfaces are recognized from the segmentation results based on the properties of the Gaussian image. Finally, the buildings are extracted from the urban point clouds based on a collection of characteristics of point cloud segments like shape, normal direction and topological relationship. Experimental results demonstrate that the proposed method can be used as a robust way to extract buildings from terrestrial laser scanning data. At the same time, the buildings are decomposed into several patches which lay a good foundation for building reconstruction.

Tree branching reconstruction from unilateral point clouds

Trees are ubiquitous in natural environment and realistic models of tree are also indispensable in computer graphics and virtual reality domains. However, their complexity in geometry and topology make it a great challenge for photo-realistic tree reconstruction. Since tree trunk is the preliminary structure of trees, its modeling is a critical step which plays an important role in tree modeling. Many existing methods focus on the overall resemblance of tree branches but omit the local geometry details. In this paper, we perform unilateral scanning of real-world trees and propose an approach that could reconstruct trees from incomplete point clouds. The core of our method contains four parts: local optimal segmentation of tree branch, skeletal point and lines extraction from unilateral branch, the cross-section construction of tree branch, and final tree branch surface generation. Experimental results demonstrate the effectiveness and robustness of our method which could keep realistic shape of trees.

A new interpolation method in mesh reconstruction from 3D point cloud

Point cloud data is a popular representation of object shape information. Triangulation of scanned point data is an important means for 3d surface reconstruction. In this paper, a new interpolation method is introduced to reconstruct object surface automatically from unorganized point cloud. Our proposed method generates some accurate interpolation points to make the results more reliable based on 3d Delaunay triangular mesh. Each center of the 3d Delaunay triangles is mapped onto the tangent plane of the three vertexes of the triangle, and further accurate interpolation points (also called growing points) are acquired by simplifying the mapping points. Triangulation by a mixture of the growing points and original point cloud data could make the triangular mesh reconstruction of the object more realistic and retaining more detailed information. Experimental results display that our method is effective and robust.