Performance Evaluation of 3D Descriptors Paired with Learned Keypoint Detectors

Matching surfaces is a challenging 3D Computer Vision problem typically addressed by local features. Although a plethora of 3D feature detectors and descriptors have been proposed in literature, it is quite difficult to identify the most effective detector-descriptor pair in a certain application. Yet, it has been shown in recent works that machine learning algorithms can be used to learn an effective 3D detector for any given 3D descriptor. In this paper, we present a performance evaluation of the detector-descriptor pairs obtained by learning a 3D detector for the most popular 3D descriptors. Purposely, we address experimental settings dealing with object recognition and surface registration. Our results show how pairing a learned detector to a learned descriptors like CGF leads to effective local features when pursuing object recognition (e.g., 0.45 recall at 0.8 precision on the UWA dataset), while there is not a clear performance gap between CGF and effective hand-crafted features like SHOT for surface registration (0.18 average precision for the former versus 0.16 for the latter).

Surface Reconstruction: Roles in the Field of Computer Vision and Computer Graphics

Surface Reconstruction is the most potent aspect of 3D computer vision. It allows recapturing or imitating of the shape of real objects. It also provides sufficient knowledge regarding the mathematical foundation for rendering applications which are widely used for analyzing medical volume data, modeling, 3D interior designing, architectural designing. In our paper, we have mentioned various algorithms and approaches for surface reconstruction and their applications. Moreover, we have tried to emphasize the necessity of surface reconstruction for solving image analysis related problem.

Proposing 3D Thermal Technology for Heritage Building Energy Monitoring

The energy monitoring of heritage buildings has, to date, been governed by methodologies and standards that have been defined in terms of sensors that record scalar magnitudes and that are placed in specific positions in the scene, thus recording only some of the values sampled in that space. In this paper, however, we present an alternative to the aforementioned technologies in the form of new sensors based on 3D computer vision that are able to record dense thermal information in a three-dimensional space. These thermal computer vision-based technologies (3D-TCV) entail a revision and updating of the current building energy monitoring methodologies. This paper provides a detailed definition of the most significant aspects of this new extended methodology and presents a case study showing the potential of 3D-TCV techniques and how they may complement current techniques. The results obtained lead us to believe that 3D computer vision can provide the field of building monitoring with a decisive boost, particularly in the case of heritage buildings.

3D Sensor-Fusion for the Documentation of Rural Heritage Buildings

3D computer vision techniques are now required for the virtual reconstruction of ancient buildings and monuments in urban environments. In this paper, we include a new subfield within the broad field of Urban Heritage that we denominate as Rural Heritage (RH), and which is focused on recovering 3D models of small buildings and facilities of significance in rural environments. We, therefore, present a multi-sensory approach whose objective is to create complete architectural documentation of the dovecotes in an extended region of central Spain. This kind of aviary construction was very common in Spanish rural environments during the 19th century and the first half of the 20th century and is representative of an RH building. Sensory fusion was developed using color cameras, 3D terrestrial laser scanners, and photogrammetric techniques with Unmanned Aerial Vehicles (UAV) and achieves precise indoor and outdoor 3D models. The sensory fusion here also refers to the fact that the information coming from different sensors is integrated into a common documentation framework. A total of 80 dovecotes have been referenced and made available to the public in open access resources. The successful results and applicability of our method lead us to believe that the current documentation and the safeguard technologies in the RH field should evolve towards the use of these 3D computer vision techniques.

Local Feature Extraction Network for Point Cloud Analysis

Geometric feature extraction of 3D point clouds plays an important role in many 3D computer vision applications such as region labeling, 3D reconstruction, object segmentation, and recognition. However, hand-designed features on point clouds lack semantic information, so cannot meet these requirements. In this paper, we propose local feature extraction network (LFE-Net) which focus on extracting local feature for point clouds analysis. Such geometric features learning from a relation of local points can be used in a variety of shape analysis problems such as classification, part segmentation, and point matching. LFE-Net consists of local geometric relation (LGR) module which aims to learn a high-dimensional local feature to express the relation between points and their neighbors. Benefiting from the additional singular values of local points and hierarchical neural networks, the learned local features are robust to permutation and rigid transformation so that they can be transformed into 3D descriptors. Moreover, we embed prior spatial information of the local points into the sub-features for combining features from multiple levels. LFE-Net achieves state-of-the-art performances on standard benchmarks including ModelNet40, ShapeNetPart.