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).

A New 3D Descriptor for Irregularly Shaped Suspended Sediment Aggregates

<p>Within coastal and estuarine environments suspended cohesive sediments that are often closely associated with carbon, nutrients, pathogens and pollutants form aggregates commonly known as ‘flocs’. Understanding the settling dynamics and eventual fate of flocculated sediment is therefore a major issue for the management of aquatic environments. Several factors have been reported to influence the hydrodynamic behaviour of flocs, including size, shape, density and porosity. Recent evidence suggests that of these shape exerts the greatest influence on settling rates. Yet means of characterising shape have been limited to easy to measure quantities such as fractal dimension and circularity measured in 2-dimensions (2D) that fail to capture the highly complex, irregular geometries of sediment flocs. However, recent improvements in sampling methods, 3D imaging capabilities and data processing software enable for the first time the characterisation of flocs based on their 3D morphology.</p><p>This study compares the morphologies of natural and artificial flocs generated under different environmental conditions. By employing a novel apparatus for the capture, immobilisation and handling of delicate floc samples, 3D X-ray micro-computed tomography (X-ray µCT) scans are successfully obtained and used to derive accurate volumetric reconstructions of tens of thousands of individual flocs. Using these datasets we compare different methods for describing shape, and test these for their ability to predict floc settling behaviours.</p>