Geometric Information Of Objects Research Articles

S$$^{2}$$P$$^{3}$$: Self-Supervised Polarimetric Pose Prediction

AbstractThis paper proposes the first self-supervised 6D object pose prediction from multimodal RGB + polarimetric images. The novel training paradigm comprises (1) a physical model to extract geometric information of polarized light, (2) a teacher–student knowledge distillation scheme and (3) a self-supervised loss formulation through differentiable rendering and an invertible physical constraint. Both networks leverage the physical properties of polarized light to learn robust geometric representations by encoding shape priors and polarization characteristics derived from our physical model. Geometric pseudo-labels from the teacher support the student network without the need for annotated real data. Dense appearance and geometric information of objects are obtained through a differentiable renderer with the predicted pose for self-supervised direct coupling. The student network additionally features our proposed invertible formulation of the physical shape priors that enables end-to-end self-supervised training through physical constraints of derived polarization characteristics compared against polarimetric input images. We specifically focus on photometrically challenging objects with texture-less or reflective surfaces and transparent materials for which the most prominent performance gain is reported.

Structure-Aware Graph Convolution Network for Point Cloud Parsing

Point clouds are becoming a popular medium to describe 3D scenes, benefitting from their accuracy and completeness in expressing the spatial and geometrical information of objects. However, due to the disorder and uneven distribution nature, merely selecting neighbors for point clouds in Euclidean space is inefficient and position-ignoring. To fill this gap, we propose a structure-aware graph convolution network (SA-GCN), which consists of an adaptive dilated KNN module (ADKNN), a learnable graph filter (LGF), and a structure-aware feature transformation module (SFT). Specially, the ADKNN module can dynamically adjust the range of grouping neighbor points, while being universal to improve the performance of arbitrary KNN-based methods. Moreover, with the localized auxiliary information provided by LGF, our SFT module disentangles the spatial details as a sort of coding guidance for better deep feature representations. Extensive experimental results on point cloud classification and segmentation tasks demonstrate the superiority of our proposed network.

On clustering shape data

ABSTRACTAmong the statistical methods to model stochastic behaviours of objects, clustering is a preliminary technique to recognize similar patterns within a group of observations in a data set. Various distances to measure differences among objects could be invoked to cluster data through numerous clustering methods. When variables in hand contain geometrical information of objects, such metrics should be adequately adapted. In fact, statistical methods for these typical data are endowed with a geometrical paradigm in a multivariate sense. In this paper, a procedure for clustering shape data is suggested employing appropriate metrics. Then, the best shape distance candidate as well as a suitable agglomerative method for clustering the simulated shape data are provided by considering cluster validation measures. The results are implemented in a real life application.

3D two-fingered caging for two types of objects: sufficient conditions and planning

Caging is a method to capture an object geometrically by robots. Position-controlled robots can make the object inescapable from the robot formation. Moreover less number of mobile robots or a robot hand with low degrees of freedom can constrain the object with considering concavity of the object. In this paper, we propose two types of caging: ring-type and waist-type, which both can be accomplished by a two-fingered hand. We derive sufficient conditions for caging of the two types and construct RRT-based motion planner for caging by a robotic arm/hand system. In motion planning, we find one of final configurations that satisfy the sufficient conditions and produce a path of robot configuration to the goal. With object recognition using AR picture markers, we can acquire geometrical information of objects and plan robot motion for caging. We show some experimental results of planning and execution of planned caging motion for four objects.

A compact system to extract topography information from scenes viewed by a miniaturized submersible explorer

In images taken underwater, it is generally difficult to correctly extract distances and geometric information of objects. Different techniques, collectively referred to as photogrammetry, exist to measure features in images. One of these is to project a reference pattern onto an object in a scene viewed by a camera, and register the distortion of this pattern, to calculate the shape of, and distance to, that object. This method is implemented here on a miniaturized submersible explorer equipped with, among many other instruments, a camera. Diffractive optical elements (DOEs) have been designed and manufactured using microsystems technology, to, together with a laser diode, camera, and in-house developed software, provide a compact system for projecting reference patterns and analyzing their deformations. The system has been characterized by measuring the distances and angles of objects in a water tank, and attempting to reproduce their shapes. The range of operation of the system, verified to be at least one meter, is limited by the compact mounting in the small submersible and the cameras’ performance. The system was found to work well under turbid conditions as well as in water containing larger particles. Together with a vehicle-mounted camera, the compact and low-power DOE laser projection system enables topographical measurement.

Shape from shading—a method of integration of sonar images and bathymetric data for ocean mapping

Abstract Shape from shading is one of the methods that derive geometric information of objects from analysis of monocular images. Application of this technique to underwater sonar images enables the conversion of imposed reflectance characteristics in sonar images to shape information, namely, slopes, about the seafloor surface. A combination of this shape information and available sparse distributed depth points results in improved dense bathymetric data. The reconstruction of shape models of seafloor surfaces from sonar images is treated as an inverse problem and is solved by the regularization theory. Sparse gridded points are used for boundary constraints. The regularization is implemented as a relaxation procedure with a hierarchical structure of multiresolution grids.