Feature Pyramid Network Based Efficient Normal Estimation and Filtering for Time-of-Flight Depth Cameras

In this paper, an efficient normal estimation and filtering method for depth images acquired by Time-of-Flight (ToF) cameras is proposed. The method is based on a common feature pyramid networks (FPN) architecture. The normal estimation method is called ToFNest, and the filtering method ToFClean. Both of these low-level 3D point cloud processing methods start from the 2D depth images, projecting the measured data into the 3D space and computing a task-specific loss function. Despite the simplicity, the methods prove to be efficient in terms of robustness and runtime. In order to validate the methods, extensive evaluations on public and custom datasets were performed. Compared with the state-of-the-art methods, the ToFNest and ToFClean algorithms are faster by an order of magnitude without losing precision on public datasets.

Forecasting Nonlinear Systems with LSTM: Analysis and Comparison with EKF

Certain difficulties in path forecasting and filtering problems are based in the initial hypothesis of estimation and filtering techniques. Common hypotheses include that the system can be modeled as linear, Markovian, Gaussian, or all at one time. Although, in many cases, there are strategies to tackle problems with approaches that show very good results, the associated engineering process can become highly complex, requiring a great deal of time or even becoming unapproachable. To have tools to tackle complex problems without starting from a previous hypothesis but to continue to solve classic challenges and sharpen the implementation of estimation and filtering systems is of high scientific interest. This paper addresses the forecast–filter problem from deep learning paradigms with a neural network architecture inspired by natural language processing techniques and data structure. Unlike Kalman, this proposal performs the process of prediction and filtering in the same phase, while Kalman requires two phases. We propose three different study cases of incremental conceptual difficulty. The experimentation is divided into five parts: the standardization effect in raw data, proposal validation, filtering, loss of measurements (forecasting), and, finally, robustness. The results are compared with a Kalman filter, showing that the proposal is comparable in terms of the error within the linear case, with improved performance when facing non-linear systems.

Improving the Performance of Dynamic Ship Positioning Systems: A Review of Filtering and Estimation Techniques

Various operations at sea, such as maintaining a constant ship position and direction, require a complex control system. Under such conditions, the ship needs an efficient positioning technique. Dynamic positioning (DP) systems provide such an application with a combination of the actuators mechanism, analyses of crucial ship variables, and environmental conditions. The natural forces of induced nonlinear waves acting on a ship’s hull interfere with the systems. To generate control signals for actuators accurately, sensor measurements should be filtered and processed. Furthermore, for safe and green routing, the forces and moments acting on the ship’s hull should be taken into account in terms of their prediction. Thus, the design of such systems takes into account the problem of obtaining data about the directional wave spectra (DWS). Sensor systems individually cannot provide high accuracy and reliability, so their measurements need to be combined and complemented. Techniques based on the recursive Kalman filter (KF) are used for this purpose. When some measurements are unavailable, the estimation procedure should predict them and, based on the comparison of theoretical and measured states, reduce the error variance of the analyzed signals. Different approaches for improving estimation algorithms have evolved over the years with the indication of improvement. This paper gives an overview of the state-of-the-art estimation and filtering techniques for providing optimum estimation states in DP systems.

Error Modelling and Optimal Estimation of Laser Scanning Aided Inertial Navigation System in GNSS-Denied Environments

A Laser Scanning aided Inertial Navigation System (LSINS) is able to provide highly accurate position and attitude information by aggregating laser scanning and inertial measurements under the assumption that the rigid transformation between sensors is known. However, a LSINS is inevitably subject to biased estimation and filtering divergence errors due to inconsistent state estimations between the inertial measurement unit and the laser scanner. To bridge this gap, this paper presents a novel integration algorithm for LSINS to reduce the inconsistences between different sensors. In this new integration algorithm, the Radial Basis Function Neural Networks (RBFNN) and Singular Value Decomposition Unscented Kalman Filter (SVDUKF) are used together to avoid inconsistent state estimations. Optimal error estimation in the LSINS integration process is achieved to reduce the biased estimation and filtering divergence errors through the error state and measurement error model built by the proposed method. Experimental tests were conducted to evaluate the navigation performance of the proposed method in Global Navigation Satellite System (GNSS)-denied environments. The navigation results demonstrate that the relationship between the laser scanner coordinates and the inertial sensor coordinates can be established to reduce sensor measurement inconsistencies, and LSINS position accuracy can be improved by 23·6% using the proposed integration method compared with the popular Extended Kalman Filter (EKF) algorithm.