Video cameras are widely used in underwater robotics to construct 3D coordinates of the workspace. However, the accuracy evaluating that takes into account the properties of the underwater environment, and technical implementation in uncontrolled conditions is still a difficult task. This assessment is especially important for robots that are focused on performing operations with items. In this paper, we propose a novel technique for accuracy analysis and demonstrate its possibilities on real data. It is based on a statistical approach that allows estimating the influence of all sources of perturbations using only experimental data obtained in an underwater environment.
Currently, intensive research is underway to develop remotely controlled and autonomous underwater robots that use technical vision systems. Typical examples of tasks that can be solved using them are: monitoring the environment; detecting objects and obstacles; approaching the robot with an object; performing operations with objects. The article focuses on the problems of constructing images of the working space of an underwater robot designed to perform operations with objects based on information received from a stereo camera. Robust algorithms for constructing 3D images of the robot's workspace based on a perspective camera model and using triangulation and clustering methods are proposed and tested on real data. Image processing algorithms take into account the presence of a waterproof shell.
The article discusses approaches to solving the problems of detecting, recognizing, and localizing an object with given distinctive features in an aquatic environment using a technical stereo vision system, taking into account restrictions. The stereo vision system is being developed as part of the task in which the AUV, for the purpose of conducting a monitoring mission, follows from the starting point of its route along a given trajectory in order to detect and classify an object with known characteristics and determine its coordinates using a technical stereo vision system at a distance up to 5 m from it with appropriate water clarity. The developed program for the system of the technical stereo vision should provide the AUV with the following information: video sequence; a frame with an image of the detected object; previously unknown characteristics of the object if it is possible to detect them (color, size or shape); distance to the object from the technical stereo vision system; and linear coordinates relative to the technical stereo vision system. Testing of the developed software was carried out on the operating module of the stereo vision installed on the AUV in the underbody compartment. The study was carried out in the pool and in open water. The experiments performed have shown the effectiveness of the developed system when used in conjunction with an underwater robot.
Distortion of underwater images can impair both the accuracy and robustness of 3D scene reconstruction algorithms. The problems that arise are related to the lack of robustness of these methods to changes in the underwater environment and features of transmitting and receiving signals under water, including, in particular, uneven illumination of the underwater environment, rapid attenuation, scattering and refraction of light when passing through an inhomogeneous medium of air-water-glass, limiting the frequency spectrum of passing light, which leads to the absorption of low-frequency components to a greater extent than light of higher frequencies. All this seriously complicates the ability to extract information about the scene as a whole and objects of interest located in the underwater environment, limits the use of standard image processing algorithms and requires their significant improvement. This article offers a new approach to analyzing the accuracy of constructing 3D coordinates of the working space of an underwater robot. The approach is based on underwater camera calibration, assessment of camera image centers taking into account the waterproof shell. We use statistical analysis that allows us to evaluate the impact of all sources of disturbances (both hardware and software) based only on experimental data. In particular, it shows how to get the error distribution using the measured values of the calibration sample and obtained by triangulation under underwater conditions. This makes it possible to simultaneously evaluate the systematic error and the distribution characteristics of the random component of the error in restoring 3D coordinates of the workspace. An important feature of the proposed approach is the ability to assess the impact of all sources of disturbances in the aggregate, including the design of a waterproof shell, based only on experimental data obtained in the underwater environment. In addition, the same approach can also provide estimates of the position of camera image centers, allowing for the presence of a waterproof shell to improve the accuracy of image processing algorithms. The proposed approach was tested on real data.
Искажение подводных изображений может ухудшать как точность, так и робастность алгоритмов 3D реконструкции сцены и визуальной одометрии, приводя к уменьшению количества обнаруживаемых сопряженных ключевых точек на парах последовательных изображений. В связи с этим, предобработка изображений и процедуры выбора ключевых точек на них являются важными факторами в этой задаче. В статье исследуется влияние различных алгоритмов предобработки и построения ключевых точек на свойства алгоритмов 3D реконструкции сцены и визуальной одометрии в условиях подводной съемки и неконтролируемого движения камеры. Ключевые слова: подводные видеоизображения, 3D реконструкция сцены, сопряженные и ключевые точки.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.