Font Size: a A A

Research On Indoor Positioning And Navigation Of Robot Based On Vision

Posted on:2020-11-11Degree:MasterType:Thesis
Country:ChinaCandidate:J W ZhuFull Text:PDF
GTID:2428330596482458Subject:Computer technology
Abstract/Summary:PDF Full Text Request
The advent of the era of artificial intelligence has set off a wave of development in the field of robotics.With the rapid development of various technologies such as computer technology and computer vision technology,robot technology is constantly moving towards intelligence.Modern robotics is no longer a program-controlled device,but an adaptive "smart" machine system that is inseparable from machine-aware systems and machine learning.In terms of perception systems,analogous to humans can rely on developed visual information to quickly construct scene models in the scene to obtain rich and effective visual information,so visual perception is undoubtedly a development trend;In machine learning,there are three learning methods: supervised learning,semi-supervised learning and unsupervised learning.The neural network is a typical representative of supervised learning and has strong learning ability.If the visual technology,machine learning technology and robot technology are effectively integrated,through the effective cooperation of various sensor data,the strengths and weaknesses will surely promote the qualitative leap of related technologies in the robot field.Autonomous navigation is one of the core of the robot intelligent system,and it is the key influencing factor for the robot to realize the environment perception and mobility.In order to make the robot have good positioning and navigation effects in the indoor scene,and at the same time have certain universality,this paper proposes to locate the robot based on the neural network from the perspective of computer vision,and combine the robot system to complete the robot positioning navigation task.And use it in the actual scene to verify the practical value of the system method.Through the visual sensor,the robot can obtain more abundant scene environment information,and solve the problem that the laser sensor can only obtain less twodimensional environment information,and the use cost is reduced.The positioning part adopts the positioning method based on the full convolutional neural network to densely return the scene coordinates,and defines the correspondence between the image and the scene space,that is,the scene positioning,so that the robot can realize itself by acquiring a single RGB image at the current perspective.The positioning function,the positioning result plays a key role in the movement of the robot and planning the path to the destination coordinate point,combined with the robot mobile platform system,and finally completes the robot navigation task.
Keywords/Search Tags:visual positioning navigation, neural network, robot
PDF Full Text Request
Related items