Font Size: a A A

Research Of Binocular Vision Positioning And Navigation Of Unmanned Express Delivery Vehicle

Posted on:2020-08-25Degree:MasterType:Thesis
Country:ChinaCandidate:W B YuanFull Text:PDF
GTID:2428330596995003Subject:Control Science and Engineering
Abstract/Summary:
This paper mainly deals with the research of positioning and navigation using two cameras in a specific situation,which is the navigation of unmanned express vehicle.The difference between the navigation of unmanned express vehicle and that of ordinary automatic driving is that the route of unmanned vehicle is fixed,known and does not change frequently,which greatly reduces the difficulty of navigation.Compared with GPS navigation,magnetic navigation and QR code navigation,the advantages of this navigation method are that it can achieve positioning and navigation without the assistance of external signals and special artificial markers.The main method discussed in this paper is to use binocular camera to scan the surrounding environment of the delivery route,and extract more special and stable feature points from the image of the environment along the route,and record the three-dimensional coordinate positions of these points in the actual environment,so as to construct the route map based on feature points.When the vehicle is working,it pairs the map in real time to get the position and orientation,so as to achieve navigation.The gradient descent method is innovatively used to solve the spatial transformation matrix between reference frames,which makes the obtained transformation matrix more practical than the traditional method.This paper mainly deals with the following contents:1.Camera distortion correction.The quality of photos directly affects the positioning accuracy and navigation.This navigation cannot be achieved if the image is overly distorted or the pixels is not enough.2.Comparison of image feature point matching algorithms.Unmanned vehicle navigation involves not only positioning accuracy,but also positioning speed.It requires real-time performance.If there is a large time lag in positioning,it will not help navigation.Therefore,it is necessary to compare the speed of each feature point extraction and matching method tofind a method of extraction and matching that meets the requirements in accuracy and speed.A series of auxiliary error correction schemes are also used after matching with existing methods.3.Map construction.It can be said that this is the core step in this design,and the quality of map construction directly affects the navigation effect.In this paper,a new method is used to solve the space transformation matrix,which is more practical than the traditional method.4.Positioning and navigation.On the premise of having a three-dimensional map,when the vehicle is actually running,it can adapt the environment to the map in real time,get the position and orientation,as well as adjust the direction in time,so as to achieve navigation.
Keywords/Search Tags:binocular vision, feature point matching, map construction, gradient descent, positioning and navigation
Related items