Positioning method based on scene matching and machine learning in urban canyon environment

Abstract

The invention provides a positioning method based on scene matching and machine learning in an urban canyon environment. The positioning method comprises the following steps of: step 1, constructing a building contour line database; step 2, calibrating a random forest training data set; step 3, training a multi-feature random forest decision tree for receiving type judgment; step 4, acquiring contour information of a building where a carrier is actually located; step 5, carrying out contour feature extraction and feature matching rough positioning on the building at the actual position of the carrier; step 6, determining an optimal candidate region; and step 7, acquiring accurate position information through interpolation in the candidate region. The positioning method is no longer limited by the number of available satellites during positioning calculation, and the problem that the urban canyon lacks satellites severely is solved.

Description

technical field [0001] The invention relates to a positioning method based on scene matching and machine learning in an urban canyon environment. Background technique [0002] The lack of stars in the urban canyon environment is very serious. In highly urbanized areas, even under the combination of multiple GNSS constellations, there may still be insufficient available stars, and the satellite signals are seriously blocked by obstacles. Special building wall materials may also This leads to the reflection and diffraction of satellite signal transmission, resulting in the degradation of satellite positioning accuracy, which cannot meet the actual application needs of urban users. [0003] The "Internet +" new format derived from the development of the Internet has put forward higher requirements for location-based services (Location Based Services, LBS). People's daily commuting, online ordering, logistics tracking and even social entertainment and many other life details ne...

AI Technical Summary

Problems solved by technology

Although the method based on multi-GNSS constellation combination alleviates the serious problem of star shortage in a single system in an urban environment to a certain extent, in highly urbanized areas with many obstacles, even multi-constellation combination cannot obtain enough visible Satellites perform positioning calculations, and even if enough satellites are obtained, there is no guarantee that satellite signals will not be affected by multipath effects

Although methods based on special antenna structures and antenna arrays can suppress multipath reception, most of these methods rely on high-cost antenna equipment and have certain limitations, such as: choke coil antennas are not suitable for high-elevation-angle NLOS signals, etc.

And this type of method will cause the device to be too bulky and cannot be popularized in mobile applications.

In addition, the correlator improvement technology based on receiver signal processing can alleviate the multipath effect to a certain extent, but this technology is only effective for Multipath and cannot deal with NLOS, and the cost of components is also a problem that cannot be ignored

The camera-based method will be limited by the defects of the camera sensor itself, such as the influence of weather, light, carrier moving speed and other factors on its image quality, and the image quality will often seriously affect the performance of such methods, and the work stability is not good

Methods based on 3D map models and ray tracing will be affected by the loading speed of 3D map models and the processing speed of ray tracing, resulting in poor calculation efficiency. It may not be suitable for high-speed users such as cars. In addition, such methods have relatively high requirements for model accuracy. high

In addition, based on methods such as camera / 3D map model and ray tracing / dual polarized antenna / machine learning to determine the type of satellite signal reception and eliminate NLOS signals to improve positioning accuracy, there is still a common problem based on NLOS elimination, that is, in urban canyons When the number of environmental satellites is in danger, the elimination is still carried out. Although multi-constellation can alleviate the problems caused by these elimination methods to a certain extent, signal elimination will inevitably lead to problems such as insufficient number of available satellites and poor satellite geometry, which will eventually lead to the degradation of positioning accuracy.

Images