CN-120195710-B - High-precision positioning method based on vehicle-road cooperation in transitional scene

CN120195710BCN 120195710 BCN120195710 BCN 120195710BCN-120195710-B

Abstract

The invention provides a high-precision positioning method under a transition scene based on vehicle-road cooperation, which comprises the steps of firstly, shooting a road scene in front of vehicle driving by using a vehicle-mounted camera to obtain an image to be matched, obtaining a positioning output result of a vehicle end based on GNSS signals by using a GNSS receiver, obtaining a visual positioning output result based on a road side end by using a road side camera, secondly, constructing a transition scene database, identifying the transition scene by combining global feature matching and local feature matching based on the image to be matched, and finally, smoothly switching the positioning output result of the vehicle end based on the GNSS signals and the visual positioning output result based on the road side end in a transition scene area. The invention provides a new solution for the defect of the traditional positioning mode of the transition scene, can effectively solve the problem of insufficient positioning precision of the current transition scene, ensures the continuity and smoothness of the track in the running process of the intelligent automobile, and has wide market prospect.

Inventors

LI DAICHENG
GAO CHAOJUN
JIANG JIN
CAI YINGFENG
WANG HAI
CHEN LONG

Assignees

江苏大学

Dates

Publication Date: 20260512
Application Date: 20250224

Claims (7)

1. A high-precision positioning method based on vehicle-road cooperation in a transition scene is characterized by comprising the following steps: Firstly, taking a picture of a road scene in front of the running of a vehicle by using a vehicle-mounted camera to obtain an image to be matched, obtaining a positioning output result of a vehicle end based on GNSS signals by using a GNSS receiver, and obtaining a visual positioning output result based on a road side end by using a road side end camera; secondly, constructing a transition scene database, and identifying the transition scene by combining global feature matching and local feature matching based on the image to be matched, wherein the method specifically comprises the following steps: The global feature matching process comprises the steps of firstly calculating Euclidean distance between a global feature descriptor of an image to be matched and global feature descriptors of images in a transition scene database, then sequentially comparing Euclidean distances of the image to be matched and all images in the transition scene database to obtain comparison of similarity between the images, and finally selecting the first N images with highest similarity as global feature matching results; The local feature matching process comprises that firstly, the local feature matching is performed on the global feature matching result, and then Is a candidate image with highest similarity with the image to be matched in the transition scene database, and is selected from the transition scene database The eight adjacent pictures are used as a search space for accurate matching and positioning, the search space S for matching all local features has 9N images, the most similar matching is described by calculating the Hamming distance between the local feature descriptors of the images to be matched and the local feature descriptors of the images in the search space S, and finally, the successfully matched images are arranged in descending order of the number of correctly matched points of the local feature descriptors to generate a node set ; Sampling point positions corresponding to the Q 1 node and nearest eight neighborhood position information in a transition scene database are selected, and vehicle positioning information (x l ,y l ) based on vehicle-road cooperation is obtained through weighted fusion, wherein the calculation formula is as follows: where n represents the number of the Q 1 node and its nearest eight neighborhood location information, n=1, 2, 9, Representing the weighting coefficients, x n representing the abscissa of the image sample points in the database, y n representing the ordinate of the image sample points in the database, Q n represents the approach weight coefficient, λ is the empirical parameter, 0< λ <1, k n represents the image matching point logarithm; And finally, in a transition scene area, smoothly switching the positioning output result of the vehicle end based on the GNSS signals and the visual positioning output result of the road side end.
2. The high-precision positioning method based on the vehicle-road cooperation under the transition scene according to claim 1, wherein the construction of the transition scene database comprises the following specific steps: And photographing the road scene by setting data sampling points for the transition scene at the road side along the road direction of the transition scene, and forming a transition scene database by using global feature descriptors and local feature descriptors of the sampling point images.
3. The high-precision positioning method based on the vehicle-road cooperation transition scene according to claim 2, wherein the global feature descriptors are extracted by adopting a CNN and NetVLAD framework method, and the local feature descriptors are obtained by extracting ORB features from each sampling point image.
4. The method for positioning with high precision under a transitional scene based on vehicle-road cooperation according to claim 1, wherein the transitional scene comprises a transitional scene moving from outside into the room and a transitional scene moving from the room to the outside.
5. The vehicle-road-collaboration-based high-precision positioning method under a transitional scene according to claim 1, wherein the smooth switching comprises a transition from an outdoor to an indoor scene positioning mode and a transition from an indoor to an outdoor scene positioning mode.
6. The high-precision positioning method based on the vehicle-road cooperation under the transitional scene of claim 5, which is characterized in that the transition of the outdoor-to-indoor scene positioning mode comprises the following specific processes: when entering a transition scene from an outdoor scene, positioning output result based on GNSS signals at the vehicle end Smooth transition to co-location output result based on vehicle road The real-time positioning coordinates (x t ,y t ) of the vehicle are obtained, and the calculation formula is as follows: Wherein, the E is a natural constant, t is a time variable taking the successful recognition moment of the transition scene as a starting zero point, the moment of the successful recognition of the crossing site scene is the 0 moment of the switching weight function, ; When entering an indoor scene from a transition scene, outputting a result based on vehicle-road co-positioning Smooth transition to roadside camera-based visual positioning output result The real-time positioning coordinates (x t ,y t ) of the vehicle are obtained, and the calculation formula is as follows: 。
7. The high-precision positioning method based on the vehicle-road cooperation under the transitional scene of claim 5, which is characterized in that the transition of the indoor-to-outdoor scene positioning mode comprises the following specific processes: when entering a transition scene from an indoor scene, outputting a result based on visual positioning of a roadside end camera Smooth transition to co-location output result based on vehicle road The real-time positioning coordinates (x t ,y t ) of the vehicle are obtained, and the calculation formula is as follows: Wherein, the E is a natural constant, t is a time variable taking the successful recognition moment of the transition scene as a starting zero point, the moment of the successful recognition of the crossing site scene is the 0 moment of the switching weight function, ; When entering an outdoor scene from a transition scene, outputting a result based on vehicle-road co-positioning Smooth transition to positioning output result based on vehicle-end GNSS signals The real-time positioning coordinates (x t ,y t ) of the vehicle are obtained, and the calculation formula is as follows: 。

Description

High-precision positioning method based on vehicle-road cooperation in transitional scene Technical Field The invention belongs to the technical field of intelligent automobile positioning, and particularly relates to a high-precision positioning method based on a vehicle-road cooperation transition scene. Background With the rapid development of technology, intelligent automobiles have become an important component of modern traffic systems. The intelligent automobile not only can improve the driving safety, but also can improve the traffic efficiency and reduce the environmental pollution. The intelligent automobile positioning technology is taken as one of the core technologies, is an important component of a modern intelligent traffic system, and is also a basis for realizing functions of automatic driving, path planning, vehicle monitoring and the like. The development of the method is subjected to the evolution process from single technology to multi-technology fusion to intelligent application. With the continuous progress of technology, intelligent automobile positioning technology will play an increasingly important role in the fields of automatic driving, traffic management, smart city construction and the like. The high-precision positioning is the basis for realizing the automatic driving path planning, the positioning error directly influences the accuracy and precision of the planning and control algorithm module, and the positioning technology of the intelligent automobile has been developed and broken through in the prior art. At present, the most widely applied vehicle positioning technology is a global navigation satellite system (Global Navigation SATELLITE SYSTEM, GNSS), satellite navigation for short, and most of devices using GNSS for positioning use single-point GNSS positioning technology, generally only achieve meter-level precision due to simple principle and low cost, and in recent years, with the development of technologies such as precise single-point positioning (Pecise Point Positioning, PPP), real-time dynamic differential (Real-TIME KINEMATIC, RTK) and the like, the positioning precision can reach centimeter level. However, due to the existence of the canyon effect, the single use of the GNSS technology for positioning has the restriction of application scenes, and in order to overcome the defect of single technology in outdoor scenes, the combined positioning technology based on the GNSS and INS (inertial navigation system ) is developed, the positioning precision and reliability can be effectively improved through data fusion, and in addition, with the development of sensor technology, various sensors such as a laser radar (LiDAR), a camera, an ultrasonic sensor and the like are introduced into a vehicle positioning system, the perception capability of the vehicle to the surrounding environment is improved, and the positioning accuracy and stability are further improved. However, the positioning technology in the indoor and outdoor transition scenes is rarely researched at present, GNSS signals are often influenced before moving from the outdoor open environment to the scenes such as indoor, tunnel, underground garage, under-overhead bridge and the like, the positioning precision of the transition scenes is severely restricted, and the requirement of the intelligent automobile on high-precision positioning is difficult to meet. Disclosure of Invention Aiming at the defects in the prior art, the invention provides a high-precision positioning method based on vehicle-road cooperation in a transitional scene. The present invention achieves the above technical object by the following means. A high-precision positioning method based on vehicle-road cooperation in a transition scene comprises the following steps: Firstly, taking a picture of a road scene in front of a vehicle running by using a vehicle-mounted camera to obtain an image to be matched, obtaining a positioning output result of a vehicle terminal based on GNSS signals by using a GNSS receiver, obtaining a visual positioning output result of the vehicle terminal based on the road side by using a road side camera, secondly, constructing a transition scene database, identifying a transition scene by combining global feature matching and local feature matching based on the image to be matched, and finally, smoothly switching the positioning output result of the vehicle terminal based on the GNSS signals and the visual positioning output result based on the road side in a transition scene area. Further, the construction of the transition scene database comprises the following specific steps: And photographing the road scene by setting data sampling points for the transition scene at the road side along the road direction of the transition scene, and forming a transition scene database by using global feature descriptors and local feature descriptors of the sampling point images. Still further, the global feature descript