CN111680709A

CN111680709A - Positioning method based on environmental picture feature matching

Info

Publication number: CN111680709A
Application number: CN201910246792.3A
Authority: CN
Inventors: 张亮; 秦伦; 彭俊
Original assignee: Wuhan Xiaoshi Technology Co ltd
Current assignee: Wuhan Xiaoshi Technology Co ltd
Priority date: 2019-03-11
Filing date: 2019-03-11
Publication date: 2020-09-18

Abstract

The invention discloses a positioning method based on environmental picture feature matching, which is characterized by collecting pictures of an environment and corresponding positions and postures of the pictures in advance to generate a data set, shooting a current picture and comparing the current picture with a final picture of the data set when the environment is entered again, finding the most similar picture and acquiring the corresponding position of the most similar picture. The positioning method can carry out rough positioning under the condition of no GPS, and can retrieve the correct position under the condition of no positioning loss of the unmanned vehicle.

Description

Positioning method based on environmental picture feature matching

Technical Field

The invention relates to a positioning technology of unmanned driving, in particular to a method for roughly positioning an unmanned vehicle under the condition that a GPS (global positioning system) is not available or only weak signals exist in the GPS.

Background

Positioning is very important in unmanned driving, and the unmanned vehicle cannot operate without positioning and knowing the position of the unmanned vehicle. The current unmanned vehicle generally adopts a multi-sensor fusion scheme for positioning. One of the schemes is to combine gps and a laser radar, the laser radar needs to give an initial value for positioning, the position acquired by gps can be provided to the laser radar through coordinate conversion under the condition of gps signals, and the initial value can be acquired only by other modes under the condition of no gps.

Generally, the appearance of a building and the environment at two sides of a road are small in change, and more feature points are provided, so that the method is suitable for matching and positioning by using pictures. The algorithm provides a method for positioning by using an environment picture, an approximate position can be provided under specific scenes (without GPS signals and the like), and then a laser radar is used for finding an accurate position, so that an unmanned vehicle can quickly find the correct position when losing the position.

Disclosure of Invention

The invention discloses a positioning method based on picture feature matching, which can carry out rough positioning under the condition of no GPS and retrieve the correct position for an unmanned vehicle under the condition of losing positioning.

In order to solve the problems, the technical scheme adopted by the invention is as follows:

the method comprises the steps of collecting environmental pictures of a site to be positioned at multiple angles by using a camera, storing position information corresponding to the pictures at positions acquired by other positioning sensors, and making into a data set. During the positioning, pictures of the surrounding environment of the current position are taken and matched with the database to find the picture which is most similar to the current taken picture, so that the current position is found.

As a preferable scheme of the positioning method based on the environmental picture feature matching, the positioning sensor is used for acquiring absolute coordinates and a posture of the device in an environment when the environmental picture is taken, and storing the absolute coordinates and the posture in a coordinate file. The serial numbers of the pictures correspond to the line numbers in the coordinate file one by one.

As a preferable scheme of the positioning method based on the environmental picture feature matching, feature points of all the shot environmental pictures are extracted, feature vectors are generated, the feature vectors are associated with the numbers of the pictures and are stored in a data set.

As a preferred scheme of the positioning method based on the environmental picture feature matching, when the positioning environment is re-entered, a camera shoots current environmental pictures from multiple angles, feature points of each picture are extracted and feature vectors are generated, the feature vectors of the current pictures are matched with the feature vectors in the data set, the picture with the closest features is found, and the approximate position is further found according to the corresponding relation.

As a preferred scheme of the positioning method based on the environmental picture feature matching, the feature vector matching is completed through quick nearest neighbor searching, and the best matching can be found in a short time.

As a preferred scheme of a positioning method based on environmental picture feature matching, the approximate picture found by matching is usually more than one picture, but more than one picture, so that several most aggregated positions are selected by adopting a clustering method corresponding to the possible positions, and the clustering center of the position is taken as the most possible position.

Drawings

FIG. 1 is a flow chart of an embodiment of a positioning method based on environmental picture feature matching according to the present application;

Detailed Description

The multi-view camera and the positioning equipment are fixed on the vehicle and connected with the computer, and the positioning program and the data acquisition program are started.

The method comprises the steps of shooting environment pictures of the left side and the right side of a road respectively along the advancing direction of the road, wherein the shooting time interval is related to the advancing speed, the speed is not too high, the overlapped part of the collected contents between adjacent pictures is required to be ensured to be more than 50%, the pictures are clear, no moving shielding objects such as automobiles and the like exist, the position and posture information corresponding to the pictures is stored while the pictures are shot, and one picture corresponds to one position and posture information. And on the same road, the environment picture and the position information are collected once to the right back and forth respectively, a more complete data set is obtained, and the position information is saved as a file.

And extracting feature points of the shot environment picture and generating a feature vector data set. According to self requirements, feature points are extracted, the number of the feature points of each picture can be determined according to the situation, the matching success rate is higher when the number of the feature points is more, and the calculation time is longer. The number of the feature vectors is set to 1000, and all the feature vectors are stored in a file as a matching database, and the arrangement order of the feature vectors matches the arrangement order of the pictures.

And when entering a road section which is collected with data next time, starting a picture positioning program, shooting a left environmental picture and a right environmental picture, extracting feature points and generating feature vectors, then performing search matching by using a nearest search algorithm, selecting a result with a large number of matched feature points, and finding out a picture number which is close to the picture shot at this time in a database file, wherein possibly a plurality of pictures are similar to the given picture. And taking out all possible positions according to the corresponding relation between the pictures and the positions, finding the position with the most aggregation by using a clustering method, and taking the center of the positions as a final positioning result.

The above-mentioned embodiments are merely preferred technical solutions of the present invention, and should not be construed as limiting the present invention. The protection scope of the present invention is defined by the claims, and includes equivalents of technical features of the claims. I.e., equivalent alterations and modifications within the scope hereof, are also intended to be within the scope of the invention.

Claims

1. A positioning method based on environmental picture feature matching comprises the following steps: the method comprises the steps of collecting environmental pictures of a site to be positioned at multiple angles by using a camera, storing position information corresponding to the pictures at positions acquired by other positioning sensors, and making into a data set. During the positioning, pictures of the surrounding environment of the current position are taken and matched with the database to find the picture which is most similar to the current taken picture, so that the current position is found.

2. The method for positioning based on environmental picture feature matching according to claim 1, wherein: the positioning sensor is used for acquiring absolute coordinates and postures of the equipment in the environment when the environment picture is shot, and storing the absolute coordinates and the postures in the coordinate file. The serial numbers of the pictures correspond to the line numbers in the coordinate file one by one.

3. The method of claim 2, wherein the method comprises: and extracting feature points of all the shot environment pictures and generating a feature vector, wherein the feature vector is associated with the numbers of the pictures and is stored in a data set.

4. The method according to claim 3, wherein the method comprises: when the mobile terminal enters the positioning environment again, the camera shoots the current environment pictures from multiple angles, the feature point of each picture is extracted and a feature vector is generated, the feature vector of the current picture is matched with the feature vector in the data set, the picture with the closest features is found, and the approximate position is further found according to the corresponding relation.

5. The method of claim 4, wherein the method comprises: the feature vector matching is done by a fast nearest neighbor search, and the best match can be found in a short time.

6. The method of claim 4, wherein the method comprises: the approximate picture found by matching is usually more than one picture, but more than one picture, so that several positions which are most gathered are selected by adopting a clustering method corresponding to a plurality of possible positions, and the clustering center of the positions is taken as the most possible position.