Method for positioning pile number of highway mileage pile

Abstract

The invention relates to the technical field of highway mileage pile number positioning, in particular to a highway mileage pile number positioning method.The highway mileage pile positioning system comprises a laser radar, an industrial camera, satellite positioning equipment, an IMU and an industrial personal computer, the IMU and the industrial personal computer are installed at the central position of the vehicle, and the laser radar and the industrial camera are used for joint calibration to obtain laser point cloud data. According to the invention, rapid, accurate and automatic matching of the highway stake number and the geodetic coordinate is realized. A laser point cloud and vision technology is applied, GPS positioning, IMU inertial navigation, computer image processing and other technologies are fused, the one-to-one correspondence relation between the highway mileage pile and geodetic coordinates is obtained, rapid conversion between the longitude and latitude of the mileage pile and the mileage pile number is achieved, the positioning precision reaches the decimeter level, and a convenient, accurate and efficient information identification method is provided for road maintenance management.

Description

Technical Field

[0001] The invention relates to the technical field of highway mileage pile number positioning, and in particular to a highway mileage pile number positioning method. Background Art

[0002] Highway positioning uses the highway mileage pile system and the national geodetic coordinate system as reference systems to achieve information identification of geographical locations such as structures along the line. The two can be converted through a specific mathematical model. In the process of highway maintenance and management, the mileage pile system is usually used as a reference system, while in the development of digital and intelligent management and maintenance, the national geodetic coordinate system is usually used as a reference system to achieve positioning. At the same time, during the operation and maintenance of mileage piles, displacement, deviation, etc. occur, resulting in positioning errors, requiring re-positioning and calibration. Therefore, it is very necessary to convert the geodetic coordinate system in digital management and maintenance into the mileage pile coordinate system in actual management and maintenance. However, the existing highway pile number and geodetic coordinate system conversion methods are not accurate and intelligent enough, and cannot meet the efficiency and accuracy of the pile number conversion method in actual applications. For example, the distance measurement method using monocular vision, due to the loss of depth information, causes the distance measurement error to increase with the increase of distance, especially in the absence of additional constraints, the error is greater. The method using binocular vision has high requirements for light, background and other environments, and has poor environmental adaptability; the above technology has the following specific defects:

[0003] The mileage post positioning algorithm based on pure vision needs to estimate the distance between the video car and the mileage post based on the GPS of the video car and obtain it through conversion. However, the existing distance calculation method uses a camera with a specific angle or binocular vision estimation, which cannot accurately calculate the relative position between the video car and the mileage post, including the azimuth and distance information;

[0004] The visual acquisition of mileage post images places high demands on the vehicle's driving lane, and it is necessary to be as close to the emergency parking lane as possible. At the same time, the image acquisition is easily affected by bad weather, and the pile number is easily missed or the acquired image is blurred, resulting in reduced distance calculation accuracy.

[0005] Therefore, in order to solve the above problems, a method for locating the mileage pile numbers of highways is proposed. Summary of the invention

[0006] The purpose of the present invention is to provide a method for locating the mileage pile numbers of highways to solve the problems raised in the above-mentioned background technology.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A method for locating the mileage pile number of a highway comprises the following steps:

[0009] Step S1, setting up a mileage post positioning system; the mileage post positioning system includes a laser radar, an industrial camera, a satellite positioning device, an IMU and an industrial computer, wherein the laser radar, the industrial camera and the satellite positioning device are installed on the roof facing the front and are set at the same height, and the IMU and the industrial computer are installed in the center of the vehicle, and the laser radar and the industrial camera are used for joint calibration to obtain laser point cloud data, and then the IMU and the industrial camera are initially calibrated to record the yaw angle, pitch angle and roll angle collected by the IMU in a stationary state;

[0010] Step S2, constructing a basic database; constructing a basic database to collect images of mileage signs and hundred-meter signs using laser radar and industrial cameras, storing images of mileage signs and hundred-meter signs and laser point cloud data corresponding to the images of mileage signs and hundred-meter signs, wherein the laser point cloud data includes the direction, intensity, and distance of the laser point; constructing a basic database to synchronously store IMU data during driving;

[0011] Step S3, retrieve 5000 images containing mileage signs collected in step S2, establish a mileage sign database, build an LPRnet model, train the model, detect mileage signs and identify the content of mileage signs;

[0012] Step S4, retrieve 10,000 images containing the 100-meter sign collected in step S2, build a YOLO series deep learning model, train the model to detect the 100-meter sign target in the image, segment and extract the 100-meter sign, and build a neural network classifier to identify the 100-meter sign value;

[0013] Step S5, calculate the location of the mileage sign and the hundred-meter sign through the positioning information of the camera; calculate the precise GPS location of the mileage sign or the hundred-meter sign through triangulation using the recorded yaw angle, pitch angle, roll angle and the distance and angle information recorded at the center of the mileage sign or the hundred-meter sign in the laser radar point cloud. The triangulation calculation formula is as follows:

[0014] long2=long1+d×sinθ / [ARC×cos(lat1)×2π / 360];

[0015] lat2=lat1+d×cosθ / (ARC×2π / 360);

[0016] Where long2, lat2 represent the longitude and latitude coordinates of the milepost to be solved, long1, lat1 represent the longitude and latitude coordinates of the industrial camera collected; d represents the distance from the industrial camera to the milepost; θ represents the deflection angle of the world coordinate system relative to the milepost position in the north direction; ARC represents the radius of the section at the current latitude;

[0017] The world coordinate system and the camera coordinate system are calculated through the translation and rotation matrix [R, t].

[0018] As a preferred solution, in step S1, the specific operation of joint calibration using laser radar and industrial camera is as follows:

[0019] Step S1-1, collect industrial camera images and radar point cloud data, extract industrial camera image points and radar points of the same name, and construct a PnP matrix;

[0020] Step S1-2: Use calibration software to complete the calibration of industrial camera and radar external parameters, obtain the transformation matrix [R, t], match the lidar point cloud with the image points, and obtain the lidar point cloud information through the image target.

[0021] As a preferred solution, the mileage post positioning system is installed on the roof of the vehicle close to the mileage post on the highway to collect images and radar data.

[0022] As a preferred solution, in step S4, the image of the hundred-meter sign is cropped with a length of 28 pixels and a width of 28 pixels with the detection result box as the coordinate center, and a label is established for the value of the hundred-meter sign, and the label value is 1-9.

[0023] As a preferred solution, in step S4, 4 convolutional layers plus 2 fully connected layers are used, ReLUs is used as the activation function, a neural network classifier is constructed, the optimizer is selected as adam, and the data is allocated to the training set and the validation set in a ratio of 8:2 to train the 100-meter sign numerical classification model.

[0024] It can be seen from the technical solution provided by the present invention that the method for locating the mileage pile number of a highway provided by the present invention has the beneficial effect of realizing fast, accurate and automatic matching of highway pile numbers with geodetic coordinates. By applying laser point cloud and visual technology, integrating GPS positioning, IMU inertial navigation, computer image processing and other technologies, a one-to-one correspondence between highway mileage piles and geodetic coordinates is obtained, and the rapid conversion between the longitude and latitude of mileage piles and mileage pile numbers is realized. The positioning accuracy reaches the decimeter level, providing a convenient, accurate and efficient information identification method for road maintenance management. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 This is a flow chart of high-precision positioning of mileage signs and hundred-meter signs in the present invention;

[0026] Figure 2 This is a schematic diagram of the coordinate calculation of the mileage sign and the hundred-meter sign in the present invention. DETAILED DESCRIPTION

[0027] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0028] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0029] like Figure 1-2 As shown, an embodiment of the present invention provides a method for locating a mileage post number of a highway, comprising the following steps:

[0030] Step S1, setting up a mileage post positioning system; the mileage post positioning system includes a laser radar, an industrial camera, a satellite positioning device, an IMU and an industrial computer, wherein the laser radar, the industrial camera and the satellite positioning device are installed on the roof facing the front and are set at the same height, and the IMU and the industrial computer are installed in the center of the vehicle. The laser radar and the industrial camera are used for joint calibration to obtain laser point cloud data, and then the IMU and the industrial camera are initially calibrated to record the yaw angle, pitch angle and roll angle collected by the IMU in a stationary state, and the laser radar and the industrial camera are used for joint calibration. The specific operations are as follows:

[0031] Step S1-1, collect industrial camera images and radar point cloud data, extract industrial camera image points and radar points of the same name, and construct a PnP matrix;

[0032] Step S1-2, use calibration software to complete the external parameter calibration of industrial cameras and radars, obtain the transformation matrix [R, t], match the laser radar point cloud with the image points, and obtain the laser radar point cloud information through the image target;

[0033] Step S2, constructing a basic database; constructing a basic database to collect images of mileage signs and hundred-meter signs using laser radar and industrial cameras, storing images of mileage signs and hundred-meter signs and laser point cloud data corresponding to the images of mileage signs and hundred-meter signs, wherein the laser point cloud data includes the direction, intensity, and distance of the laser point; constructing a basic database to synchronously store IMU data during driving;

[0034] Step S3, retrieve 5000 images containing mileage signs collected in step S2, establish a mileage sign database, build an LPRnet model, train the model, detect mileage signs and identify the content of mileage signs;

[0035] Step S4, retrieve 10,000 images containing the 100-meter sign collected in step S2, build a YOLO series deep learning model, train the model to detect the 100-meter sign target in the image, segment and extract the 100-meter sign, and build a neural network classifier to identify the 100-meter sign value;

[0036] The 100-meter sign is cut with the detection result frame as the coordinate center, and the 100-meter sign image with a length of 28 pixels and a width of 28 pixels is cropped. At the same time, a label is established for the 100-meter sign value, and the label value is 1-9;

[0037] Four convolutional layers plus two fully connected layers were used, ReLUs was used as the activation function, a neural network classifier was constructed, the optimizer was selected as adam, the data was divided into training set and validation set in a ratio of 8:2, and the 100-meter card numerical classification model was trained;

[0038] Step S5, calculate the location of the mileage sign and the hundred-meter sign through the positioning information of the camera; calculate the precise GPS location of the mileage sign or the hundred-meter sign through triangulation using the recorded yaw angle, pitch angle, roll angle and the distance and angle information recorded at the center of the mileage sign or the hundred-meter sign in the laser radar point cloud. The triangulation calculation formula is as follows:

[0039] long2=long1+d×sinθ / [ARC×cos(lat1)×2π / 360];

[0040] lat2=lat1+d×cosθ / (ARC×2π / 360);

[0041] Where long2, lat2 represent the longitude and latitude coordinates of the milepost to be solved, long1, lat1 represent the longitude and latitude coordinates of the industrial camera collected; d represents the distance from the industrial camera to the milepost; θ represents the deflection angle of the world coordinate system relative to the milepost position in the north direction; ARC represents the radius of the section at the current latitude;

[0042] The world coordinate system and the camera coordinate system are calculated through the translation and rotation matrix [R, t].

[0043] In this embodiment, the mileage post positioning system is installed on the roof of the vehicle close to the mileage post of the highway to collect images and radar data.

[0044] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for locating the mileage pile number of a highway, characterized by: The following steps are involved: Step S1, setting up a mileage post positioning system; the mileage post positioning system includes a laser radar, an industrial camera, a satellite positioning device, an IMU and an industrial computer, wherein the laser radar, the industrial camera and the satellite positioning device are installed on the roof facing the front and are set at the same height, and the IMU and the industrial computer are installed in the center of the vehicle, and the laser radar and the industrial camera are used for joint calibration to obtain laser point cloud data, and then the IMU and the industrial camera are initially calibrated to record the yaw angle, pitch angle and roll angle collected by the IMU in a stationary state; Step S2, constructing a basic database; the basic database is constructed by using laser radar and industrial cameras to collect images of mileage signs and hundred-meter signs, and storing images of mileage signs and hundred-meter signs and laser point cloud data corresponding to the images of mileage signs and hundred-meter signs, wherein the laser point cloud data includes the direction, intensity and distance of the laser point; the basic database is constructed to synchronously store IMU data of the driving process; Step S3, retrieve 5000 images containing mileage signs collected in step S2, establish a mileage sign database, build an LPRnet model, train the model, detect mileage signs and identify the content of mileage signs; Step S4, retrieve 10,000 images containing the 100-meter sign collected in step S2, build a YOLO series deep learning model, train the model to detect the 100-meter sign target in the image, segment and extract the 100-meter sign, and build a neural network classifier to identify the 100-meter sign value; Step S5, calculate the location of the mileage sign and the hundred-meter sign through the positioning information of the camera; calculate the precise GPS location of the mileage sign or the hundred-meter sign through triangulation using the recorded yaw angle, pitch angle, roll angle and the distance and angle information recorded at the center of the mileage sign or the hundred-meter sign in the laser radar point cloud. The triangulation calculation formula is as follows: long2=long1+d×sinθ / [ARC×cos(lat1)×2π / 360]; lat2=lat1+d×cosθ / (ARC×2π / 360); Where long2, lat2 represent the longitude and latitude coordinates of the milepost to be solved, long1, lat1 represent the longitude and latitude coordinates of the industrial camera collected; d represents the distance from the industrial camera to the milepost; θ represents the deflection angle of the world coordinate system relative to the milepost position in the north direction; ARC represents the radius of the section at the current latitude; The world coordinate system and the camera coordinate system are calculated through the translation and rotation matrix [R, t].

2. A method for locating the mileage pile number of a highway according to claim 1, characterized in that: In step S1, the specific operation of joint calibration using laser radar and industrial camera is as follows: Step S1-1, collect industrial camera images and radar point cloud data, extract industrial camera image points and radar points of the same name, and construct a PnP matrix; Step S1-2: Use calibration software to complete the calibration of industrial cameras and radar external parameters, obtain the transformation matrix [R, t], match the lidar point cloud with the image points, and obtain the lidar point cloud information through the image target.

3. A method for locating the mileage pile number of a highway according to claim 1, characterized in that: The mileage post positioning system is installed on the roof of the vehicle close to the mileage post on the highway to collect images and radar data.

4. A method for locating the mileage pile number of a highway according to claim 1, characterized in that: In step S4, the image of the hundred-meter sign is cropped with a length of 28 pixels and a width of 28 pixels, with the detection result frame as the coordinate center, and a label is established for the value of the hundred-meter sign, and the label value is 1-9.

5. The method for locating the mileage pile number of a highway according to claim 1, characterized in that: In step S4, four convolutional layers plus two fully connected layers are used, ReLUs is used as the activation function, a neural network classifier is constructed, the optimizer is selected as adam, and the data is allocated to the training set and the validation set in a ratio of 8:2 to train the 100-meter card numerical classification model.

Images