Vehicle positioning method, device, computer equipment and medium

Abstract

The present application provides a vehicle positioning method, apparatus, computer equipment, and medium, which acquire a two-dimensional image captured by a camera and three-dimensional point cloud data captured by a laser radar. The two-dimensional image includes a two-dimensional image of a target in an uneven area, and the three-dimensional point cloud data includes three-dimensional point cloud data of the target in the uneven area. The target two-dimensional image and the target three-dimensional point cloud data are matched to calculate the degree of ground potholes in the uneven area. The ground potholes are input into a target neural network model to obtain the ground flatness of the uneven area. When filtering the collected first position information of the vehicle using a Kalman filter algorithm, the observation error matrix parameters are adjusted according to the ground flatness to obtain the second position information of the vehicle. The method can adjust the degree of influence of sensor error on the second position information and compensate for the positioning data, thereby improving the accuracy of vehicle positioning and making the positioning result more robust and stable.

Description

Technical Field

[0001] The present application relates to the field of vehicles, and in particular to a vehicle positioning method, device, computer equipment and medium. Background Art

[0002] High-precision positioning is fundamental to vehicle operation. Vehicle positioning primarily utilizes GNSS (Global Navigation Satellite System), which offers centimeter-level accuracy and generally meets the required accuracy on smooth surfaces. However, when driving on uneven surfaces, the presence of potholes and bumps can cause the vehicle to jolt as it moves forward. This can cause high-frequency positioning information to deviate significantly from the planned trajectory, significantly impacting vehicle control accuracy.

[0003] For example, when unmanned agricultural machinery is performing fine operations in farmland, if the farmland surface is uneven and has large potholes and bumps, there will be bumps in the forward movement of the agricultural machinery. Due to the tilting and shaking of the vehicle, the positioning information will have a high frequency and large deviation from the planned trajectory, resulting in low positioning accuracy of the agricultural machinery. Summary of the Invention

[0004] In view of this, the purpose of this application is to provide a vehicle positioning method, apparatus, computer equipment, and medium to improve the accuracy of vehicle positioning and make the positioning results more robust and stable. The specific solution is as follows:

[0005] In one aspect, the present application provides a vehicle positioning method, comprising:

[0006] Acquire a two-dimensional image captured by a camera and three-dimensional point cloud data captured by a laser radar; the two-dimensional image includes a two-dimensional image of a target in an uneven area, and the three-dimensional point cloud data includes three-dimensional point cloud data of a target in an uneven area;

[0007] Matching the target two-dimensional image with the target three-dimensional point cloud data to calculate the degree of potholes in the uneven area;

[0008] Inputting the ground pothole degree into the target neural network model to obtain the ground flatness of the uneven area;

[0009] When filtering the collected first position information of the vehicle using the Kalman filter algorithm, the observation error matrix parameters are adjusted according to the ground flatness to obtain the second position information of the vehicle.

[0010] In another aspect, an embodiment of the present application further provides a vehicle positioning device, comprising:

[0011] An acquisition unit, configured to acquire a two-dimensional image acquired by a camera and three-dimensional point cloud data acquired by a lidar; the two-dimensional image includes a two-dimensional image of a target in an uneven area, and the three-dimensional point cloud data includes three-dimensional point cloud data of a target in an uneven area;

[0012] a first calculation unit, configured to match the target two-dimensional image with the target three-dimensional point cloud data to calculate the degree of potholes in the uneven area;

[0013] A second calculation unit is used to input the ground pothole degree into the target neural network model to obtain the ground flatness of the uneven area;

[0014] The processing unit is used to adjust the observation error matrix parameters according to the ground flatness when filtering the collected first position information of the vehicle through the Kalman filter algorithm to obtain the second position information of the vehicle.

[0015] In another aspect, an embodiment of the present application provides a computer device, comprising a processor and a memory:

[0016] The memory is used to store program code and transmit the program code to the processor;

[0017] The processor is configured to execute the vehicle positioning method described above according to the instructions in the program code.

[0018] On the other hand, an embodiment of the present application provides a computer-readable storage medium, which is used to store a computer program, and the computer program is used to execute the vehicle positioning method described in the above aspects.

[0019] On the other hand, an embodiment of the present application provides a computer program product including instructions, which, when executed on a computer, enables the computer to execute the vehicle positioning method described in the above aspects.

[0020] Embodiments of the present application provide a vehicle positioning method, apparatus, computer equipment, and medium, which obtain a two-dimensional image captured by a camera and three-dimensional point cloud data captured by a lidar; the two-dimensional image includes a target two-dimensional image of an uneven area, and the three-dimensional point cloud data includes target three-dimensional point cloud data of the uneven area; the target two-dimensional image and the target three-dimensional point cloud data are matched to calculate the degree of ground potholes in the uneven area; the degree of ground potholes is input into a target neural network model to obtain the ground flatness of the uneven area; when filtering the collected first position information of the vehicle through a Kalman filter algorithm, the observation error matrix parameters are adjusted according to the ground flatness to obtain the second position information of the vehicle.

[0021] In an embodiment of the present application, the collected first position information of the vehicle is processed by the Kalman filter algorithm, which can filter out noise and perform smoothing, and can effectively reduce the lateral control error. At the same time, the observation error matrix parameters are parameters required for the calculation of the Kalman filter algorithm, and their source is the sensor error, that is, the inaccuracy of the sensor. During the processing, the observation error matrix parameters are adjusted according to the flatness of the ground, which can adjust the degree of influence of the sensor error on the second position information and compensate for the positioning data, thereby improving the accuracy of vehicle positioning and making the positioning results more robust and stable. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0023] Figure 1 A schematic diagram of a vehicle positioning method according to an embodiment of the present application is shown;

[0024] Figure 2 A structural block diagram of a vehicle positioning device provided in an embodiment of the present application is shown;

[0025] Figure 3 A schematic diagram of the structure of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0026] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are described in detail below with reference to the accompanying drawings.

[0027] In the following description, many specific details are set forth to facilitate a full understanding of the present application. However, the present application may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present application. Therefore, the present application is not limited to the specific embodiments disclosed below.

[0028] As described in the background technology, when unmanned agricultural machinery is performing fine operations in farmland, when the farmland surface is uneven and has large potholes and bumps, there will be bumps during the forward movement of the agricultural machinery. Due to the tilting and shaking of the vehicle, the positioning information has a high frequency and a large deviation from the planned trajectory, resulting in low positioning accuracy of the agricultural machinery.

[0029] Based on the above technical problems, the embodiments of the present application provide a vehicle positioning method, device, computer equipment and medium to obtain a two-dimensional image captured by a camera and three-dimensional point cloud data captured by a lidar; the two-dimensional image includes a target two-dimensional image of an uneven area, and the three-dimensional point cloud data includes target three-dimensional point cloud data of the uneven area; the target two-dimensional image and the target three-dimensional point cloud data are matched to calculate the degree of ground potholes in the uneven area; the degree of ground potholes is input into a target neural network model to obtain the ground flatness of the uneven area; when the collected first position information of the vehicle is filtered by the Kalman filter algorithm, the observation error matrix parameters are adjusted according to the ground flatness to obtain the second position information of the vehicle.

[0030] In an embodiment of the present application, the collected first position information of the vehicle is processed by the Kalman filter algorithm, which can filter out noise and perform smoothing, and can effectively reduce the lateral control error. At the same time, the observation error matrix parameters are parameters required for the calculation of the Kalman filter algorithm, and their source is the sensor error, that is, the inaccuracy of the sensor. During the processing, the observation error matrix parameters are adjusted according to the flatness of the ground, which can adjust the degree of influence of the sensor error on the second position information and compensate for the positioning data, thereby improving the accuracy of vehicle positioning and making the positioning results more robust and stable.

[0031] For ease of understanding, a vehicle positioning method, apparatus, computer equipment, and medium provided in an embodiment of the present application are described in detail below with reference to the accompanying drawings.

[0032] refer to Figure 1 As shown, it is a flow chart of a vehicle positioning method provided in an embodiment of the present application, and the method may include the following steps.

[0033] S101, obtaining a two-dimensional image captured by a camera and three-dimensional point cloud data captured by a lidar.

[0034] In the embodiments of the present application, the camera can be an RGB camera or other monocular camera, which can be mounted on a vehicle to capture a two-dimensional image of the road ahead of the vehicle. For example, an RGB camera can be mounted centered on the front of a tractor hood. A laser radar can also be mounted on the vehicle to acquire three-dimensional point cloud data of the road ahead of the vehicle. For example, a laser radar can be mounted centered on the top of the hood.

[0035] In an embodiment of the present application, a two-dimensional image captured by a camera and three-dimensional point cloud data captured by a lidar can be obtained, and a two-dimensional image and three-dimensional point cloud data of the road ahead of the vehicle can be obtained.

[0036] Specifically, after the camera and lidar are fixed on the vehicle, they can be jointly calibrated to establish a correspondence between the two-dimensional image and the three-dimensional point cloud data to ensure that the camera and lidar have a unified reference standard and can be converted into each other, and then the two-dimensional image and three-dimensional point cloud data can be collected.

[0037] Specifically, since the road ahead has flat areas and uneven areas, and the uneven areas have potholes or bumps, the two-dimensional image includes the target two-dimensional image of the uneven area, and may also include the two-dimensional image of the flat area. The three-dimensional point cloud data includes the target three-dimensional point cloud data of the uneven area, and may also include the three-dimensional point cloud data of the flat area.

[0038] S102 , matching the target two-dimensional image with the target three-dimensional point cloud data, and calculating the degree of potholes in the uneven area.

[0039] In the embodiment of the present application, the road surface area can be divided. Since the vehicle mainly operates in a straight line, the road surface on both sides of the vehicle will basically have no impact on its driving. Therefore, it is necessary to pre-process the 2D image and 3D point cloud data to obtain the region of interest (ROI). For the region of interest, flat areas and uneven areas can be distinguished, so that the degree of potholes in the uneven area of ​​the region of interest can be determined.

[0040] Specifically, preprocessing the 2D image and 3D point cloud data involves cropping the 2D image captured by the camera and the 3D point cloud data collected by the LiDAR based on the vehicle's width and direction of travel to obtain a region of interest (ROI). This region is defined based on the vehicle's width and direction of travel. This region, known as the ROI, is then cropped, and only the pixels and 3D point cloud within this region are processed.

[0041] Then, when the uneven road area in the region of interest is determined, the region of interest can be divided into flat areas and uneven areas using a semantic segmentation algorithm, thereby obtaining a target two-dimensional image of the uneven area and target three-dimensional point cloud data of the uneven area. Specifically, semantic segmentation is a typical computational vision problem that involves taking some raw data such as a planar image as input and converting them into a mask with a highlighted region of interest. The two-dimensional image is input into the semantic segmentation algorithm. Since the image is composed of many pixels, semantic segmentation can segment the pixels according to the different semantic meanings expressed in the image, thereby outputting flat areas and uneven areas.

[0042] In an embodiment of the present application, for uneven areas, the target two-dimensional image and the target three-dimensional point cloud data can be matched to obtain the three-dimensional point cloud coordinates corresponding to each pixel point of the two-dimensional image, and the degree of ground potholes in the uneven area can be calculated. Since the vehicle will be bumpy when the ground is uneven, which will affect the positioning accuracy of the vehicle, the positioning position can be adaptively adjusted according to the degree of ground potholes to improve the accuracy of vehicle positioning. The degree of ground potholes can be determined based on the height of the potholes or bumps, or based on the height and area of ​​the potholes or bumps.

[0043] In one possible implementation, S102 may include matching the target two-dimensional image and the target three-dimensional point cloud data to obtain the point cloud coordinates of the uneven area, that is, obtaining the boundary information and depth information of the uneven area; then calculating the area S of the uneven area based on the point cloud coordinates of the uneven area, and then screening out the point cloud with the largest depth from the point cloud coordinates of the uneven area to obtain the depth H of the uneven area; and calculating the degree of potholes in the uneven area based on the area S and the depth H of the uneven area.

[0044] Specifically, the degree of potholes can be calculated using the formula K × H / S, where K is the conversion factor. Area S represents the size of the uneven area, and depth H represents the vertical height of the pothole or bump. Determining the degree of potholes based on S and H allows for more accurate determination of pothole conditions and, consequently, more precise positioning adjustments.

[0045] In another possible implementation, the degree of potholes on the ground may be determined based on the height of the potholes or bumps, which can reduce the amount of calculation and save time.

[0046] S103, inputting the degree of ground potholes into the target neural network model to obtain the ground flatness of the uneven area.

[0047] In the embodiment of the present application, before S103, the preset neural network model can be trained to obtain a trained target neural network model to output the ground flatness. The ground flatness can be the vertical deviation of the road surface from the ideal plane.

[0048] Specifically, the gyroscope can be used to collect multiple vehicle roll angles corresponding to multiple sampling points. The ground flatness can be determined based on the multiple vehicle roll angles and the number of sampling points. Specifically, the gyroscope installed on the vehicle can be used to obtain the vehicle roll angle when the agricultural machinery is traveling on an uneven area. The ground flatness calculation formula can be used to determine the ground flatness of the agricultural machinery passing through the potholes. The ground flatness calculation formula is:

[0049]

[0050] Among them, P represents the flatness of the ground in degrees (°), N represents the number of sampling points, r i Indicates the vehicle roll angle at each sampling point, in degrees (°).

[0051] Next, the preset neural network model can be trained based on the ground flatness and the degree of ground potholes. The degree of ground potholes is used as a vector input, and the ground flatness calculated by the ground flatness calculation formula is used as a vector output to train the target neural network model.

[0052] The target neural network model can be a BP neural network. The BP neural network is a multi-layer feedforward network trained by error back propagation. Its algorithm is called the BP algorithm. Its basic idea is the gradient descent method, which uses gradient search technology to minimize the mean square error between the actual output value and the expected output value of the network.

[0053] In an embodiment of the present application, when the target neural network model is applied, the degree of ground potholes can be input into the trained target neural network model to obtain the predicted ground flatness of the uneven area. The ground flatness can indicate the change in the vehicle state caused by the degree of ground potholes acting on the vehicle, and can reflect the degree of influence of the vehicle on the degree of ground potholes, that is, the bumpy condition of the vehicle, so as to subsequently adjust the positioning position of the vehicle according to the degree of influence of the vehicle.

[0054] S104 , when filtering the collected first position information of the vehicle using a Kalman filter algorithm, adjusting observation error matrix parameters according to the flatness of the ground to obtain second position information of the vehicle.

[0055] In an embodiment of the present application, when the first position information measured by GPS is used to locate an unmanned agricultural machinery in a farmland, the positioning information will fluctuate to varying degrees due to the size of potholes on the road, increasing the lateral control error of the vehicle and failing to meet the requirements of precise operation. Therefore, the collected first position information of the vehicle is filtered, and the Kalman filter algorithm is used to filter out noise and smooth the positioning data. The filtered positioning data will greatly reduce the error.

[0056] Specifically, the Kalman filter algorithm has an observation error matrix parameter R and a process error covariance matrix Q. The source of the observation error matrix parameter R is sensor error, that is, sensor inaccuracy. Each element of the process error covariance matrix Q is the covariance between the element errors of state X, caused by uncertain noise. The size of R and Q will affect whether the filtered data is more reliable, that is, the measured value measured by the sensor or the predicted value calculated by the model. Here, the measured value measured by the sensor refers to the first position information obtained by GPS positioning, and the predicted value calculated by the model refers to the predicted position information calculated by the prior model.

[0057] In an embodiment of the present application, the first position information of the vehicle can be collected through GPS positioning. When the first position information is filtered by the Kalman filter algorithm, the observation error matrix parameter R is adjusted according to the flatness of the ground to obtain the second position information of the vehicle. The second position information of the vehicle is used as the optimized vehicle position. The observation error matrix parameter is a parameter required for the calculation of the Kalman filter algorithm, and its source is the sensor error, that is, the inaccuracy of the sensor. During the processing, the observation error matrix parameter is adjusted according to the flatness of the ground, which can adjust the degree of influence of the sensor error on the second position information and compensate for the positioning data, thereby improving the accuracy of vehicle positioning and making the positioning result more robust and stable.

[0058] Specifically, the Kalman filter algorithm is an algorithm that uses a linear system state equation to optimally estimate system data through system input and output observation data. The Kalman filter algorithm used in this application can be a linear Kalman filter algorithm, etc.

[0059] In one possible implementation, when adjusting the observation error matrix parameters according to the ground flatness, when the ground flatness is less than a first preset threshold, it means that the ground is relatively flat, and the measured first position information will not deviate significantly and is relatively accurate. The observation error matrix parameters can be adjusted smaller, and when calculating the second position information, the sensor's measurement value, i.e., the first position information, can be more trusted, thereby improving the accuracy of vehicle positioning.

[0060] Specifically, when the ground flatness is greater than or equal to the first preset threshold, it means that the ground is more bumpy and uneven, and the vehicle is more bumpy. The measured first position information will produce a large offset as the sensor shakes significantly. The measured first position information is not very accurate. The observation error matrix parameters should be increased to reduce the trust in the first position information when calculating the second position information. This can further improve the vehicle's positioning accuracy and make the positioning results more robust and stable.

[0061] In an embodiment of the present application, S104 may specifically include adjusting observation error matrix parameters R and process error matrix parameters Q based on the flatness of the ground when filtering the collected first vehicle position information using the Kalman filter algorithm to obtain the second vehicle position information. Specifically, both observation error matrix parameters R and process error matrix parameters Q may be adjusted to further improve the vehicle's positioning accuracy and enhance the robustness and stability of the positioning result.

[0062] Specifically, when adjusting the observation error matrix parameters and the process error matrix parameters according to the ground flatness, when the ground flatness is less than the second preset threshold, it means that the ground is relatively flat, and the first position information obtained by the sensor measurement is more accurate than the predicted position information. The observation error matrix parameters can be reduced to increase the trust in the first position information, and the process error matrix parameters can be increased to reduce the trust in the predicted position information. Adaptive adjustments are made to both the measured values ​​and the predicted values, which can further improve the vehicle's positioning accuracy and make the positioning results more robust and stable.

[0063] Specifically, when the ground flatness is greater than or equal to the second preset threshold, it means that the ground is uneven with many potholes. The measured first position information will produce a large offset as the sensor shakes significantly. The first position information is less accurate than the predicted position information. The observation error matrix parameters can be increased to reduce the trust in the first position information, and the process error matrix parameters can be reduced to increase the trust in the predicted position calculated by the prior model. This can further improve the vehicle's positioning accuracy and make the positioning results more robust and stable.

[0064] The sizes of the first preset threshold and the second preset threshold are not specifically limited here, and those skilled in the art can set them according to actual conditions.

[0065] An embodiment of the present application provides a vehicle positioning method, which performs Kalman filter algorithm processing on the collected first position information of the vehicle, can filter out noise and perform smoothing, and can effectively reduce lateral control errors. At the same time, the observation error matrix parameters are parameters required for calculation of the Kalman filter algorithm, and their source is sensor error, that is, sensor inaccuracy. During the processing, the observation error matrix parameters are adjusted according to the flatness of the ground, which can adjust the degree of influence of the sensor error on the second position information and compensate for the positioning data, thereby improving the accuracy of vehicle positioning and making the positioning results more robust and stable.

[0066] Based on the above vehicle positioning method, the present application embodiment also provides a vehicle positioning device, referring to Figure 2FIG. 1 is a block diagram of a vehicle positioning device provided in an embodiment of the present application, and the device may include:

[0067] An acquisition unit 201 is configured to acquire a two-dimensional image captured by a camera and three-dimensional point cloud data captured by a lidar; the two-dimensional image includes a two-dimensional image of a target in an uneven area, and the three-dimensional point cloud data includes three-dimensional point cloud data of a target in an uneven area;

[0068] The first calculation unit 202 is configured to match the target two-dimensional image with the target three-dimensional point cloud data to calculate the degree of potholes in the uneven area;

[0069] The second calculation unit 203 is used to input the ground pothole degree into the target neural network model to obtain the ground flatness of the uneven area;

[0070] The processing unit 204 is configured to adjust observation error matrix parameters according to the ground flatness when filtering the collected first position information of the vehicle using a Kalman filter algorithm to obtain second position information of the vehicle.

[0071] Optionally, when adjusting the observation error matrix parameters according to the ground flatness, when the ground flatness is less than a first preset threshold, the observation error matrix parameters are adjusted down; when the ground flatness is greater than or equal to the first preset threshold, the observation error matrix parameters are adjusted up.

[0072] Optionally, the processing unit 204 is configured to:

[0073] When filtering the collected first position information of the vehicle using the Kalman filter algorithm, the observation error matrix parameters and the process error matrix parameters are adjusted according to the ground flatness to obtain the second position information of the vehicle.

[0074] Optionally, when adjusting the observation error matrix parameters and the process error matrix parameters according to the ground flatness, when the ground flatness is less than a second preset threshold, the observation error matrix parameters are adjusted down and the process error matrix parameters are adjusted up; when the ground flatness is greater than or equal to the second preset threshold, the observation error matrix parameters are adjusted up and the process error matrix parameters are adjusted down.

[0075] Optionally, the device further comprises:

[0076] a cropping unit, configured to crop the two-dimensional image captured by the camera and the three-dimensional point cloud data captured by the lidar to obtain a region of interest according to the width of the vehicle and the driving direction of the vehicle;

[0077] The segmentation unit is used to segment the region of interest by a semantic segmentation algorithm to obtain a target two-dimensional image of the uneven region and target three-dimensional point cloud data of the uneven region.

[0078] Optionally, the first computing unit 202 is configured to:

[0079] Matching the target two-dimensional image with the target three-dimensional point cloud data to obtain point cloud coordinates of the uneven area;

[0080] Calculating the area of ​​the uneven area according to the point cloud coordinates of the uneven area;

[0081] Filtering the point cloud with the largest depth from the point cloud coordinates of the uneven area to obtain the depth of the uneven area;

[0082] The degree of potholes in the uneven area is calculated according to the area of ​​the uneven area and the depth of the uneven area.

[0083] An embodiment of the present application provides a vehicle positioning device, which performs Kalman filter algorithm processing on the collected first position information of the vehicle, can filter out noise and perform smoothing, and can effectively reduce lateral control errors. At the same time, the observation error matrix parameters are parameters required for Kalman filter algorithm calculation, and their source is sensor error, that is, sensor inaccuracy. During the processing, the observation error matrix parameters are adjusted according to the flatness of the ground, which can adjust the degree of influence of the sensor error on the second position information and compensate for the positioning data, thereby improving the accuracy of vehicle positioning and making the positioning results more robust and stable.

[0084] On the other hand, the embodiment of the present application provides a computer device, referring to Figure 3 FIG. 1 is a schematic diagram of the structure of a computer device provided in an embodiment of the present application, wherein the computer device includes a processor 320 and a memory 310:

[0085] The memory is used to store program code and transmit the program code to the processor;

[0086] The processor is used to execute the vehicle positioning method provided by the above embodiment according to the instructions in the program code.

[0087] The computer device may include a terminal device or a server, and the aforementioned vehicle positioning apparatus may be configured in the computer device.

[0088] On the other hand, an embodiment of the present application further provides a storage medium, which is used to store a computer program, and the computer program is used to execute the vehicle positioning method provided by the above embodiment.

[0089] In addition, an embodiment of the present application further provides a computer program product including instructions, which, when executed on a computer, enables the computer to execute the vehicle positioning method provided in the above embodiment.

[0090] Those skilled in the art will understand that all or part of the steps of implementing the above-mentioned method embodiment can be completed by program instruction hardware, and the above-mentioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above-mentioned method embodiment; and the above-mentioned storage medium can be at least one of the following media: read-only memory (English: Read-only Memory, abbreviated: ROM), RAM, magnetic disk or optical disk, etc., various media that can store program codes.

[0091] The various embodiments in this specification are described in a progressive manner. Similar parts between the various embodiments can be referred to in conjunction with each other. Each embodiment focuses on the differences from the other embodiments. In particular, the device embodiments are generally similar to the method embodiments, so the description is relatively simple. For relevant parts, refer to the description of the method embodiments.

[0092] The above is only a preferred embodiment of the present application. Although the present application has been disclosed as a preferred embodiment, it is not intended to limit the present application. Any technician familiar with the art can use the above-disclosed methods and technical contents to make many possible changes and modifications to the technical solution of the present application without departing from the scope of the technical solution of the present application, or modify it into an equivalent embodiment with equivalent changes. Therefore, any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present application without departing from the content of the technical solution of the present application are still within the scope of protection of the technical solution of the present application.

Claims

1. A vehicle positioning method, characterized in that: include: Acquire a two-dimensional image captured by a camera and three-dimensional point cloud data captured by a laser radar; the two-dimensional image includes a two-dimensional image of a target in an uneven area, and the three-dimensional point cloud data includes three-dimensional point cloud data of a target in an uneven area; Matching the target two-dimensional image with the target three-dimensional point cloud data to calculate the degree of potholes in the uneven area; Inputting the ground pothole degree into the target neural network model to obtain the ground flatness of the uneven area; When filtering the collected first position information of the vehicle using a Kalman filter algorithm, adjusting observation error matrix parameters according to the ground flatness to obtain second position information of the vehicle; When filtering the collected first position information of the vehicle by the Kalman filter algorithm, adjusting the observation error matrix parameters according to the ground flatness to obtain the second position information of the vehicle includes: When filtering the collected first position information of the vehicle using a Kalman filter algorithm, adjusting observation error matrix parameters and process error matrix parameters according to the ground flatness to obtain second position information of the vehicle; When adjusting the observation error matrix parameters and the process error matrix parameters according to the ground flatness, when the ground flatness is less than a second preset threshold, the observation error matrix parameters are adjusted down and the process error matrix parameters are adjusted up; when the ground flatness is greater than or equal to the second preset threshold, the observation error matrix parameters are adjusted up and the process error matrix parameters are adjusted down.

2. The method according to any one of claim 1, characterized in that Before matching the target two-dimensional image with the target three-dimensional point cloud data to calculate the pothole degree of the uneven area, the method further includes: According to the width of the vehicle and the driving direction of the vehicle, the two-dimensional image captured by the camera and the three-dimensional point cloud data captured by the lidar are cropped to obtain an area of ​​interest; The region of interest is divided by a semantic segmentation algorithm to obtain a target two-dimensional image of the uneven region and target three-dimensional point cloud data of the uneven region.

3. The method according to claim 1, characterized in that The matching of the target two-dimensional image and the target three-dimensional point cloud data to calculate the degree of potholes in the uneven area includes: Matching the target two-dimensional image with the target three-dimensional point cloud data to obtain point cloud coordinates of the uneven area; Calculating the area of ​​the uneven area according to the point cloud coordinates of the uneven area; Filtering the point cloud with the largest depth from the point cloud coordinates of the uneven area to obtain the depth of the uneven area; The degree of potholes in the uneven area is calculated according to the area of ​​the uneven area and the depth of the uneven area.

4. The method according to claim 1, wherein Before inputting the ground pothole degree into the target neural network model to obtain the ground flatness of the uneven area, the method further includes: Multiple vehicle roll angles corresponding to multiple sampling points are collected by the gyroscope; determining the flatness of the ground based on the plurality of vehicle roll angles and the number of sampling points; The preset neural network model is trained according to the flatness of the ground and the degree of potholes in the ground to obtain a target neural network model.

5. A vehicle positioning device, characterized in that: include: An acquisition unit, configured to acquire a two-dimensional image acquired by a camera and three-dimensional point cloud data acquired by a lidar; the two-dimensional image includes a two-dimensional image of a target in an uneven area, and the three-dimensional point cloud data includes three-dimensional point cloud data of a target in an uneven area; a first calculation unit, configured to match the target two-dimensional image with the target three-dimensional point cloud data to calculate the degree of potholes in the uneven area; A second calculation unit is used to input the ground pothole degree into the target neural network model to obtain the ground flatness of the uneven area; a processing unit configured to adjust observation error matrix parameters according to the ground flatness when filtering the collected first position information of the vehicle using a Kalman filter algorithm to obtain second position information of the vehicle; The processing unit is configured to: When filtering the collected first position information of the vehicle using a Kalman filter algorithm, adjusting observation error matrix parameters and process error matrix parameters according to the ground flatness to obtain second position information of the vehicle; When adjusting the observation error matrix parameters and the process error matrix parameters according to the ground flatness, when the ground flatness is less than a second preset threshold, reducing the observation error matrix parameters and increasing the process error matrix parameters; When the ground flatness is greater than or equal to a second preset threshold, the observation error matrix parameters are increased and the process error matrix parameters are decreased.

6. A computer device, characterized in that: The computer device includes a processor and a memory: The memory is used to store program code and transmit the program code to the processor; The processor is configured to execute the vehicle positioning method according to any one of claims 1 to 4 according to instructions in the program code.

7. A computer-readable storage medium, characterized in that The computer-readable storage medium is used to store a computer program, and the computer program is used to execute the vehicle positioning method according to any one of claims 1 to 4.

Images