Vehicle location positioning method, device, computer equipment and storage medium

Abstract

The present application relates to a vehicle positioning method, device, computer equipment and storage medium. The above method includes: obtaining the roll angle and pitch angle corresponding to the vehicle; inputting the roll angle and pitch angle into a ground posture map model, extracting the roll angle and pitch angle through the ground posture map model, and outputting the corresponding position coordinates; determining the corresponding position of the vehicle according to the position coordinates. The use of this method can improve the accuracy of the vehicle position.

Description

Technical Field

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

[0002] Among the various positioning methods of autonomous driving, traditional inertial navigation and wheel speed dead reckoning will accumulate errors over time. The global navigation satellite system is affected by satellite signal obstruction, multipath effects and signal interference. LiDAR matching positioning also relies on high-precision maps (HDMap) and surrounding environment features, and visual positioning also relies on high-precision maps (HDMap) and visual semantic features.

[0003] Although these positioning methods complement each other to ensure high availability of high-precision positioning, there are still great risks in some scenarios. For example, when a vehicle drives into a long tunnel, there is a degradation problem in lidar matching positioning and visual positioning. Although the longitudinal direction can be obtained by recursion using the inertial measurement unit and wheel speed data, when the recursion time of the inertial measurement unit and wheel speed data is too long, the longitudinal accuracy cannot be guaranteed. In highway scenarios, once RTK (Real-time kinematic) encounters signal interference or an active ionosphere, longitudinal positioning cannot be guaranteed if it only relies on lidar matching and visual positioning, because the longitudinal features of lidar and visual highway scenes are very similar in length, and longitudinal degradation is very likely to occur. Summary of the invention

[0004] Based on this, it is necessary to provide a vehicle positioning method, device, computer equipment and storage medium that can improve the accuracy of the vehicle position in order to solve the above technical problems.

[0005] A vehicle positioning method, the method comprising:

[0006] Get the roll angle and pitch angle corresponding to the vehicle;

[0007] The roll angle and pitch angle are input into the ground attitude map model, and the features of the roll angle and pitch angle are extracted through the ground attitude map model to output the corresponding position coordinates;

[0008] Determine the corresponding position of the vehicle according to the position coordinates.

[0009] In one embodiment, the ground posture map model includes a long short-term memory network and a convolutional neural network, and the roll angle and the pitch angle are input into the ground posture map model, and the roll angle and the pitch angle are extracted by the ground posture map model to output the corresponding position coordinates, including:

[0010] Inputting the roll angle and the pitch angle into the long short-term memory network, extracting features of the roll angle and the pitch angle through the long short-term memory network, and outputting a first feature sequence;

[0011] The roll angle and the pitch angle are input into a convolutional neural network, and features of the roll angle and the pitch angle are extracted through the convolutional neural network to output a second feature sequence;

[0012] The first feature sequence and the second feature sequence are fused and the corresponding position coordinates are output.

[0013] In one of the embodiments, the training step of the ground posture map model includes: obtaining a training sample set, the training sample set includes at least one training sample, the training sample includes a training roll angle and a training pitch angle, and the training sample carries a corresponding real position coordinate;

[0014] Input the training sample set into the untrained ground posture map model, extract features of the training samples through the untrained ground posture map model, and output the corresponding training position coordinates;

[0015] The loss value is calculated based on the training position coordinates and the corresponding real position coordinates;

[0016] Determine whether the untrained ground posture map model has reached the convergence condition based on the loss value;

[0017] When the convergence condition is reached, the trained ground posture map model is obtained.

[0018] In one embodiment, obtaining a training sample set includes:

[0019] Determine a plurality of different sampling intervals and corresponding sequence lengths;

[0020] The corresponding training roll angle and training pitch angle are collected according to the sampling interval and sequence length to obtain a training sample set.

[0021] In one embodiment, the ground posture map model includes a long short-term memory network and a convolutional neural network, and the training sample set is input into the untrained ground posture map model, and the untrained ground posture map model is used to extract features of the training samples and output corresponding training position coordinates, including:

[0022] The training sample set is divided into a long-distance training sample set and a short-distance training sample set according to different sequence lengths and sampling intervals;

[0023] Input the long-distance training sample set into the long short-term memory network, extract features from the long-distance training sample set through the long short-term memory network, and obtain a long-distance feature sequence;

[0024] The short-distance training sample set is input into the convolutional neural network, and the convolutional neural network is used to extract features of the short-distance training sample set to obtain a short-distance feature sequence;

[0025] The corresponding training position coordinates are generated according to the long-distance feature sequence and the short-distance feature sequence.

[0026] In one embodiment, the method further comprises:

[0027] When the convergence condition is not reached, the model parameters of the untrained ground posture map model are continuously adjusted, and the step of obtaining the training sample set is returned;

[0028] Until the convergence condition is reached, a trained ground posture map model is obtained.

[0029] In one of the embodiments, the training samples in the training sample set also include heading angle, turning angles of the front two wheels, vehicle speed, angular velocity and acceleration.

[0030] A vehicle position positioning device, the device comprising:

[0031] An acquisition module is used to acquire the roll angle and pitch angle corresponding to the vehicle;

[0032] A prediction module is used to input the roll angle and the pitch angle into the ground attitude map model, extract the features of the roll angle and the pitch angle through the ground attitude map model, and output the corresponding position coordinates;

[0033] The determination module is used to determine the position corresponding to the vehicle according to the position coordinates.

[0034] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the following steps are implemented:

[0035] Get the roll angle and pitch angle corresponding to the vehicle;

[0036] The roll angle and pitch angle are input into the ground attitude map model, and the features of the roll angle and pitch angle are extracted through the ground attitude map model to output the corresponding position coordinates;

[0037] Determine the corresponding position of the vehicle according to the position coordinates.

[0038] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the following steps:

[0039] Get the roll angle and pitch angle corresponding to the vehicle;

[0040] The roll angle and pitch angle are input into the ground attitude map model, and the features of the roll angle and pitch angle are extracted through the ground attitude map model to output the corresponding position coordinates;

[0041] Determine the corresponding position of the vehicle according to the position coordinates.

[0042] The above-mentioned vehicle positioning method, device, computer equipment and storage medium obtain the roll angle and pitch angle corresponding to the vehicle; input the roll angle and pitch angle into the ground posture map model, extract the features of the roll angle and pitch angle through the ground posture map model, and output the corresponding position coordinates; determine the corresponding position of the vehicle according to the position coordinates. Therefore, the roll angle and pitch angle generated by the vehicle during driving can be used to predict and locate the current position of the vehicle, thereby improving the accuracy of the vehicle's position. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A diagram showing an application environment of a vehicle positioning method in an embodiment;

[0044] Figure 2 A schematic diagram of a process flow of a vehicle positioning method in one embodiment;

[0045] Figure 3 A schematic diagram of a process for extracting features of a ground posture map model in one embodiment;

[0046] Figure 4 A schematic diagram of a process of training a ground posture map model in one embodiment;

[0047] Figure 5 A flowchart of a training sample set acquisition step in one embodiment;

[0048] Figure 6 A schematic diagram of a process of training a ground posture map model in one embodiment;

[0049] Figure 7 A schematic diagram of a process of training a ground posture map model in one embodiment;

[0050] Figure 8 is a structural block diagram of a vehicle position positioning device in one embodiment;

[0051] Fig. 9 is an internal structure diagram of a computer device in one embodiment;

[0052] Fig.10 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application 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 application and are not used to limit the present application.

[0054] The vehicle positioning method provided in this application can be applied to Figure 1 In the application environment shown, the vehicle terminal 102 communicates with the server 104 through a network. The vehicle terminal 102 can be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, and portable wearable devices connected to the vehicle, and the server 104 can be implemented as an independent server or a server cluster composed of multiple servers.

[0055] Specifically, the vehicle-mounted terminal 102 obtains the roll angle and pitch angle corresponding to the vehicle and sends them to the server 104. The server 104 inputs the received roll angle and pitch angle into the ground posture map model, extracts features of the roll angle and pitch angle through the ground posture map model, outputs the corresponding position coordinates, and determines the corresponding position of the vehicle based on the position coordinates.

[0056] In one embodiment, the vehicle-mounted terminal 102 obtains the roll angle and pitch angle corresponding to the vehicle, inputs the roll angle and pitch angle into a ground posture map model, performs feature extraction on the roll angle and pitch angle through the ground posture map model, outputs the corresponding position coordinates, and determines the corresponding position of the vehicle according to the position coordinates.

[0057] In one embodiment, Figure 2 As shown, a vehicle positioning method is provided, which is applied to Figure 1 The vehicle terminal or server in the example is used to illustrate, including the following steps:

[0058] Step 202, obtaining the roll angle and pitch angle corresponding to the vehicle.

[0059] Among them, the uneven road surface characteristics include the posture characteristics of the road surface, and the posture characteristics of the road surface include roll angle and pitch angle. The current position of the vehicle can be located through the posture characteristics of the road surface, wherein the roll angle and pitch angle of the vehicle can be called the posture angle of the vehicle. By monitoring the posture angle of the vehicle, the tilt direction and angle of the vehicle can be determined, thereby the tilt degree and posture of the road surface can be seen.

[0060] Among them, during the driving process, the vehicle can collect the roll angle and pitch angle through the relevant hardware equipment. The hardware equipment can be an inertial measurement unit, which is a device for measuring the three-axis attitude angle (or angular rate) and acceleration of an object. Generally, an inertial measurement unit contains three single-axis accelerometers and three single-axis gyroscopes. The accelerometer detects the acceleration signal of the object in the independent three-axis of the carrier coordinate system, and the gyroscope detects the angular velocity signal of the carrier relative to the navigation coordinate system, measures the angular velocity and acceleration of the object in three-dimensional space, and uses this to calculate the attitude of the object.

[0061] The roll angle is the angle of rotation of the vehicle around the longitudinal axis, that is, the degree to which the vehicle tilts left or right. When the vehicle tilts to the right, the roll angle is positive; when the vehicle tilts to the left, the roll angle is negative.

[0062] The pitch angle is the angle of rotation of the vehicle around the horizontal axis, that is, the degree to which the vehicle tilts forward and backward. When the vehicle tilts forward, the pitch angle is positive; when the vehicle tilts backward, the pitch angle is negative.

[0063] Step 204: input the roll angle and the pitch angle into a ground attitude map model, extract features of the roll angle and the pitch angle through the ground attitude map model, and output corresponding position coordinates.

[0064] Among them, the ground posture map model is a neural network model that predicts the current position of the vehicle based on the posture of the road surface. The neural network model can be but is not limited to a long short-term memory network LSTM and a convolutional neural network CNN.

[0065] Among them, the ground posture map model is trained through a large number of training samples (including roll angle and pitch angle), and the model parameters are continuously adjusted until the convergence condition is reached to obtain a trained ground posture map model. That is, the trained ground posture map model can be used for predictive positioning, which can achieve high-precision positioning, and use the roll angle and pitch angle generated by the vehicle during driving to predict and locate the current position of the vehicle.

[0066] Specifically, a trained ground posture map model is obtained, and the roll angle and pitch angle generated by the vehicle during driving are input into the trained ground posture map model. The roll angle and pitch angle are extracted through the ground posture map model, and the corresponding position coordinates are output.

[0067] Step 206, determining the position corresponding to the vehicle according to the position coordinates.

[0068] Specifically, after obtaining the position coordinates output by the ground posture map model, the current position of the vehicle is located through the position coordinates. Therefore, the trained ground posture map model can be used for predictive positioning, which can achieve high-precision positioning, and use the roll angle and pitch angle generated by the vehicle during driving to predict and locate the current position of the vehicle.

[0069] In the above vehicle positioning method, the roll angle and pitch angle corresponding to the vehicle are obtained; the roll angle and pitch angle are input into the ground posture map model, and the roll angle and pitch angle are extracted by the ground posture map model to output the corresponding position coordinates; the corresponding position of the vehicle is determined according to the position coordinates. Therefore, the roll angle and pitch angle generated by the vehicle during driving can be used to predict and locate the current position of the vehicle, thereby improving the accuracy of the vehicle's position.

[0070] In one embodiment, Figure 3 As shown in FIG. 1 , the ground posture map model includes a long short-term memory network and a convolutional neural network. The roll angle and the pitch angle are input into the ground posture map model. The ground posture map model is used to extract the features of the roll angle and the pitch angle, and the corresponding position coordinates are output, including:

[0071] Step 302: input the roll angle and the pitch angle into a long short-term memory network, extract features of the roll angle and the pitch angle through the long short-term memory network, and output a first feature sequence.

[0072] Step 304: input the roll angle and the pitch angle into a convolutional neural network, perform feature extraction on the roll angle and the pitch angle through the convolutional neural network, and output a second feature sequence.

[0073] Step 306: fuse the first feature sequence and the second feature sequence, and output corresponding position coordinates.

[0074] Among them, the ground posture map model includes a long short-term memory network and a convolutional neural network. The long short-term memory network is a network used to perform feature processing on long-distance roll and pitch angles in the ground posture map model, while the convolutional neural network is a network used to perform feature processing on short-distance roll and pitch angles in the ground posture map model.

[0075] Specifically, after obtaining the corresponding roll angle and pitch angle of the vehicle, the roll angle and pitch angle are input into the long short-term memory network and the convolutional neural network respectively. The long short-term memory network extracts features of the roll angle and the pitch angle to obtain the corresponding first feature sequence.

[0076] At the same time, the convolutional neural network extracts features of the roll angle and pitch angle to obtain the corresponding second feature sequence. Finally, the first feature sequence and the second feature sequence are fused to output the corresponding position coordinates.

[0077] In one embodiment, in order to improve the accuracy of position positioning, the input data of the ground posture map model may include the heading angle, the turning angles of the front two wheels, the angular velocity and the acceleration during the vehicle's driving process in addition to the roll angle and pitch angle generated during the vehicle's driving process. The angular velocity and acceleration include the speed on the horizontal axis x, the longitudinal axis y and the vertical axis z. These data are input into the ground posture map model, and feature extraction is performed through the ground posture map model to output the corresponding position coordinates.

[0078] Among them, the ground posture map model includes a long short-term memory network LSTM and a convolutional neural network CNN. The feature dimension of the long short-term memory network LSTM input in the trained ground posture map model is 6 dimensions (including sampling interval, roll angle, pitch angle, heading angle, speed, and the direction angles of the first two wheels). The training adopts a 3-layer LSTM. The first layer consists of 128 hidden units, the second layer consists of 64 hidden units, the third layer consists of 32 hidden units, and the fourth layer is the attention mechanism layer. Finally, it is output by the fully connected layer, where the output dimension of the fully connected layer is 8, and the activation function adopts the ReLU function.

[0079] The feature dimension of the convolutional neural network CNN input is 13 dimensions (including sequence number, sampling interval, roll angle, pitch angle, heading angle, speed, the turning angles of the first two wheels, angular velocity x, angular velocity y, angular velocity z, acceleration x, acceleration y, acceleration z). CNN consists of 3 layers of 1D convolutional neural networks. The number of convolutional filters in the first layer is 128, the convolution kernel size is 3, and the activation function is "ReLU". Then it passes through a one-dimensional maximum pooling layer, the pooling window length is 2, and the pooling window moving step is 2. The number of convolutional filters in the second layer is 64, the convolution kernel size is 3, the activation function is "ReLU", and then it passes through a one-dimensional maximum pooling layer, with the same parameters as above. The number of convolutional filters in the third layer is 32, and other parameters are the same as above. It passes through a one-dimensional maximum pooling layer and then a layer of attention mechanism.

[0080] Finally, the output layers of LSTM and CNN are merged, and then pass through the fully connected layer and the output layer, and the output layer outputs in 2 dimensions.

[0081] In one embodiment, Figure 4 As shown, the training steps of the ground posture map model include:

[0082] Step 402, obtaining a training sample set, where the training sample set includes at least one training sample, the training sample includes a training roll angle and a training pitch angle, and the training sample carries corresponding real position coordinates;

[0083] Step 404: input the training sample set into the untrained ground posture map model, extract features from the training samples through the untrained ground posture map model, and output corresponding training position coordinates.

[0084] Step 406, calculating the loss value based on the training position coordinates and the corresponding real position coordinates.

[0085] Step 408, determining whether the untrained ground posture map model has reached a convergence condition based on the loss value.

[0086] Step 410: When the convergence condition is reached, a trained ground posture map model is obtained.

[0087] Among them, the ground posture map model needs to be trained through a large amount of training sample data. By comparing the real label with the output label, the loss value obtained by comparison is used to determine whether the ground posture map model has reached the convergence condition.

[0088] Specifically, the roll angle and pitch angle are collected by the inertial measurement unit of the vehicle as training samples. A large number of training samples form a training sample set, which includes training roll angles and training pitch angles. Each training sample is associated with a corresponding real position coordinate.

[0089] Furthermore, the training sample set is input into the untrained ground posture map model, and the ground posture map model extracts features from the training samples in the training sample set, outputs the corresponding training position coordinates, and calculates the loss value based on the training position coordinates and the corresponding real position coordinates. The loss value can be calculated using an adaptive learning rate algorithm (Adam: Adaptive Moment Estimation) and a Huber loss function.

[0090] Finally, it is determined whether the ground posture map model has reached the convergence condition based on the loss value. The convergence condition can be that when the loss value reaches a preset threshold, it is determined that the convergence condition has been reached. The preset threshold can be determined based on actual business needs, actual product needs, or actual application scenarios. It can also be that when a preset number of training times is reached, it is determined that the convergence condition has been reached. The preset number of training times can be determined based on actual business needs, actual product needs, or actual application scenarios. Among them, when the loss value reaches the convergence condition, a trained ground posture map model is obtained.

[0091] In one embodiment, Figure 5 As shown, obtaining a training sample set includes:

[0092] Step 502, determining a plurality of different sampling intervals and corresponding sequence lengths;

[0093] Step 504: collect corresponding training roll angles and training pitch angles according to the sampling interval and sequence length to obtain a training sample set.

[0094] Among them, the sampling interval is the distance interval between two samples, and the sequence length is the number of training samples. Multiple different sampling intervals and corresponding sequence lengths are determined, and they can be divided into long distance and short distance according to the sampling intervals and sequence lengths. Specifically, multiple different sampling intervals and corresponding sequence lengths can be determined according to actual business needs, actual product needs or actual application scenarios.

[0095] Furthermore, the corresponding training roll angle and training pitch angle are collected according to the sampling interval and sequence length to obtain a training sample set. For example, one is that the sequence length is 200 and the samples are collected every 1.5 meters, and the other is that the sequence length is 100 and the samples are collected every 0.5 meters.

[0096] In one embodiment, Figure 6 As shown, the ground posture map model includes a long short-term memory network and a convolutional neural network. The training sample set is input into the untrained ground posture map model, and the untrained ground posture map model is used to extract features of the training samples and output the corresponding training position coordinates, including:

[0097] Step 602: Divide the training sample set into a long-distance training sample set and a short-distance training sample set according to different sequence lengths and sampling intervals.

[0098] Step 604: input the long-distance training sample set into the long short-term memory network, extract features from the long-distance training sample set through the long short-term memory network, and obtain a long-distance feature sequence.

[0099] Step 606: input the short-distance training sample set into the convolutional neural network, perform feature extraction on the short-distance training sample set through the convolutional neural network, and obtain a short-distance feature sequence.

[0100] Step 608: Generate corresponding training position coordinates according to the long-distance feature sequence and the short-distance feature sequence.

[0101] Among them, the ground posture map model includes a long short-term memory network and a convolutional neural network. The long short-term memory network is a network used to perform feature processing on long-distance roll and pitch angles in the ground posture map model, while the convolutional neural network is a network used to perform feature processing on short-distance roll and pitch angles in the ground posture map model.

[0102] Specifically, after the corresponding training roll angle and training pitch angle are collected according to the sampling interval and sequence length to obtain the training sample set, the training sample set can be divided according to different sampling intervals and sequence lengths to obtain a long-distance training sample set and a short-distance training sample set.

[0103] Furthermore, the long-distance training sample set is input into the long short-term memory network, and the features of the training samples in the long-distance training sample set are extracted through the long short-term memory network to obtain the corresponding long-distance feature sequence. At the same time, the short-distance training sample set is input into the convolutional neural network, and the features of the short-distance training sample set are extracted through the convolutional neural network to obtain the corresponding short-distance feature sequence.

[0104] Finally, the feature sequences output by the long short-term memory network and the convolutional neural network are fused, that is, the long-distance feature sequence and the short-distance feature sequence are fused to obtain the corresponding training position coordinates.

[0105] In one embodiment, Figure 7 As shown, the above method also includes:

[0106] Step 702, when the convergence condition is not reached, the model parameters of the untrained ground posture map model are continuously adjusted, and the process returns to the step of obtaining the training sample set.

[0107] Step 704, until the convergence condition is reached, a trained ground posture map model is obtained.

[0108] Specifically, it is determined whether the ground posture map model has reached the convergence condition based on the loss value. The convergence condition may be that when the loss value reaches a preset threshold, it is determined that the convergence condition has been reached. The preset threshold may be determined based on actual business needs, actual product needs, or actual application scenarios. It may also be that when a preset number of training times is reached, it is determined that the convergence condition has been reached. The preset number of training times may be determined based on actual business needs, actual product needs, or actual application scenarios. When the loss value reaches the convergence condition, a trained ground posture map model is obtained.

[0109] Furthermore, when the convergence condition is not reached, the model parameters of the untrained ground posture map model are continuously adjusted, and the step of obtaining the training sample set is returned until the convergence condition is reached according to the loss value, and a trained ground posture map model is obtained.

[0110] In one embodiment, the training samples in the training sample set further include heading angle, turning angles of the front two wheels, vehicle speed, angular velocity and acceleration.

[0111] In one embodiment, in order to improve the accuracy of position positioning, the training input data of the ground posture map model may include, in addition to the roll angle and pitch angle generated during vehicle driving, the heading angle, the turning angles of the front two wheels, the angular velocity and the acceleration during vehicle driving. The angular velocity and acceleration include components on the horizontal axis x, the longitudinal axis y and the vertical axis z. Inputting these data into the ground posture map model for training can improve the prediction accuracy of the ground posture map model.

[0112] It should be understood that, although the various steps in the above-mentioned flow chart are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps in the above-mentioned flow chart may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these sub-steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0113] In one embodiment, Figure 8 As shown, a vehicle position positioning device 800 is provided, comprising: an acquisition module 802, a prediction module 804 and a determination module 806, wherein:

[0114] The acquisition module 802 is used to acquire the roll angle and pitch angle corresponding to the vehicle.

[0115] The prediction module 804 is used to input the roll angle and the pitch angle into the ground posture map model, extract features of the roll angle and the pitch angle through the ground posture map model, and output corresponding position coordinates.

[0116] The determination module 806 is used to determine the position corresponding to the vehicle according to the position coordinates.

[0117] In one embodiment, the ground posture map model includes a long short-term memory network and a convolutional neural network. The prediction module 804 inputs the roll angle and the pitch angle into the long short-term memory network, extracts features of the roll angle and the pitch angle through the long short-term memory network, and outputs a first feature sequence. The roll angle and the pitch angle are input into the convolutional neural network, extracts features of the roll angle and the pitch angle through the convolutional neural network, and outputs a second feature sequence. The first feature sequence and the second feature sequence are fused to output the corresponding position coordinates.

[0118] In one embodiment, the vehicle positioning device 800 obtains a training sample set, the training sample set includes at least one training sample, the training sample includes a training roll angle and a training pitch angle, the training sample carries the corresponding real position coordinates, the training sample set is input into an untrained ground posture map model, features are extracted from the training samples through the untrained ground posture map model, and the corresponding training position coordinates are output, a loss value is calculated based on the training position coordinates and the corresponding real position coordinates, and whether the untrained ground posture map model meets the convergence condition is determined based on the loss value. When the convergence condition is met, a trained ground posture map model is obtained.

[0119] In one embodiment, the vehicle position positioning device 800 determines a plurality of different sampling intervals and corresponding sequence lengths, collects corresponding training roll angles and training pitch angles according to the sampling intervals and sequence lengths, and obtains a training sample set.

[0120] In one embodiment, the ground posture map model includes a long short-term memory network and a convolutional neural network. The vehicle position positioning device 800 divides the training sample set into a long-distance training sample set and a short-distance training sample set according to different sequence lengths and sampling intervals, inputs the long-distance training sample set into the long short-term memory network, extracts features from the long-distance training sample set through the long short-term memory network to obtain a long-distance feature sequence, inputs the short-distance training sample set into the convolutional neural network, extracts features from the short-distance training sample set through the convolutional neural network to obtain a short-distance feature sequence, and generates corresponding training position coordinates based on the long-distance feature sequence and the short-distance feature sequence.

[0121] In one embodiment, when the convergence condition is not met, the vehicle position positioning device 800 continuously adjusts the model parameters of the untrained ground posture map model, returns to the step of obtaining the training sample set, and obtains the trained ground posture map model until the convergence condition is met.

[0122] For the specific definition of the vehicle position positioning device, please refer to the definition of the vehicle position positioning method above, which will not be repeated here. Each module in the above-mentioned vehicle position positioning device can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0123] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Fig. 9As shown. The computer device includes a processor, a memory, a network interface and a database connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store a ground posture map model. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a vehicle position positioning method is implemented.

[0124] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Fig.10 As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a vehicle positioning method is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0125] Those skilled in the art will understand that Fig. 9 or Fig.10 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0126] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the following steps are implemented: obtaining a roll angle and a pitch angle corresponding to a vehicle; inputting the roll angle and the pitch angle into a ground posture map model, performing feature extraction on the roll angle and the pitch angle through the ground posture map model, and outputting corresponding position coordinates; and determining a corresponding position of the vehicle according to the position coordinates.

[0127] In one embodiment, the ground posture map model includes a long short-term memory network and a convolutional neural network, and the processor further implements the following steps when executing the computer program: inputting the roll angle and the pitch angle into the long short-term memory network, extracting features of the roll angle and the pitch angle through the long short-term memory network, and outputting a first feature sequence;

[0128] The roll angle and the pitch angle are input into a convolutional neural network, and features of the roll angle and the pitch angle are extracted through the convolutional neural network to output a second feature sequence;

[0129] The first feature sequence and the second feature sequence are fused and the corresponding position coordinates are output.

[0130] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0131] Acquire a training sample set, where the training sample set includes at least one training sample, where the training sample includes a training roll angle and a training pitch angle, and the training sample carries corresponding real position coordinates;

[0132] Input the training sample set into the untrained ground posture map model, extract features of the training samples through the untrained ground posture map model, and output the corresponding training position coordinates;

[0133] The loss value is calculated based on the training position coordinates and the corresponding real position coordinates;

[0134] Determine whether the untrained ground posture map model has reached the convergence condition based on the loss value;

[0135] When the convergence condition is reached, the trained ground posture map model is obtained.

[0136] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0137] Determine a plurality of different sampling intervals and corresponding sequence lengths;

[0138] The corresponding training roll angle and training pitch angle are collected according to the sampling interval and sequence length to obtain a training sample set.

[0139] In one embodiment, the ground posture map model includes a long short-term memory network and a convolutional neural network, and the processor further implements the following steps when executing the computer program:

[0140] The training sample set is divided into a long-distance training sample set and a short-distance training sample set according to different sequence lengths and sampling intervals;

[0141] Input the long-distance training sample set into the long short-term memory network, extract features from the long-distance training sample set through the long short-term memory network, and obtain a long-distance feature sequence;

[0142] The short-distance training sample set is input into the convolutional neural network, and the convolutional neural network is used to extract features of the short-distance training sample set to obtain a short-distance feature sequence;

[0143] The corresponding training position coordinates are generated according to the long-distance feature sequence and the short-distance feature sequence.

[0144] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0145] When the convergence condition is not reached, the model parameters of the untrained ground posture map model are continuously adjusted, and the step of obtaining the training sample set is returned;

[0146] Until the convergence condition is reached, a trained ground posture map model is obtained. In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented: obtaining the roll angle and pitch angle corresponding to the vehicle; inputting the roll angle and pitch angle into the ground posture map model, extracting features of the roll angle and pitch angle through the ground posture map model, and outputting corresponding position coordinates; determining the corresponding position of the vehicle according to the position coordinates.

[0147] In one embodiment, the ground posture map model includes a long short-term memory network and a convolutional neural network, and the processor further implements the following steps when executing the computer program: inputting the roll angle and the pitch angle into the long short-term memory network, extracting features of the roll angle and the pitch angle through the long short-term memory network, and outputting a first feature sequence;

[0148] The roll angle and the pitch angle are input into a convolutional neural network, and features of the roll angle and the pitch angle are extracted through the convolutional neural network to output a second feature sequence;

[0149] The first feature sequence and the second feature sequence are fused and the corresponding position coordinates are output.

[0150] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0151] Acquire a training sample set, where the training sample set includes at least one training sample, where the training sample includes a training roll angle and a training pitch angle, and the training sample carries corresponding real position coordinates;

[0152] Input the training sample set into the untrained ground posture map model, extract features of the training samples through the untrained ground posture map model, and output the corresponding training position coordinates;

[0153] The loss value is calculated based on the training position coordinates and the corresponding real position coordinates;

[0154] Determine whether the untrained ground posture map model has reached the convergence condition based on the loss value;

[0155] When the convergence condition is reached, the trained ground posture map model is obtained.

[0156] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0157] Determine a plurality of different sampling intervals and corresponding sequence lengths;

[0158] The corresponding training roll angle and training pitch angle are collected according to the sampling interval and sequence length to obtain a training sample set.

[0159] In one embodiment, the ground posture map model includes a long short-term memory network and a convolutional neural network, and the processor further implements the following steps when executing the computer program:

[0160] The training sample set is divided into a long-distance training sample set and a short-distance training sample set according to different sequence lengths and sampling intervals;

[0161] Input the long-distance training sample set into the long short-term memory network, extract features from the long-distance training sample set through the long short-term memory network, and obtain a long-distance feature sequence;

[0162] The short-distance training sample set is input into the convolutional neural network, and the convolutional neural network is used to extract features of the short-distance training sample set to obtain a short-distance feature sequence;

[0163] The corresponding training position coordinates are generated according to the long-distance feature sequence and the short-distance feature sequence.

[0164] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0165] When the convergence condition is not reached, the model parameters of the untrained ground posture map model are continuously adjusted, and the step of obtaining the training sample set is returned;

[0166] Until the convergence condition is reached, a trained ground posture map model is obtained.

[0167] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0168] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0169] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A vehicle positioning method, the method comprising: Get the roll angle and pitch angle corresponding to the vehicle; Inputting the roll angle and the pitch angle into a ground posture map model, performing feature extraction on the roll angle and the pitch angle through the ground posture map model, and outputting corresponding position coordinates; Determine the position corresponding to the vehicle according to the position coordinates; The ground posture map model includes a long short-term memory network and a convolutional neural network, and the roll angle and the pitch angle are input into the ground posture map model, and the roll angle and the pitch angle are extracted by the ground posture map model to output corresponding position coordinates, including: Inputting the roll angle and the pitch angle into the long short-term memory network, performing feature extraction on the roll angle and the pitch angle through the long short-term memory network, and outputting a first feature sequence; Inputting the roll angle and the pitch angle into the convolutional neural network, performing feature extraction on the roll angle and the pitch angle through the convolutional neural network, and outputting a second feature sequence; Merging the first feature sequence and the second feature sequence, and outputting corresponding position coordinates; The training steps of the ground posture map model include: Acquire a training sample set, wherein the training sample set includes at least one training sample, the training sample includes a training roll angle and a training pitch angle, and the training sample carries corresponding real position coordinates; Inputting the training sample set into an untrained ground posture map model, performing feature extraction on the training samples through the untrained ground posture map model, and outputting corresponding training position coordinates; Calculate the loss value based on the training position coordinates and the corresponding real position coordinates; Determining whether the untrained ground posture map model reaches a convergence condition according to the loss value; When the convergence condition is reached, a trained ground posture map model is obtained; Wherein, obtaining a training sample set comprises: Determine a plurality of different sampling intervals and corresponding sequence lengths; Collecting corresponding training roll angles and training pitch angles according to the sampling interval and the sequence length to obtain the training sample set; The ground posture map model includes a long short-term memory network and a convolutional neural network, and the training sample set is input into an untrained ground posture map model, and features of the training samples are extracted by the untrained ground posture map model to output corresponding training position coordinates, including: Dividing the training sample set into a long-distance training sample set and a short-distance training sample set according to different sequence lengths and sampling intervals; Inputting the long-distance training sample set into the long short-term memory network, performing feature extraction on the long-distance training sample set through the long short-term memory network to obtain a long-distance feature sequence; Inputting the short-distance training sample set into the convolutional neural network, performing feature extraction on the short-distance training sample set through the convolutional neural network to obtain a short-distance feature sequence; The corresponding training position coordinates are generated according to the long-distance feature sequence and the short-distance feature sequence.

2. The method according to claim 1, characterized in that: The method further comprises: When the convergence condition is not met, continuously adjusting the model parameters of the untrained ground posture map model, and returning to the step of obtaining the training sample set; Until the convergence condition is reached, the trained ground posture map model is obtained.

3. The method according to claim 1 or 2, characterized in that: The training samples in the training sample set also include heading angle, turning angles of the front two wheels, vehicle speed, angular velocity and acceleration.

4. A vehicle position positioning device, characterized in that: The device comprises: An acquisition module is used to acquire the roll angle and pitch angle corresponding to the vehicle; A prediction module, used for inputting the roll angle and the pitch angle into a ground posture map model, performing feature extraction on the roll angle and the pitch angle through the ground posture map model, and outputting corresponding position coordinates; A determination module, used to determine the position corresponding to the vehicle according to the position coordinates; Wherein, the ground posture map model includes a long short-term memory network and a convolutional neural network, the prediction module is further used to input the roll angle and the pitch angle into the long short-term memory network, extract features of the roll angle and the pitch angle through the long short-term memory network, output a first feature sequence, input the roll angle and the pitch angle into the convolutional neural network, extract features of the roll angle and the pitch angle through the convolutional neural network, output a second feature sequence, fuse the first feature sequence and the second feature sequence, and output corresponding position coordinates; The device is further used to obtain a training sample set, the training sample set includes at least one training sample, the training sample includes a training roll angle and a training pitch angle, the training sample carries a corresponding real position coordinate, the training sample set is input into an untrained ground posture map model, features of the training samples are extracted through the untrained ground posture map model, and the corresponding training position coordinates are output, a loss value is calculated based on the training position coordinates and the corresponding real position coordinates, and whether the untrained ground posture map model meets the convergence condition is determined based on the loss value, and when the convergence condition is met, a trained ground posture map model is obtained; The device is further used to determine a plurality of different sampling intervals and corresponding sequence lengths, and collect corresponding training roll angles and training pitch angles according to the sampling intervals and the sequence lengths to obtain the training sample set; The device is also used to divide the training sample set into a long-distance training sample set and a short-distance training sample set according to different sequence lengths and sampling intervals, input the long-distance training sample set into the long short-term memory network, perform feature extraction on the long-distance training sample set through the long short-term memory network to obtain a long-distance feature sequence, input the short-distance training sample set into the convolutional neural network, perform feature extraction on the short-distance training sample set through the convolutional neural network to obtain a short-distance feature sequence, and generate corresponding training position coordinates according to the long-distance feature sequence and the short-distance feature sequence.

5. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 3 are implemented.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 3 are implemented.

Images