Method, device and electronic device for determining steering angle of vehicle

Abstract

The present invention discloses a method, an apparatus, and an electronic device for determining a steering angle of a vehicle. Among them, the method includes: acquiring a lane line image, where the lane line image is an image captured during the driving of the vehicle; inputting the lane line image into a first convolutional module of an angle prediction model to obtain a first feature map; inputting the first feature map into a second convolutional module of the angle prediction model to obtain a second feature map; inputting the second feature map into a result prediction module of the angle prediction model to obtain a target angle value, where the target angle value is the angle value by which the vehicle should steer according to the lane line. The present invention solves the technical problem that the determined steering angle is inaccurate when determining the steering angle of a vehicle in the related art.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving, and in particular, to a method, a device, and an electronic device for determining a steering angle of a vehicle. Background Art

[0002] In current deep intelligent vehicles, since the neural network model needs data for optimization training to predict the angle of vehicle travel. However, in the related art methods, when the scene is complex, it may run randomly due to inconsistent data comparison, such as problems like the light dimming and lane line occlusion, resulting in inaccurate determined angles.

[0003] In view of the above problems, no effective solution has been proposed yet. Summary of the Invention

[0004] Embodiments of the present invention provide a method, a device, and an electronic device for determining a steering angle of a vehicle, so as to at least solve the technical problem that the determined steering angle is inaccurate when determining the steering angle of a vehicle in the related art.

[0005] According to an aspect of an embodiment of the present invention, a method for determining a steering angle of a vehicle is provided, including: obtaining a lane line image, where the lane line image is an image captured during the driving of the vehicle; inputting the lane line image into a first convolution module of an angle prediction model to obtain a first feature map, where the first convolution module includes a first number of first convolution layers, the size of the first convolution layer is a first size, and the angle prediction model is a model trained based on sample data, and the sample data includes a sample lane line image and a sample angle value corresponding to the sample lane line image; inputting the first feature map into a second convolution module of the angle prediction model to obtain a second feature map, where the second convolution module includes a second number of second convolution layers, the size of the second convolution layer is a second size, the second size is smaller than the first size, and the second number is greater than the first number; inputting the second feature map into a result prediction module of the angle prediction model to obtain a target angle value, where the target angle value is the angle value that the vehicle should turn according to the lane line.

[0006] Optionally, before inputting the lane line image into the first convolution module of the angle prediction model to obtain a first feature map, it further includes: obtaining an initial model; determining the execution task of the initial model; determining the first number and the first size based on the execution task, and determining the second number and the second size based on the execution task.

[0007] Optionally, determining the first quantity and the first dimension based on the execution task, and determining the second quantity and the second dimension based on the execution task includes: obtaining the sample data; determining the complexity index and the quantity index of the sample data; determining the first quantity and the first dimension based on the execution task and the complexity index and the quantity index of the sample data, and determining the second quantity and the second dimension based on the execution task and the complexity index and the quantity index of the sample data.

[0008] Optionally, before inputting the lane line image into the first convolutional module of the angle prediction model to obtain the first feature map, it further includes: obtaining initial data; determining the error term in the execution task; transforming the initial data according to the error term to obtain the sample data.

[0009] Optionally, after inputting the lane line image into the first convolutional module of the angle prediction model to obtain the first feature map, it further includes: performing binarization processing on the lane line image to obtain the binarized lane line image.

[0010] Optionally, inputting the first feature map into the second convolutional module of the angle prediction model to obtain the second feature map includes: inputting the first feature map into the first pooling module of the angle prediction model to obtain the first feature map after dimensionality reduction processing; inputting the first feature map after dimensionality reduction processing into the second convolutional module of the angle prediction model to obtain the second feature map.

[0011] Optionally, after inputting the second feature map into the result prediction module of the angle prediction model to obtain the target angle value, it further includes: comparing the target angle value with a predetermined threshold; in the case where the target angle value is less than the predetermined threshold, sending a steering instruction to the steering module of the vehicle, where the steering instruction carries the target angle value.

[0012] According to one aspect of an embodiment of the present invention, a steering angle determination device for a vehicle is provided, including: an acquisition module configured to acquire a lane line image, where the lane line image is an image captured during the driving of the vehicle; a first determination module configured to input the lane line image into a first convolution module of an angle prediction model to obtain a first feature map, where the first convolution module includes a first number of first convolution layers, the size of the first convolution layer is a first size, and the angle prediction model is a model trained based on sample data, the sample data including sample lane line images and corresponding sample angle values; a second determination module configured to input the first feature map into a second convolution module of the angle prediction model to obtain a second feature map, where the second convolution module includes a second number of second convolution layers, the size of the second convolution layer is a second size, the second size is smaller than the first size, and the second number is greater than the first number; a third determination module configured to input the second feature map into a result prediction module of the angle prediction model to obtain a target angle value, where the target angle value is the angle value by which the vehicle should steer according to the lane line.

[0013] According to one aspect of an embodiment of the present invention, an electronic device is provided, including: a processor; a memory for storing executable instructions of the processor; wherein the processor is configured to execute the instructions to implement the vehicle steering angle determination method described in any one of the above.

[0014] According to one aspect of an embodiment of the present invention, a computer-readable storage medium is provided. When the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the vehicle steering angle determination method described in any one of the above.

[0015] In an embodiment of the present invention, a lane line image is acquired. The lane line image is input into the first convolutional module of the angle prediction model to obtain a first feature map. The first feature map is input into the second convolutional module of the angle prediction model to obtain a second feature map. The second feature map is input into the result prediction module of the angle prediction model to obtain a target angle value. Wherein, the target angle value is the angle value by which the vehicle should turn according to the lane line, thereby being able to guide the driving angle of the vehicle. Since the lane line image is an image captured during the driving process of the vehicle, the finally determined vehicle angle can conform to the guidance of the lane line. Moreover, since the finally determined angle is obtained by the angle prediction model, and the angle prediction model is a model trained based on sample data, the sample data includes sample lane line images and corresponding sample angle values, therefore, features can be better recognized, so that the determined angle is more accurate. Also, because the first convolutional module in the angle prediction model includes a first number of first convolutional layers, the size of the first convolutional layer is a first size, the second convolutional module includes a second number of second convolutional layers, the size of the second convolutional layer is a second size, the second size is smaller than the first size, and the second number is greater than the first number, therefore, the required features can be extracted more specifically, and then prediction can be performed, making the prediction more accurate, thus solving the technical problem that the determined steering angle is inaccurate when determining the steering angle of a vehicle in the related art. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings described herein are used to provide a further understanding of the present invention and form a part of this application. The illustrative embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation to the present invention. In the drawings:

[0017] Figure 1 is a flowchart of a method for determining the steering angle of a vehicle according to an embodiment of the present invention;

[0018] Figure 2 is a structural block diagram of a device for determining the steering angle of a vehicle according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0020] It should be noted that the terms "first", "second", etc. in the description, claims and the above-mentioned drawings of the present invention are used to distinguish similar objects, and do not necessarily have to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present invention described here can be implemented in an order other than those illustrated or described here. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0021] Embodiment 1

[0022] According to an embodiment of the present invention, an embodiment of a method for determining the steering angle of a vehicle is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. And although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that here.

[0023] Figure 1 is a flowchart of a method for determining the steering angle of a vehicle according to an embodiment of the present invention, as Figure 1 shown, the method includes the following steps:

[0024] Step S102, obtain a lane line image, where the lane line image is an image obtained by taking a picture during the driving of the vehicle;

[0025] In step S102 provided in the present application, the lane line image refers to a photo or image of the lane lines on the road, used to check the identification of the lane lines, and enable the vehicle to drive according to the identification and rules of the lane lines. For example, it is not allowed to cross the line and drive in the direction indicated by the line, etc.

[0026] The lane line image can be obtained by a camera device provided in the vehicle body. So as to accurately capture the lane line image corresponding to the current driving position of the vehicle to better guide the vehicle to move forward.

[0027] Step S104, input the lane line image into the first convolution module of the angle prediction model to obtain a first feature map, where the first convolution module includes a first number of first convolution layers, the size of the first convolution layer is the first size, and the angle prediction model is a model trained based on sample data, and the sample data includes sample lane line images and corresponding sample angle values;

[0028] In step S104 provided in this application, it is first explained that the angle value that the vehicle should turn is determined by an angle prediction model. By setting this model, the model can better learn the differences and connections between samples, so as to better predict the angle value that the vehicle should turn. Moreover, when the model predicts, the speed and accuracy of obtaining the angle value will also be greatly improved, enhancing the autonomous driving experience.

[0029] In this step, it is explained that the lane line image is input into the first convolutional module. Among them, there are a first number of first convolutional layers in the first convolutional module, and the size of the first convolutional layer is a first size. Among them, the first size is a larger size, and the first number is a smaller number.

[0030] When the first size is a larger size, for example, it is set as a 5x5 convolutional kernel, so as to be able to capture more extensive low-level features in the spatial domain. Low-level features such as identifying objects, bounding boxes for objects, identifying lane lines, bounding boxes for lane lines, and so on.

[0031] When the first number is a smaller number, for example, it is one, which can reduce the computational complexity.

[0032] Step S106: Input the first feature map into the second convolutional module of the angle prediction model to obtain a second feature map. Among them, the second convolutional module includes a second number of second convolutional layers, and the size of the second convolutional layer is a second size. The second size is smaller than the first size, and the second number is larger than the first number;

[0033] In step S106 provided in this application, it is explained that the first feature map is input into the second convolutional module. Among them, there are a second number of second convolutional layers in the second convolutional module, and the size of the second convolutional layer is a second size. Among them, the second size is a smaller size, and the second number is a larger number.

[0034] When the first size is a smaller size, for example, it is set as a 3x3 convolutional kernel, which helps to more finely abstract high-level features, such as identifying the specific features of objects, identifying the specific features of lane lines, and so on.

[0035] When the first number is a larger number, for example, it is 3, which can extract more refined high-level features in different aspects.

[0036] Step S108: Input the second feature map into the result prediction module of the angle prediction model to obtain a target angle value. Among them, the target angle value is the angle value that the vehicle should turn according to the lane line.

[0037] In step S108 provided in this application, the second feature map is input into the result prediction module of the angle prediction model to obtain a target angle value, achieving the purpose of determining the target angle value.

[0038] Through the above steps S102 - S108, a lane line image is obtained. The lane line image is input into the first convolutional module of the angle prediction model to obtain a first feature map. The first feature map is input into the second convolutional module of the angle prediction model to obtain a second feature map. The second feature map is input into the result prediction module of the angle prediction model to obtain a target angle value. Wherein, the target angle value is the angle value by which the vehicle should turn according to the lane line, thereby being able to guide the driving angle of the vehicle. Since the lane line image is an image captured during the vehicle's driving process, the finally determined vehicle angle can conform to the guidance of the lane line. Moreover, since the finally determined angle is obtained by the angle prediction model, and the angle prediction model is a model trained based on sample data, the sample data includes sample lane line images and corresponding sample angle values, therefore, it can better recognize features, making the determined angle more accurate. Also, because the first convolutional module in the angle prediction model includes a first number of first convolutional layers, the size of the first convolutional layer is a first size, the second convolutional module includes a second number of second convolutional layers, the size of the second convolutional layer is a second size, the second size is smaller than the first size, and the second number is greater than the first number, therefore, it can more specifically extract the required features, and then make a prediction, making the prediction more accurate, thus solving the technical problem in the related art that the determined steering angle is inaccurate when determining the steering angle of a vehicle.

[0039] As an alternative embodiment, before inputting the lane line image into the first convolutional module of the angle prediction model to obtain a first feature map, it further includes: obtaining an initial model; determining the execution task of the initial model; determining the first number and the first size based on the execution task, and determining the second number and the second size based on the execution task.

[0040] In this embodiment, the operations before inputting the lane line image into the first convolutional module of the angle prediction model to obtain a first feature map are described. Before that, it is necessary to obtain an initial model and determine the execution task of the initial model, so as to determine the internal structure of the model based on the execution task. For example, in the case where the execution task is a lane line detection task, the first number and the first size are determined based on the lane line detection task, and the second number and the second size are determined based on the lane line detection task.

[0041] When determining the first number, the second number, the first size, and the second size according to the execution task, it can be based on the difficulty of the execution task. For example, the generalization index and the refinement index of the execution task. In the scenario of lane line detection, since the obtained images are relatively single, mostly lane line images, therefore, it can be determined that its generalization index is low. And there are various lane lines that need to be refined for analysis, so it can be determined that its refinement index is high.

[0042] Therefore, a smaller number of convolution kernels with a larger size can be determined according to a lower generalization index, and a larger number of convolution kernels with a smaller size can be determined according to a higher refinement index, so as to better adapt to the application scenario and achieve better results.

[0043] As an alternative embodiment, determining a first quantity and a first size based on the execution of a task, and determining a second quantity and a second size based on the execution of a task, includes: obtaining sample data; determining a complexity index and a quantity index of the sample data; determining the first quantity and the first size based on the execution of the task and the complexity index and the quantity index of the sample data, and determining the second quantity and the second size based on the execution of the task and the complexity index and the quantity index of the sample data.

[0044] In this embodiment, the steps of determining the first quantity, the second quantity, the first size, and the second size are described. In this embodiment, the sample data is also considered. That is, the complexity level and the quantity of the sample data are considered to comprehensively determine the first quantity, the second quantity, the first size, and the second size in combination with the execution of the task.

[0045] Since the processing speed will be slower when setting multiple layers of convolution kernels, if the sample is too complex and the quantity is too large, it will inevitably increase the burden. Therefore, a more suitable model structure can be constructed according to the complexity index and the quantity index of the sample data, so that the model can also be better trained during training.

[0046] As an alternative embodiment, before inputting the lane line image into the first convolution module of the angle prediction model to obtain a first feature map, it further includes: obtaining initial data; determining an error term in the execution of the task; and transforming the initial data according to the error term to obtain sample data.

[0047] In this embodiment, the steps before inputting the lane line image into the first convolution module of the angle prediction model to obtain a first feature map are described. In this step, it is described that the initial data needs to be obtained first, the error term in the execution of the task is determined, and the initial data is transformed according to the error term to expand the initial data, thereby obtaining the above-mentioned sample data.

[0048] It should be noted that the above error term can be data that interferes with the accuracy of the execution of the task. For example, in the task of detecting lane lines, the goal is to detect lane lines, but the result of determining the lane lines is easily affected by factors such as light.

[0049] At this time, the error term can be light, and transforming the initial data can be to transform data under different light conditions to obtain sample data.

[0050] As described above, the error term can also be an occluder. When the error term is an occluder, the images in the initial data can be occluded at different positions to obtain the above sample data.

[0051] Thus, while the sample data is richer, the error brought by the error term can be more clearly identified, avoiding inaccurate results.

[0052] As an optional embodiment, before inputting the lane line image into the first convolution module of the angle prediction model to obtain the first feature map, it further includes: binarizing the lane line image to obtain the binarized lane line image.

[0053] In this embodiment, the steps before inputting the lane line image into the first convolution module of the angle prediction model to obtain the first feature map are described. Before that, the lane line image can be binarized to obtain the binarized lane line image. By binarizing the image, the data information volume can be reduced. That is, through binarization, the gray information in the image is converted into two colors, black and white, reducing the data information volume of the image, which is convenient for storage and processing. It can also improve the image quality: binarization can make the edges of the image clearer and the details more prominent. It can also facilitate image segmentation. Binarization can separate the target from the background in the image, facilitating image segmentation and target recognition. Moreover, binarization can simplify the image processing algorithm and process, improving the efficiency and accuracy of image processing.

[0054] As an optional embodiment, inputting the first feature map into the second convolution module of the angle prediction model to obtain the second feature map includes: inputting the first feature map into the first pooling module of the angle prediction model to obtain the first feature map after dimensionality reduction; inputting the first feature map after dimensionality reduction into the second convolution module of the angle prediction model to obtain the second feature map.

[0055] In this embodiment, the steps of inputting the first feature map into the second convolution module of the angle prediction model to obtain the second feature map are described. In this step, the first feature map can be first input into the first pooling module of the angle prediction model to obtain the first feature map after dimensionality reduction. Through pooling, the spatial dimension of the feature map is reduced. This helps to reduce the computational complexity while ensuring the retention of important information in the image and improving the perception of the spatial structure.

[0056] It should be noted that a pooling module can be introduced after each convolution module. The setting of the pooling layer can be based on the corresponding convolution module. Since pooling reduces the computational complexity, the calculation can be performed according to the processing difficulty in its corresponding convolution module.

[0057] After that, the first feature map after dimensionality reduction processing is input into the second convolutional module of the angle prediction model to obtain a second feature map, so as to achieve a better effect of feature extraction.

[0058] As an optional embodiment, after inputting the second feature map into the result prediction module of the angle prediction model to obtain the target angle value, it further includes: comparing the target angle value with a predetermined threshold; in the case where the target angle value is less than the predetermined threshold, sending a steering instruction to the steering module of the vehicle, wherein the steering instruction carries the target angle value.

[0059] In this embodiment, the steps after inputting the second feature map into the result prediction module of the angle prediction model to obtain the target angle value are described. After that, the target angle value will be compared with the predetermined threshold to determine the comparison result between the target angle value and the predetermined threshold. The predetermined threshold can be the angle that the vehicle can turn. For example, if the wheel can only turn 90 degrees, then it is compared with 90 degrees. In the case where the target angle value is less than the predetermined threshold, it means that the vehicle can turn, and a steering instruction is sent to the steering module of the vehicle to make the vehicle turn and achieve the steering effect.

[0060] Based on the above embodiments and optional embodiments, an optional implementation manner is provided, which is specifically described below.

[0061] In the related art, due to the high dependence of the deep learning model on training data, when facing complex scenarios such as dimming of light and occlusion of lane lines, the inconsistency of the training data will lead to a decline in the generalization ability of the model. This may cause problems such as random running and misjudgment of the intelligent vehicle during actual inspection. It is difficult to accurately determine the angle value at which the vehicle should turn.

[0062] In view of this, an optional implementation manner of the present invention provides a method for determining the steering angle of a vehicle, which can accurately determine the angle value at which the vehicle needs to turn, and can achieve better results with less data.

[0063] Premise steps:

[0064] S1, obtain an initial model;

[0065] S2, determine the execution task of the initial model;

[0066] S3, obtain initial data;

[0067] S4, determine the error term in the execution task;

[0068] S5, transform the initial data according to the error term to obtain sample data, where the sample data includes a sample lane line image and a sample angle value corresponding to the sample lane line image;

[0069] S6. Determine the complexity index and quantity index of the sample data;

[0070] S7. Based on the task execution and the complexity index and quantity index of the sample data, determine the quantity and size of the first convolutional layer to obtain the first convolutional module, and based on the task execution and the complexity index and quantity index of the sample data, determine the quantity and size of the second convolutional layer to obtain the second convolutional module;

[0071] That is, in the lane inspection task scenario, the convolutional kernel selection strategy can be to use a larger 5x5 convolutional kernel in the initial convolutional layer to capture a wider range of low-level features in the spatial domain. Smaller 3x3 convolutional kernels are used in subsequent convolutional layers. Through multi-layer stacking, it helps to more finely abstract high-level features. That is, larger convolutional kernels are used to capture richer primary features, while in subsequent stages, more attention is paid to using smaller convolutional kernels to more finely describe features.

[0072] S8. Construct a pooling module after the convolutional module;

[0073] That is, dimension adjustment of the pooling layer: A pooling layer is introduced after each convolutional layer to reduce the spatial dimension of the feature map. This helps to reduce computational complexity while ensuring the retention of important information in the image and improving the perception of the spatial structure.

[0074] S9. Finally, set up the fully connected layer;

[0075] That is, the final fully connected layer is designed to learn the global features and more advanced semantic information of the overall image. The introduction of this layer can provide stronger semantic discrimination performance for the abstract representation of specific tasks.

[0076] S10. Based on the above first convolutional module, second convolutional module, pooling module, and fully connected layer, construct the initial model;

[0077] S11. Train the initial model based on the sample data to obtain the angle prediction model.

[0078] It should be noted that in the process of constructing the initial model above, in order to better adapt to the actual application scenario, an advanced parameter adjustment algorithm is used to optimize the parameters of the model. An advanced parameter adjustment algorithm is also introduced. Through adaptive learning and fine adjustment, the model can better adapt to different actual application scenarios. During the training process, combined with the latest progress in deep learning, an effective training strategy is designed to improve the generalization ability and convergence speed of the model.

[0079] It should also be noted that in the above model, in order to improve the generalization ability of the model, regularization techniques are adopted in the network, such as the Dropout technique. The introduction of the regularization technique effectively alleviates the overfitting problem by randomly turning off neurons, making the network more robust to different inputs.

[0080] When more data needs to be collected in the same scenario in related technologies to achieve a relatively accurate effect, the method provided by the optional implementation manner of the present invention can achieve a more accurate effect with fewer data.

[0081] Application steps:

[0082] S1. Obtain a lane line image, where the lane line image is an image captured during the driving of a vehicle;

[0083] For example, receive the picture obtained by the camera, perform binary extraction, and only retain the required part.

[0084] S2. Input the lane line image into the first convolution module of the angle prediction model to obtain a first feature map, where the first convolution module includes a first number of first convolution layers, the size of the first convolution layer is the first size, and the angle prediction model is a model trained based on sample data, and the sample data includes sample lane line images and corresponding sample angle values;

[0085] S3. Input the first feature map into the first pooling module of the angle prediction model to obtain a first feature map after dimensionality reduction processing;

[0086] S4. Input the first feature map after dimensionality reduction processing into the second convolution module of the angle prediction model to obtain a second feature map, where the second convolution module includes a second number of second convolution layers, the size of the second convolution layer is the second size, the second size is smaller than the first size, and the second number is greater than the first number;

[0087] S5. Input the second feature map into the result prediction module of the angle prediction model, and after processing such as pooling and fully connected layers, finally obtain a target angle value, where the target angle value is the angle value that the vehicle should turn according to the lane line;

[0088] S6. Compare the target angle value with a predetermined threshold;

[0089] S7. When the target angle value is less than the predetermined threshold, send a steering instruction to the steering module of the vehicle, where the steering instruction carries the target angle value;

[0090] For example, the angle value can be output to the lower computer control system.

[0091] S8. When the target angle value is greater than a predetermined threshold, send a detection instruction to the detection module and the calibration module, and delete the angle value of this time, so as to detect faults and calibrate the angle value.

[0092] Through the above optional implementation manners, at least the following beneficial effects can be achieved:

[0093] (1) Use an advanced parameter adjustment algorithm to perform adaptive learning and adjustment of model parameters;

[0094] (2) Design a training strategy for specific tasks to improve the generalization ability and convergence speed of the model;

[0095] (3) Effectively reduce the overfitting problem and make the network more robust to different inputs;

[0096] (4) Adopt a larger convolutional kernel to capture richer primary features, while in the subsequent stage, pay more attention to using smaller convolutional kernels to describe features more finely;

[0097] (5) Can achieve more accurate results with fewer data.

[0098] It should be noted that for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that the present invention is not limited by the described action sequence, because according to the present invention, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the present invention.

[0099] Through the description of the above implementation manners, those skilled in the art can clearly understand that the method according to the above embodiments can be implemented by means of software plus a necessary general hardware platform. Of course, it can also be implemented by hardware, but in many cases, the former is a better implementation manner. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes several instructions for causing a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods of the various embodiments of the present invention.

[0100] Embodiment 2

[0101] According to an embodiment of the present invention, there is also provided a device for implementing the method for determining the steering angle of the vehicle as described above. Figure 2 It is a structural block diagram of the device for determining the steering angle of the vehicle according to an embodiment of the present invention, as Figure 2As shown in the figure, the device includes: an acquisition module 202, a first determination module 204, a second determination module 206, and a third determination module 208. The device will be described in detail below.

[0102] The acquisition module 202 is configured to acquire a lane line image, where the lane line image is an image captured during the driving of the vehicle; the first determination module 204 is connected to the above acquisition module 202 and is configured to input the lane line image into the first convolution module of the angle prediction model to obtain a first feature map. The first convolution module includes a first number of first convolution layers, and the size of the first convolution layer is a first size. The angle prediction model is a model trained based on sample data, and the sample data includes sample lane line images and corresponding sample angle values; the second determination module 206 is connected to the above first determination module 204 and is configured to input the first feature map into the second convolution module of the angle prediction model to obtain a second feature map. The second convolution module includes a second number of second convolution layers, the size of the second convolution layer is a second size, the second size is smaller than the first size, and the second number is greater than the first number; the third determination module 208 is connected to the above second determination module 206 and is configured to input the second feature map into the result prediction module of the angle prediction model to obtain a target angle value, where the target angle value is the angle value by which the vehicle should turn according to the lane line.

[0103] It should be noted here that the above acquisition module 202, first determination module 204, second determination module 206, and third determination module 208 correspond to steps S102 to S108 in the method for determining the steering angle of the vehicle. The instances and application scenarios implemented by the multiple modules and the corresponding steps are the same, but are not limited to the content disclosed in the above Embodiment 1.

[0104] Embodiment 3

[0105] According to another aspect of the embodiments of the present invention, an electronic device is further provided, including: a processor; a memory for storing processor-executable instructions, where the processor is configured to execute the instructions to implement the method for determining the steering angle of the vehicle in any of the above items.

[0106] Embodiment 4

[0107] According to another aspect of the embodiments of the present invention, a computer-readable storage medium is further provided. When the instructions in the computer-readable storage medium are executed by the processor of the electronic device, the electronic device can execute the method for determining the steering angle of the vehicle in any of the above items.

[0108] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0109] In the above embodiments of the present invention, the descriptions of the various embodiments each have their own focuses. For parts not detailed in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.

[0110] In the several embodiments provided by the present application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only illustrative. For example, the division of the units can be a logical function division. In actual implementation, there can be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces. The indirect couplings or communication connections of the units or modules can be in electrical or other forms.

[0111] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0112] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.

[0113] If the above-mentioned integrated units are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The foregoing storage medium includes: USB flash drives, read-only memories (ROMs), random access memories (RAMs), mobile hard disks, magnetic disks, or optical disks and other various media that can store program codes.

[0114] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A method for determining the steering angle of a vehicle, characterized in that, Including: Obtain a lane line image, where the lane line image is an image captured during the driving of a vehicle; Input the lane line image into the first convolutional module of an angle prediction model to obtain a first feature map, where the first convolutional module includes a first number of first convolutional layers, the size of the first convolutional layer is a first size, and the angle prediction model is a model trained based on sample data, and the sample data includes sample lane line images and corresponding sample angle values; Input the first feature map into the second convolutional module of the angle prediction model to obtain a second feature map, where the second convolutional module includes a second number of second convolutional layers, the size of the second convolutional layer is a second size, the second size is smaller than the first size, and the second number is greater than the first number; Input the second feature map into the result prediction module of the angle prediction model to obtain a target angle value, where the target angle value is the angle value by which the vehicle should turn when driving along the lane line; Where, before inputting the lane line image into the first convolutional module of the angle prediction model to obtain a first feature map, it further includes: obtaining an initial model; determining the execution task of the initial model; determining the first number and the first size based on the execution task, and determining the second number and the second size based on the execution task; Where, determining the first number and the first size based on the execution task, and determining the second number and the second size based on the execution task includes: obtaining the sample data; determining the complexity index and the quantity index of the sample data; determining the first number and the first size based on the execution task and the complexity index and the quantity index of the sample data, and determining the second number and the second size based on the execution task and the complexity index and the quantity index of the sample data.

2. The method according to claim 1, characterized in that, Before inputting the lane line image into the first convolutional module of the angle prediction model to obtain a first feature map, it further includes: Obtain initial data; Determine the error term in the execution task; Transform the initial data according to the error term to obtain the sample data.

3. The method according to claim 1, characterized in that, Before inputting the lane line image into the first convolutional module of the angle prediction model to obtain a first feature map, it further includes: Perform binarization processing on the lane line image to obtain a binarized lane line image.

4. The method according to claim 1, characterized in that, Inputting the first feature map into the second convolutional module of the angle prediction model to obtain a second feature map includes: Input the first feature map into the first pooling module of the angle prediction model to obtain a first feature map after dimensionality reduction processing; Input the first feature map after dimensionality reduction processing into the second convolutional module of the angle prediction model to obtain the second feature map.

5. The method according to any one of claims 1 to 4, characterized in that, After inputting the second feature map into the result prediction module of the angle prediction model to obtain a target angle value, it further includes: Compare the target angle value with a predetermined threshold; When the target angle value is less than the predetermined threshold, a steering instruction is sent to the steering module of the vehicle, where the target angle value is carried in the steering instruction.

6. A device for determining the steering angle of a vehicle, characterized in that, Comprising: An acquisition module, configured to acquire a lane line image, where the lane line image is an image captured during the driving of the vehicle; A first determination module, configured to input the lane line image into a first convolution module of an angle prediction model to obtain a first feature map, where the first convolution module includes a first number of first convolution layers, the size of the first convolution layer is a first size, and the angle prediction model is a model trained based on sample data, and the sample data includes a sample lane line image and a sample angle value corresponding to the sample lane line image; A second determination module, configured to input the first feature map into a second convolution module of the angle prediction model to obtain a second feature map, where the second convolution module includes a second number of second convolution layers, the size of the second convolution layer is a second size, the second size is smaller than the first size, and the second number is greater than the first number; A third determination module, configured to input the second feature map into a result prediction module of the angle prediction model to obtain a target angle value, where the target angle value is the angle value by which the vehicle should steer according to the lane line; Wherein, before the first determination module is further configured to input the lane line image into the first convolution module of the angle prediction model to obtain a first feature map, it further includes: acquiring an initial model; determining the execution task of the initial model; determining the first number and the first size based on the execution task, and determining the second number and the second size based on the execution task; Wherein, the first determination module is further configured to determine the first number and the first size based on the execution task, and determine the second number and the second size based on the execution task, including: acquiring the sample data; determining the complexity index and the quantity index of the sample data; determining the first number and the first size based on the execution task and the complexity index and the quantity index of the sample data, and determining the second number and the second size based on the execution task and the complexity index and the quantity index of the sample data.

7. An electronic device, characterized in that, Comprising: A processor; A memory for storing instructions executable by the processor; Wherein, the processor is configured to execute the instructions to implement the vehicle steering angle determination method according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, When the instructions in the computer-readable storage medium are executed by the processor of the electronic device, the electronic device is enabled to execute the vehicle steering angle determination method according to any one of claims 1 to 5.

Images