Vehicle driving area detection method and device and storage medium

Abstract

The invention discloses a method and a device for detecting a drivable area of a vehicle and a storage medium. The method comprises the following steps: acquiring a to-be-processed image of a preset area corresponding to a target vehicle; performing feature extraction on the to-be-processed image to obtain distance features corresponding to boundary points and boundary point category features; the distance feature represents a distance between a boundary point of the to-be-processed image and a center point of the to-be-processed image; according to the distance features corresponding to the boundary points, determining predicted positions of the boundary points of the to-be-processed image; determining a prediction category of the boundary point according to the prediction position of the boundary point of the to-be-processed image and the category feature of the boundary point; and determining a drivable area of the target vehicle according to the predicted position of the boundary point of the to-be-processed image and the predicted category of the boundary point. According to the invention, the accuracy of detecting the drivable area by the vehicle is improved, and the safety of the vehicle in the driving process is improved.

Description

Technical Field

[0001] The present application relates to the field of intelligent automatic parking technology, and in particular to a method, device and storage medium for detecting a vehicle drivable area. Background Art

[0002] With the development of autonomous driving technology, automatic parking has also become an indispensable implementation scenario in autonomous driving. The drivable area detection algorithm is a key component of autonomous driving technology, which is crucial to ensure that autonomous vehicles can drive safely and effectively on the road. In the existing technology, the progress in the field of computer vision provides a basis for drivable area detection. Through image processing and analysis technology, the algorithm can extract road features and obstacle information from the visual data captured by the camera. The development of deep learning, especially convolutional neural network (CNN), has greatly promoted the progress of image segmentation technology. Deep learning models can automatically learn feature representations in images, so as to more accurately identify and classify drivable areas. Sensors such as laser radar (LiDAR), radar, GPS and inertial measurement unit (IMU) provide autonomous vehicles with rich environmental perception capabilities. These sensor data can be combined with visual data to provide more accurate and robust drivable area detection. The data of a single sensor may have limitations, and multi-sensor fusion technology can effectively combine information from different sensors to improve the accuracy and reliability of detection.

[0003] However, in daily life, vehicles may face a variety of complex environments when parking automatically, including urban streets, highways, rural roads, etc., and in order to ensure the safe driving of the vehicle, the drivable area detection algorithm must be able to process sensor data and make decisions within milliseconds, which places high demands on the algorithm's computational efficiency; at the same time, the drivable area detection algorithm needs to maintain high accuracy under various weather conditions and lighting changes; in the existing technology, the real-time performance and computational efficiency of the algorithm are poor, and the detection accuracy is not precise enough. Summary of the invention

[0004] The present application provides a method, device and storage medium for detecting a vehicle's drivable area, which can improve the accuracy of the vehicle's detection of the drivable area and enhance safety during driving.

[0005] In one aspect, the present application provides a method for detecting a vehicle drivable area, the method comprising:

[0006] Acquire an image to be processed of a preset area corresponding to a target vehicle; the preset area is an area whose distance from the target vehicle meets a preset distance condition;

[0007] Extracting features of the image to be processed to obtain distance features and boundary point category features corresponding to the boundary points; the distance features represent the distance between the boundary points of the image to be processed and the center point of the image to be processed;

[0008] Determining predicted positions of the boundary points of the image to be processed according to the distance features corresponding to the boundary points;

[0009] Determining the predicted category of the boundary point according to the predicted position of the boundary point of the image to be processed and the category characteristics of the boundary point;

[0010] The drivable area of ​​the target vehicle is determined according to the predicted positions of the boundary points of the image to be processed and the predicted categories of the boundary points.

[0011] In an exemplary embodiment, the extracting features of the image to be processed to obtain distance features and boundary point category features corresponding to the boundary points includes:

[0012] Input the image to be processed into the distance prediction network in the feature extraction model for distance prediction processing, predict the distance between the center point of the image to be processed and the boundary point of the image to be processed at each preset increment along the target direction of the image to be processed, and obtain the target number of distances between the center point of the image to be processed and the boundary point of the image to be processed;

[0013] Determining the distance feature of the target number dimension according to the distance between the center point of the target number of images to be processed and the boundary point of the target number of images to be processed;

[0014] Inputting the image to be processed into the boundary point category recognition network in the feature extraction model to perform boundary point category recognition processing, identifying the category of the boundary points of the image to be processed at intervals of the preset increment along the target direction of the image to be processed, and obtaining the target number of boundary point categories;

[0015] According to the boundary point categories of the target quantity, boundary point category features of the target quantity dimension are determined.

[0016] In an exemplary embodiment, determining the predicted position of the boundary point of the image to be processed according to the distance feature corresponding to the boundary point includes:

[0017] Determine the coordinate information of any target boundary point according to the distance between any target boundary point and the center point of the image to be processed; the target boundary point is any boundary point among the target number of boundary points;

[0018] Determining the predicted position of the target boundary point according to the coordinate information of the target boundary point;

[0019] The predicted position of each boundary point in the image to be processed is determined according to the predicted position of the target boundary point.

[0020] In an exemplary embodiment, determining the predicted category of the boundary point according to the predicted position of the boundary point of the image to be processed and the category feature of the boundary point includes:

[0021] Determining, according to the boundary point category characteristics, the position category corresponding to the boundary point of the image to be processed;

[0022] Determining a target prediction category of the target boundary point according to any position category corresponding to the target boundary point and the predicted position of the target boundary point;

[0023] According to the target prediction category of the target boundary point, the prediction category of each boundary point in the image to be processed is determined.

[0024] In an exemplary embodiment, determining the drivable area of ​​the target vehicle according to the predicted position of the boundary point of the image to be processed and the predicted category of the boundary point includes:

[0025] Connecting the boundary points of the image to be processed according to the predicted positions of the boundary points of the image to be processed to obtain a closed area;

[0026] Determining the boundary points of the drivable area of ​​the image to be processed and the boundary points of the prohibited drivable area of ​​the image to be processed according to the predicted categories of the boundary points;

[0027] The drivable area of ​​the target vehicle is determined according to the drivable area boundary points, the prohibited drivable area boundary points and the closed area.

[0028] In an exemplary embodiment, the method for training the feature extraction model includes:

[0029] Acquire sample overhead view image data; the sample overhead view image data is annotated with a sample distance label and a sample boundary point category label;

[0030] Based on a feature extraction network of a preset model, feature extraction processing is performed on the sample overhead view image data to obtain sample distance features and sample boundary point category features corresponding to the sample overhead view image data;

[0031] Performing distance prediction processing on the sample distance feature based on the distance prediction network to obtain a sample distance result of the sample overhead view image data;

[0032] Performing boundary point category recognition processing on the sample boundary point category features based on the boundary point category recognition network to obtain a sample boundary point category result of the overhead image data;

[0033] The preset model is trained according to a first difference between the sample distance result and the sample distance label, and a second difference between the sample boundary point category result and the sample boundary point category label, and the feature extraction network in the preset model after the training is used as the feature extraction model.

[0034] In an exemplary embodiment, the method of training the preset model according to the first difference between the sample distance result and the sample distance label, and the second difference between the sample boundary point category result and the sample boundary point category label, and using the feature extraction network in the preset model after the training as the feature extraction model includes:

[0035] determining first loss information according to a first difference between the sample distance result and the sample distance label;

[0036] Determining second loss information according to a second difference between the sample boundary point classification result and the sample boundary point classification label;

[0037] Determining target loss information according to the first loss information and the second loss information;

[0038] According to the target loss information, the parameters of the preset model are adjusted, and the feature extraction network in the preset model after training is used as the feature extraction model.

[0039] Another aspect provides a device for detecting a vehicle drivable area, the device comprising:

[0040] The image module to be processed is used to obtain an image to be processed of a preset area corresponding to the target vehicle; the preset area is an area whose distance from the target vehicle meets a preset distance condition;

[0041] A feature extraction module is used to extract features from the image to be processed to obtain distance features and boundary point category features corresponding to the boundary points; the distance features represent the distance between the boundary points of the image to be processed and the center point of the image to be processed;

[0042] A prediction position module, used to determine the predicted position of the boundary point of the image to be processed according to the distance feature corresponding to the boundary point;

[0043] A prediction category module, used to determine the prediction category of the boundary point according to the predicted position of the boundary point of the image to be processed and the category characteristics of the boundary point;

[0044] The drivable area module is used to determine the drivable area of ​​the target vehicle according to the predicted positions of the boundary points of the image to be processed and the predicted categories of the boundary points.

[0045] On the other hand, an electronic device is provided, comprising a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or at least one program is loaded by the processor and executes the method for detecting a vehicle drivable area as described above.

[0046] On the other hand, a computer-readable storage medium is provided, in which at least one instruction or at least one program is stored, and the at least one instruction or at least one program is loaded and executed by a processor to implement the vehicle drivable area detection method as described above.

[0047] On the other hand, a computer program product or a computer program is provided, the computer program product or the computer program includes computer instructions, the computer instructions are stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes to implement the above-mentioned vehicle drivable area detection method.

[0048] The present application provides a method, device and storage medium for detecting a vehicle drivable area, which have the following technical effects:

[0049] The present application obtains an image to be processed of a preset area corresponding to a target vehicle; the preset area is an area whose distance from the target vehicle meets a preset distance condition; performs feature extraction on the image to be processed to obtain distance features and boundary point category features corresponding to the boundary points; the distance features characterize the distance between the boundary points of the image to be processed and the center point of the image to be processed; according to the distance features corresponding to the boundary points, the predicted positions of the boundary points of the image to be processed are determined; according to the predicted positions of the boundary points of the image to be processed and the boundary point category features, the predicted categories of the boundary points are determined; according to the predicted positions of the boundary points of the image to be processed and the predicted categories of the boundary points, the drivable area of ​​the target vehicle is determined. The present application obtains an image to be processed corresponding to a target vehicle, performs feature extraction on the image to be processed, obtains distance features and boundary point category features of the image to be processed, and determines the predicted positions and boundary point categories of the boundary points of the image to be processed respectively through these two features, thereby determining the drivable area, improving the accuracy of the prediction of the drivable area, improving the accuracy and reliability of the vehicle detection of the drivable area, and improving the safety of the vehicle during driving. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions and advantages of the embodiments of this specification or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0051] Figure 1 It is a flow chart of a method for detecting a vehicle drivable area provided in an embodiment of this specification;

[0052] Figure 2 It is a flow chart of a method for detecting a vehicle drivable area provided in an embodiment of this specification;

[0053] Figure 3 It is a flow chart of a method for detecting a vehicle drivable area provided in an embodiment of this specification;

[0054] Figure 4 is a structural diagram of a feature extraction network provided in an embodiment of this specification;

[0055] Figure 5 It is a flow chart of a method for detecting a vehicle drivable area provided in an embodiment of this specification;

[0056] Figure 6 It is a flow chart of a method for detecting a vehicle drivable area provided in an embodiment of this specification;

[0057] Figure 7 is a schematic diagram of a boundary point prediction provided by an embodiment of this specification;

[0058] Figure 8 It is a flow chart of a method for detecting a vehicle drivable area provided in an embodiment of this specification;

[0059] Fig. 9 It is a flow chart of a method for detecting a vehicle drivable area provided in an embodiment of this specification;

[0060] Fig.10 is a schematic diagram of a vehicle drivable area provided in an embodiment of this specification;

[0061] Fig.11 is a schematic diagram of a vehicle drivable area provided in an embodiment of this specification;

[0062] Fig.12 It is a structural schematic diagram of a vehicle drivable area detection device provided in an embodiment of this specification.

[0063] Fig.13It is a structural diagram of a server of a method for detecting a vehicle drivable area provided in an embodiment of this specification. DETAILED DESCRIPTION

[0064] The following will be combined with the drawings in the embodiments of this specification to clearly and completely describe the technical solutions in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0065] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0066] The following describes a method for detecting a vehicle drivable area of ​​the present application. Figure 1 It is a flow chart of a method for detecting a vehicle drivable area provided in an embodiment of this specification. This specification provides method operation steps as described in the embodiment or flow chart, but may include more or fewer operation steps based on conventional or non-creative labor. The order of steps listed in the embodiment is only one way of executing the steps among many orders, and does not represent the only order of execution. When the actual system or server product is executed, it can be executed in sequence or in parallel (for example, in a parallel processor or multi-threaded processing environment) according to the method shown in the embodiment or the accompanying drawings. Specifically, Figure 1 As shown, the method can be applied to a control unit in a vehicle, and the method comprises:

[0067] S101: Acquire an image to be processed of a preset area corresponding to a target vehicle; the preset area is an area whose distance from the target vehicle meets a preset distance condition.

[0068] In the embodiments of the present specification, the image to be processed may be a bird's-eye view image obtained by performing IPM (Inverse Perspective Mapping) processing on an image collected by a camera, a camera or other image acquisition device on a target vehicle. In the image collected by the image acquisition device, due to the existence of the perspective effect, the originally parallel objects do not appear parallel in the image, and the idea of ​​IPM is to convert the original perspective result to the field of view of a bird's-eye view, eliminate perspective distortion, and make the image look like it is observed vertically from above. Among them, the preset distance condition can be centered on the target vehicle, and the distance value to the target vehicle is less than or equal to the distance value a, and the area that meets this condition is the preset area. The perspective image is converted into a bird's-eye view through the IPM transformation, so that the lane lines, road edges and other ground features are more linear, which simplifies the subsequent feature extraction process, and IPM helps the system to understand the relative position and distance of objects more accurately, which is conducive to vehicle obstacle avoidance.

[0069] S103: extracting features of the image to be processed to obtain distance features and boundary point category features corresponding to boundary points; the distance features represent the distance between the boundary points of the image to be processed and the center point of the image to be processed.

[0070] In the embodiments of the present specification, feature extraction is performed on the image to be processed, and distance features and boundary point category features are extracted, thereby facilitating the subsequent determination of the drivable area and improving the accuracy of vehicle detection of the drivable area.

[0071] In the embodiment of this specification, the feature extraction is performed on the image to be processed to obtain the distance feature and the boundary point category feature corresponding to the boundary point, such as Figure 2 As shown, Figure 2 The following is a flowchart of a method for detecting a vehicle drivable area, including:

[0072] S201: Input the image to be processed into the distance prediction network in the feature extraction model for distance prediction processing, predict the distance between the center point of the image to be processed and the boundary points of the image to be processed at preset increments along the target direction of the image to be processed, and obtain the target number of distances between the center point of the image to be processed and the boundary points of the image to be processed.

[0073] In the embodiment of this specification, for example, the target direction may be a clockwise direction starting from the 12 o'clock direction of the top view of the image to be processed; the preset increment may be 1 degree; the distance between the center point of the image to be processed and the boundary point of the image to be processed is predicted at each preset increment along the target direction of the image to be processed, that is, along the 12 o'clock direction of the top view of the image to be processed, a boundary point is predicted at each 1 degree interval, and the predicted result is the distance between the center point of the image to be processed and the current boundary point of the image to be processed. Since the image has a total of 360 degrees, a total of 360 boundary points can be predicted, and the target number can be 360.

[0074] In the embodiments of this specification, the training method of the feature extraction model is as follows: Figure 3 As shown, Figure 3 The following is a flowchart of a method for detecting a vehicle drivable area, including:

[0075] S301: Acquire sample overhead view image data; the sample overhead view image data is annotated with sample distance labels and sample boundary point category labels.

[0076] In the embodiments of this specification, Figure 4 As shown, Figure 4 It is a structural diagram of a feature extraction network. Figure 4 401 is an input image of size 1024*1024*3, 402 is a feature extraction network using Regnet as the backbone network, and 403 is two sets of 360-dimensional feature vectors output by the last layer of the feature extraction network. One set represents the distance of the drivable area points, and the other represents the category of the drivable area points. Different category points can be represented by different color channels. For example, the drivable area can be represented by blue area points, the obstacle can be represented by red points as a car, and the obstacle can be represented by green area points as a wall. Specifically, the sample overhead view image data is the input data, and the size of the sample overhead view image data is 1024*1024*3, that is, 1024x1024 pixels, and the sample overhead view image data has 3 color channels (RGB). This means that each pixel is composed of three colors: red, green, and blue.

[0077] S302: Based on a feature extraction network of a preset model, feature extraction processing is performed on the sample overhead view image data to obtain sample distance features and sample boundary point category features corresponding to the sample overhead view image data.

[0078] In the embodiments of the present specification, specifically, the feature extraction model uses Regnet as the backbone network, i.e., the feature extraction network. The sample overhead view image data is input into a preset model, and feature extraction processing is performed through the feature extraction network Regnet. First, shallow texture, color and boundary information are learned to obtain shallow features, and then semantic feature information is learned through a deep network to fuse shallow features and high-level semantic information. Finally, the last layer of the network obtains sample distance features and sample boundary point category features; and the dimensions of the sample distance features and the sample boundary point category features are the same, and exemplarily, can be 360 ​​dimensions. The sample distance feature characterizes the predicted distance between the boundary point in the sample overhead image data and the center point of the sample overhead image data; the sample boundary point category feature characterizes the predicted category of the boundary point in the sample overhead image data; and by selecting Regnet as the backbone network, since the Regnet network optimizes the ratio of network width and depth, each layer can use its computing resources more efficiently, reducing the amount of calculation, thereby accelerating the training process and improving the convergence speed of the model, reducing the total time required for training iterations, and the design of the Regnet network takes into account the effective utilization of computing resources, thereby saving hardware resources; at the same time, since the Regnet architecture can flexibly adjust the width and depth according to needs, a lightweight version of the model can be constructed. These models significantly reduce the requirements for computing resources while maintaining high accuracy, optimize the model inference process, and the model is designed with computing efficiency in mind to ensure that input data can be processed and results can be output in a timely manner in practical applications, ensuring the real-time processing capability of the model on various hardware platforms. Moreover, when the image is input to the model, the model directly outputs a highly structured feature vector, which can be directly used in the decision-making logic of the autonomous driving system. It can be seen that while an end-to-end network is implemented, there is almost no post-processing calculation, which reduces the resource usage during actual use.

[0079] S303: Perform distance prediction processing on the sample distance feature based on the distance prediction network to obtain a sample distance result of the sample overhead view image data.

[0080] In the embodiments of the present specification, a network architecture specifically used for distance prediction is designed, and the distance prediction processing is performed on the sample distance features through the distance prediction network, so that the vehicle can achieve rapid response and accurate distance from the center point of the image to be processed to the boundary point of the drivable area. At the same time, it can work under different environmental conditions and maintain a high prediction accuracy even in the presence of light changes, occlusion or other interference factors.

[0081] S304: performing boundary point category recognition processing on the sample boundary point category features based on the boundary point category recognition network to obtain a sample boundary point category result of the overhead image data.

[0082] In the embodiments of the present specification, a network architecture specifically used for category recognition is designed, and distance prediction processing is performed on sample boundary point category features through the boundary point category network, so that the vehicle can respond quickly and accurately identify object categories, and the vehicle can maintain a high recognition accuracy under different lighting conditions, viewing angles and occlusion conditions.

[0083] S305: Training the preset model according to a first difference between the sample distance result and the sample distance label, and a second difference between the sample boundary point category result and the sample boundary point category label, and using the feature extraction network in the preset model after training as the feature extraction model.

[0084] In the embodiments of the present specification, the model is trained by combining distance differences and boundary point category differences, so that the model can achieve accurate recognition even in complex environments, thereby improving the accuracy, robustness and real-time performance of the model.

[0085] In the embodiment of the present specification, the preset model is trained according to the first difference between the sample distance result and the sample distance label, and the second difference between the sample boundary point category result and the sample boundary point category label, and the feature extraction network in the preset model after the training is used as the feature extraction model, such as Figure 5 As shown, Figure 5 The following is a flowchart of a method for detecting a vehicle drivable area, including:

[0086] S501: Determine first loss information according to a first difference between the sample distance result and the sample distance label.

[0087] In the embodiment of the present specification, the sample distance result is the predicted distance between the boundary point in the sample overhead image data and the center point of the sample overhead image data, and the sample distance label is the true distance between the boundary point in the sample overhead image data and the center point of the sample overhead image data. The first loss information is determined according to the difference between the true distance and the predicted distance. The first loss information can be determined using the loss function of the drivable area detection algorithm. For example, the loss function of Smooth L1 (smoothed mean absolute error function) can be used to measure the distance between the predicted distance and the true distance. The formula of the Smooth L1 loss function can be:

[0088]

[0089] Among them, l loss is the loss value of a single sample, x n is the true value distance, y nTo predict the distance, beta is a hyperparameter that determines the turning point where the loss function transitions from quadratic loss to linear loss.

[0090] The Smooth L1 loss function is used so that when the absolute value of the difference between the predicted distance and the true value distance is less than beta, the loss function takes the form of mean square error, that is, 0.5×(x n -y n ) 2 This form encourages the model to minimize the residual when the error is small, because the penalty for the error in the loss function is squared. When the absolute value of the difference between the predicted distance and the true value distance is greater than or equal to beta, the loss function takes the form of absolute error, but is linearly scaled, i.e. |x n -y n |-0.5×beta, this form gives a gentler penalty to larger errors, avoids excessive impact of outliers on model training, and improves the robustness of the model. It can be seen that the Smooth L1 loss function combines the advantages of mean square error and absolute error, balancing the sensitivity of the model when the error is small and the robustness when the error is large.

[0091] S502: Determine second loss information according to a second difference between the sample boundary point classification result and the sample boundary point classification label.

[0092] In the embodiment of the present specification, the sample boundary point classification result is the predicted category of the boundary point in the sample overhead image data, and the sample boundary point category label is the actual category of the boundary point in the sample overhead image data. The second loss information is determined according to the difference between the actual category and the predicted category. The second loss information can be determined by a category loss function. For example, a cross entropy loss function can be used to measure the difference between the predicted category and the actual category. The formula of the cross entropy loss function is:

[0093]

[0094] Among them, L CE is the loss value of a single sample, N is the total number of categories, i represents the number of categories, and y is the vector of the actual category. is the probability vector of the predicted category, y i is the value of the i-th element in the vector y of the actual category, is the probability vector of the predicted category The value of the i-th element in .

[0095] Using the cross entropy loss function, when calculating the category loss information, if the probability of the predicted category is close to 1, then will be a small negative number so that If the probability of the predicted category is far from 1, then It will be a large negative number, resulting in an increase in loss; by minimizing the cross entropy loss to predict a probability distribution closer to the actual category, the accuracy and reliability of the predicted category are improved.

[0096] S503: Determine target loss information according to the first loss information and the second loss information.

[0097] In the embodiment of the present specification, the target loss information indicating the distance difference and the category difference is determined based on the first loss information indicating the difference between the true value distance and the predicted distance and the second loss information indicating the difference between the actual category and the predicted category. Exemplarily, the target loss information is determined based on the loss function of the drivable area detection algorithm and the category loss function. The formula of the target loss information can be:

[0098] L total =λ·L SmoothL1 +(1-λ)·L CE

[0099] Among them, L total represents the target loss information, λ is the hyperparameter for adjusting the weight between the two loss information, and L SmoothL1 is the first loss information, L CE It is the second loss information.

[0100] Since different loss functions have different numerical ranges, the category loss function fluctuates within a smaller range, while the drivable area detection algorithm loss function has a larger value. For example, the weight of the category loss function can be 0.3, and the weight of the drivable area detection algorithm loss function can be 1. The effects of different losses on model training can be balanced by adjusting the hyperparameters.

[0101] S504: According to the target loss information, the parameters of the preset model are adjusted, and the feature extraction network in the preset model after training is used as the feature extraction model.

[0102] In the embodiments of this specification, after determining the target loss information, the target loss information can be used to calculate the gradient, and then the parameters of the preset model can be updated through back propagation, and the feature extraction network in the preset model after training is used as the feature extraction model. By combining the first loss information and the second loss information, the model parameters are adjusted, which not only balances the sensitivity of the model when the error is small and the robustness when the error is large, but also improves the accuracy and reliability of the model prediction results.

[0103] S202: Determine the distance feature of the target number dimension according to the distance between the center point of the target number of images to be processed and the boundary point of the image to be processed.

[0104] In the embodiment of this specification, since 360 ​​boundary points of the images to be processed are predicted, 360-dimensional distance features can be obtained, and each dimension represents the distance between the center point of an image to be processed and the boundary point of the image to be processed in the current dimension. By predicting the distance from the center point of the image to be processed to the boundary point of the drivable area in each dimension, the accuracy and reliability of the detection are improved.

[0105] S203: Input the image to be processed into the boundary point category recognition network in the feature extraction model for boundary point category recognition processing, identify the category of the boundary points of the image to be processed at each preset increment along the target direction of the image to be processed, and obtain the target number of boundary point categories.

[0106] In the embodiment of this specification, for example, the target direction may be a clockwise direction starting from the 12 o'clock direction of the top view of the image to be processed; the preset increment may be 1 degree; the category of the boundary point of the image to be processed is identified at each interval of the preset increment along the target direction of the image to be processed, that is, along the 12 o'clock direction of the top view of the image to be processed, the category of a boundary point is predicted at each interval of 1 degree, and the predicted result is the category of the boundary point of the current image to be processed. Since the image has a total of 360 degrees, a total of 360 boundary point categories can be predicted, and the target number can be 360.

[0107] S204: Determine boundary point category characteristics of the target quantity dimension according to the boundary point category of the target quantity.

[0108] In the embodiments of the present specification, since the categories of the boundary points of 360 images to be processed are predicted, 360-dimensional boundary point category features can be obtained, and each dimension represents the category of the boundary points of the image to be processed in the current dimension. By predicting the category of the boundary points of each dimension of the image to be processed, it is possible to predict the drivable area while detecting the categories of complex boundaries and obstacles, thereby improving the accuracy and diversity of detection.

[0109] S105: Determine predicted positions of boundary points of the image to be processed according to distance features corresponding to the boundary points.

[0110] In the embodiments of the present specification, the predicted position of the boundary point of each dimension in the image to be processed is determined by the distance between the boundary point of each dimension in the image to be processed and the center point of the image to be processed, thereby achieving a comprehensive judgment of the drivable area in the image to be processed and improving the accuracy and reliability of vehicle detection of the drivable area.

[0111] In the embodiment of this specification, the predicted position of the boundary point of the image to be processed is determined according to the distance feature corresponding to the boundary point, such as Figure 6 As shown, Figure 6 The following is a flowchart of a method for detecting a vehicle drivable area, including:

[0112] S601: Determine coordinate information of any target boundary point according to the distance between any target boundary point and the center point of the image to be processed; the target boundary point is any boundary point among the target number of boundary points.

[0113] In an embodiment of the present specification, the distance value between any target boundary point and the center point of the image to be processed is mapped to the center point of the image to be processed, and the distance value is converted into a coordinate value to determine the coordinate information of the target boundary point. For example, assuming that the horizontal direction of the image to be processed is the x-axis, the vertical direction of the image to be processed is the y-axis, the center point of the image to be processed is the origin, and the predicted distance of the target boundary point is y, the coordinate information of the target boundary point is determined according to the predicted distance y as (a, b).

[0114] S602: Determine the predicted position of the target boundary point according to the coordinate information of the target boundary point.

[0115] In the embodiments of this specification, for example, it is assumed that the horizontal direction of the image to be processed is the x-axis, the vertical direction of the image to be processed is the y-axis, the center point of the image to be processed is the origin, and the coordinates of the target boundary point are (a, b). According to the coordinate information of the target boundary point, a point with a horizontal value of a and a vertical value of b is determined, thereby determining the predicted position of the target boundary point.

[0116] S603: Determine the predicted position of each boundary point in the image to be processed according to the predicted position of the target boundary point.

[0117] In the embodiments of this specification, Figure 7 As shown, Figure 7 A schematic diagram of boundary point prediction. Figure 7 The points on the boundary represent the boundary points of the image to be processed. In fact, there should be 360 ​​boundary points for each dimension. The figure is only for illustration. By converting the predicted distance of each dimension into coordinate values ​​and then determining the predicted position of the boundary points of each dimension, the accuracy and reliability of vehicle detection in the drivable area are improved.

[0118] S107: Determine the predicted category of the boundary point according to the predicted position of the boundary point of the image to be processed and the category feature of the boundary point.

[0119] In the embodiments of the present specification, the predicted category of the boundary points in each dimension is determined based on the boundary points mapped on the boundary of the image to be processed and combined with the category features of the boundary points, thereby achieving all-round category prediction of the boundary of the image to be processed and achieving comprehensiveness and reliability of detection.

[0120] In the embodiment of the present specification, the predicted category of the boundary point is determined according to the predicted position of the boundary point of the image to be processed and the category feature of the boundary point, such as Figure 8 As shown, Figure 8 The following is a flowchart of a method for detecting a vehicle drivable area, including:

[0121] S801: Determine, according to the boundary point category features, a position category corresponding to the boundary point of the image to be processed.

[0122] In the embodiments of the present specification, the boundary point category feature represents the category of the boundary points of the image to be processed in each dimension, wherein the boundary point here is a fuzzy concept and is not precise to a point, but rather represents an approximate category of the boundary point position corresponding to this dimension. Therefore, the position category corresponding to the boundary point of the image to be processed in each dimension can be determined based on the boundary point category feature.

[0123] S802: Determine a target prediction category of the target boundary point according to any position category corresponding to the target boundary point and the predicted position of the target boundary point.

[0124] In an embodiment of the present specification, the position category of the dimension corresponding to the target boundary point is determined based on any position category corresponding to the target boundary point, and the specific boundary point position of the current dimension corresponding to the target boundary point is determined based on the predicted position of the target boundary point, thereby determining the target prediction category of the target boundary point.

[0125] S803: Determine the prediction category of each boundary point in the image to be processed according to the target prediction category of the target boundary point.

[0126] In the embodiments of this specification, Figure 7 As shown, Figure 7 A schematic diagram of boundary point prediction. Figure 7In the figure, 701 is a car, 702 is a wall, and 703 is no obstacle. Different category points can be represented by different color channels according to the output, so that the blue area points represent the drivable area, the red points represent the obstacle as a car, and the green area points represent the obstacle as a wall. Therefore, the boundary points next to obstacle 701 can be red, the boundary points next to obstacle 702 can be green, and the boundary points next to 703 can be blue. Combine the target prediction categories of all target boundary points to determine the prediction category of each boundary point in the image to be processed. By determining the category of the boundary point corresponding to each dimension, the category corresponding to all boundary points in the image to be processed can be accurately identified, which is convenient for identifying the category of the drivable area or obstacle, and improving the accuracy and efficiency of detection.

[0127] S109: Determine a drivable area of ​​the target vehicle according to the predicted positions of the boundary points of the image to be processed and the predicted categories of the boundary points.

[0128] In the embodiments of the present specification, the drivable area detected by the target vehicle is determined by the predicted boundary points on the boundary of the image to be processed and the predicted category corresponding to each boundary point. This not only detects the drivable area, but also ensures the integrity of the complex boundary prediction, thereby achieving all-round area detection and improving the accuracy and efficiency of detection.

[0129] In the embodiment of this specification, the drivable area of ​​the target vehicle is determined according to the predicted position of the boundary point of the image to be processed and the predicted category of the boundary point, such as Fig. 9 As shown, Fig. 9 The following is a flowchart of a method for detecting a vehicle drivable area, including:

[0130] S901: Connecting the boundary points of the image to be processed according to the predicted positions of the boundary points of the image to be processed to obtain a closed area.

[0131] In the embodiment of this specification, the specific position of each boundary point is located according to the predicted position of the boundary points of the image to be processed, and the boundary points of the image to be processed are connected to obtain a closed area, which simplifies the detection process and improves the accuracy and reliability of the detection.

[0132] S902: Determine the boundary points of the drivable area of ​​the image to be processed and the boundary points of the prohibited driving area of ​​the image to be processed according to the predicted categories of the boundary points.

[0133] In the embodiment of this specification, the boundary points of the drivable area represent all drivable areas, and the boundary points of the prohibited drivable area represent the existence of obstacles and the types of obstacles, where the types of obstacles can include walls, curbs, people, cars, pillars and other obstacles. The type of the boundary of the drivable area is accurately detected, the integrity of the complex boundary prediction is guaranteed, and the vehicle can avoid obstacles in time.

[0134] S903: Determine a drivable area for the target vehicle according to the drivable area boundary points, the prohibited drivable area boundary points and the closed area.

[0135] In the embodiments of this specification, Figure 10-11 As shown, Figure 10-11 A schematic diagram of a drivable area. Fig.10 Among them, 1001 is the target vehicle, 1003 is the obstacle vehicle, 1004 is the obstacle vehicle, and 1002 represents the closed area. Fig.11 1101 is the target vehicle, 1102 is the closed area, and 1103 is the obstacle wall. The drivable area is determined according to the boundary points of the drivable area, the boundary points of the prohibited drivable area, and the connected closed area. In practical applications, the drivable area can be marked in green to achieve a striking effect, which is convenient for users to distinguish the drivable area and avoid obstacles in time, thereby improving the safety during driving.

[0136] This specification also provides a detection device for a vehicle drivable area, such as Fig.12 As shown, the device comprises:

[0137] The image module 1201 to be processed is used to obtain an image to be processed in a preset area corresponding to a target vehicle; the preset area is an area whose distance from the target vehicle meets a preset distance condition;

[0138] The feature extraction module 1202 is used to extract features from the image to be processed to obtain distance features and boundary point category features corresponding to the boundary points; the distance features represent the distance between the boundary points of the image to be processed and the center point of the image to be processed;

[0139] A predicted position module 1203, used to determine the predicted position of the boundary point of the image to be processed according to the distance feature corresponding to the boundary point;

[0140] A prediction category module 1204, configured to determine a prediction category of the boundary point according to the predicted position of the boundary point of the image to be processed and the category characteristics of the boundary point;

[0141] The drivable area module 1205 is used to determine the drivable area of ​​the target vehicle according to the predicted positions of the boundary points of the image to be processed and the predicted categories of the boundary points.

[0142] In some embodiments, the feature extraction module further includes:

[0143] The target quantity submodule is used to input the image to be processed into the distance prediction network in the feature extraction model for distance prediction processing, and predict the distance between the center point of the image to be processed and the boundary point of the image to be processed at each preset increment along the target direction of the image to be processed, so as to obtain the target number of distances between the center point of the image to be processed and the boundary point of the image to be processed.

[0144] The target quantity dimension submodule is used to determine the distance feature of the target quantity dimension according to the distance between the center point of the target quantity of the image to be processed and the boundary point of the image to be processed.

[0145] The boundary point category submodule is used to input the image to be processed into the boundary point category recognition network in the feature extraction model for boundary point category recognition processing, identify the category of the boundary points of the image to be processed at each preset increment along the target direction of the image to be processed, and obtain the target number of boundary point categories.

[0146] The category feature submodule is used to determine the category features of the boundary points of the target quantity dimension according to the category of the boundary points of the target quantity.

[0147] In some embodiments, the predicted location module further includes:

[0148] The coordinate information submodule is used to determine the coordinate information of the target boundary point according to the distance between any target boundary point and the center point of the image to be processed; the target boundary point is any boundary point among the target number of boundary points.

[0149] The predicted position determination submodule is used to determine the predicted position of the target boundary point according to the coordinate information of the target boundary point.

[0150] The position determination submodule is used to determine the predicted position of each boundary point in the image to be processed according to the predicted position of the target boundary point.

[0151] In some embodiments, the prediction category module further includes:

[0152] The position category submodule is used to determine the position category corresponding to the boundary point of the image to be processed according to the boundary point category characteristics.

[0153] The target prediction category submodule is used to determine the target prediction category of the target boundary point according to any position category corresponding to the target boundary point and the predicted position of the target boundary point.

[0154] The prediction category determination submodule is used to determine the prediction category of each boundary point in the image to be processed according to the target prediction category of the target boundary point.

[0155] In some embodiments, the drivable area module further includes:

[0156] The closed area submodule is used to connect the boundary points of the image to be processed according to the predicted positions of the boundary points of the image to be processed to obtain a closed area.

[0157] The boundary point submodule is used to determine the boundary points of the drivable area of ​​the image to be processed and the boundary points of the prohibited driving area of ​​the image to be processed according to the predicted categories of the boundary points.

[0158] The drivable area determination submodule is used to determine the drivable area of ​​the target vehicle according to the drivable area boundary points, the prohibited driving area boundary points and the closed area.

[0159] In some embodiments, the target quantity submodule further includes:

[0160] The sample overhead view image data unit is used to obtain sample overhead view image data; the sample overhead view image data is annotated with a sample distance label and a sample boundary point category label.

[0161] The sample feature unit is used to perform feature extraction processing on the sample overhead view image data based on a feature extraction network of a preset model to obtain sample distance features and sample boundary point category features corresponding to the sample overhead view image data.

[0162] The sample distance result unit is used to perform distance prediction processing on the sample distance feature based on the distance prediction network to obtain a sample distance result of the sample overhead view image data.

[0163] The sample boundary point category result unit is used to perform boundary point category recognition processing on the sample boundary point category features based on the boundary point category recognition network to obtain the sample boundary point category result of the overhead image data.

[0164] A training unit is used to train the preset model according to a first difference between the sample distance result and the sample distance label, and a second difference between the sample boundary point category result and the sample boundary point category label, and use the feature extraction network in the preset model after the training as the feature extraction model.

[0165] In some embodiments, the training unit further comprises:

[0166] The first loss information subunit is used to determine first loss information according to a first difference between the sample distance result and the sample distance label.

[0167] The second loss information subunit is used to determine the second loss information according to a second difference between the sample boundary point classification result and the sample boundary point classification label.

[0168] The target loss information subunit is used to determine target loss information based on the first loss information and the second loss information.

[0169] A parameter updating subunit is used to adjust the parameters of the preset model according to the target loss information, and use the feature extraction network in the preset model after training as the feature extraction model.

[0170] The device and method embodiments in the described device embodiments are based on the same inventive concept.

[0171] An embodiment of the present specification provides a vehicle electronic device, which includes a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or at least one program is loaded and executed by the processor to implement a vehicle drivable area detection method as provided in the above method embodiment.

[0172] An embodiment of the present application also provides a computer storage medium, which can be set in a terminal to store at least one instruction or at least one program related to a vehicle drivable area detection method in a method embodiment, and the at least one instruction or at least one program is loaded and executed by the processor to implement the vehicle drivable area detection method provided in the above method embodiment.

[0173] The embodiment of the present application also provides a computer program product or a computer program, which includes computer instructions stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes to implement the vehicle drivable area detection method provided in the above method embodiment.

[0174] The memory described in the embodiments of this specification can be used to store software programs and modules, and the processor executes various functional applications and data processing by running the software programs and modules stored in the memory. The memory may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, application programs required for functions, etc.; the data storage area may store data created according to the use of the device, etc. In addition, the memory may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other volatile solid-state storage device. Accordingly, the memory may also include a memory controller to provide the processor with access to the memory.

[0175] The vehicle drivable area detection method embodiment provided in the embodiments of this specification can be executed in a mobile terminal, a computer terminal, a server or a similar computing device. Taking running on a server as an example, Fig.13 1 is a hardware structure block diagram of a server of a vehicle drivable area detection method provided in an embodiment of this specification. Fig.13 As shown, the server 1300 may have relatively large differences due to different configurations or performances, and may include one or more central processing units (CPU) 1310 (the central processing unit 1310 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 1330 for storing data, and one or more storage media 1320 (such as one or more mass storage devices) for storing application programs 1323 or data 1322. Among them, the memory 1330 and the storage medium 1320 can be short-term storage or permanent storage. The program stored in the storage medium 1320 may include one or more modules, each of which may include a series of instruction operations on the server. Furthermore, the central processing unit 1310 can be configured to communicate with the storage medium 1320 and execute a series of instruction operations in the storage medium 1320 on the server 1300. The server 1300 may also include one or more power supplies 1360, one or more wired or wireless network interfaces 1350, one or more input and output interfaces 1340, and / or one or more operating systems 1321, such as Windows ServerTM, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM, etc.

[0176] The input / output interface 1340 may be used to receive or send data via a network. The specific example of the network may include a wireless network provided by a communication provider of the server 1300. In one example, the input / output interface 1340 includes a network adapter (Network Interface Controller, NIC), which may be connected to other network devices through a base station so as to communicate with the Internet. In one example, the input / output interface 1340 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0177] It can be understood by those skilled in the art that Fig.13 The structure shown is for illustration only and does not limit the structure of the above electronic device. Fig.13 More or fewer components as shown, or with Fig.13 Different configurations shown.

[0178] It can be seen from the embodiments of the vehicle drivable area detection method, device and storage medium provided by the present application that the present application obtains a to-be-processed image of a preset area corresponding to a target vehicle; the preset area is an area whose distance from the target vehicle satisfies a preset distance condition; feature extraction is performed on the to-be-processed image to obtain distance features and boundary point category features corresponding to boundary points; the distance features characterize the distance between the boundary points of the to-be-processed image and the center point of the to-be-processed image; based on the distance features corresponding to the boundary points, the predicted positions of the boundary points of the to-be-processed image are determined; based on the predicted positions of the boundary points of the to-be-processed image and the boundary point category features, the predicted categories of the boundary points are determined; based on the predicted positions of the boundary points of the to-be-processed image and the boundary point category features, the drivable area of ​​the target vehicle is determined. This application uses Regnet as the backbone network to extract features from the image to be processed after IPM processing to obtain the distance features corresponding to the 360-dimensional boundary points and the 360-dimensional boundary point category features, thereby achieving feature vectors based on an end-to-end network and improving the efficiency, accuracy and flexibility of detection; the 360-degree prediction method is used to predict the boundary points and boundary point categories corresponding to each dimension on the image to be processed under the IPM perspective, thereby obtaining the drivable area, ensuring that the drivable area around the vehicle body in the top view is accurately predicted under limited computational complexity, improving the prediction accuracy in the case of blind spots, and because of the predicted boundary point categories, the integrity of the complex boundary prediction and the type of the drivable area boundary are detected. The comprehensiveness and accuracy of vehicle detection of the drivable area are achieved, and the safety of the vehicle during driving is improved.

[0179] It should be noted that the above sequence of the embodiments of this specification is for description only and does not represent the advantages and disadvantages of the embodiments. The above describes specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0180] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device, equipment, and storage medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0181] Those skilled in the art can understand that all or part of the steps of the above embodiments can be completed by hardware, or by a program to instruct the relevant hardware to complete, and the program can be stored in a computer storage medium, and the storage medium mentioned above can be a read-only memory, a disk or an optical disk, etc. The above is only a preferred embodiment of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for detecting a vehicle drivable area, characterized in that: The method comprises: Acquire an image to be processed of a preset area corresponding to a target vehicle; the preset area is an area whose distance from the target vehicle meets a preset distance condition; Extracting features of the image to be processed to obtain distance features and boundary point category features corresponding to the boundary points; the distance features represent the distance between the boundary points of the image to be processed and the center point of the image to be processed; Determining predicted positions of the boundary points of the image to be processed according to the distance features corresponding to the boundary points; Determining the predicted category of the boundary point according to the predicted position of the boundary point of the image to be processed and the category characteristics of the boundary point; The drivable area of ​​the target vehicle is determined according to the predicted positions of the boundary points of the image to be processed and the predicted categories of the boundary points.

2. The detection method according to claim 1, characterized in that: The extracting features of the image to be processed to obtain distance features and boundary point category features corresponding to the boundary points includes: Input the image to be processed into the distance prediction network in the feature extraction model for distance prediction processing, predict the distance between the center point of the image to be processed and the boundary point of the image to be processed at each preset increment along the target direction of the image to be processed, and obtain the target number of distances between the center point of the image to be processed and the boundary point of the image to be processed; Determining the distance feature of the target number dimension according to the distance between the center point of the target number of images to be processed and the boundary point of the target number of images to be processed; Inputting the image to be processed into the boundary point category recognition network in the feature extraction model to perform boundary point category recognition processing, identifying the category of the boundary points of the image to be processed at intervals of the preset increment along the target direction of the image to be processed, and obtaining the target number of boundary point categories; According to the boundary point categories of the target quantity, boundary point category features of the target quantity dimension are determined.

3. The detection method according to claim 2, characterized in that: The step of determining the predicted position of the boundary point of the image to be processed according to the distance feature corresponding to the boundary point includes: Determine the coordinate information of any target boundary point according to the distance between any target boundary point and the center point of the image to be processed; the target boundary point is any boundary point among the target number of boundary points; Determining the predicted position of the target boundary point according to the coordinate information of the target boundary point; According to the predicted position of the target boundary point, the predicted position of each boundary point in the image to be processed is determined.

4. The detection method according to claim 3, characterized in that: The step of determining the predicted category of the boundary point according to the predicted position of the boundary point of the image to be processed and the category feature of the boundary point includes: Determining, according to the boundary point category characteristics, the position category corresponding to the boundary point of the image to be processed; Determining a target prediction category of the target boundary point according to any position category corresponding to the target boundary point and the predicted position of the target boundary point; According to the target prediction category of the target boundary point, the prediction category of each boundary point in the image to be processed is determined.

5. The detection method according to claim 3, characterized in that: The determining the drivable area of ​​the target vehicle according to the predicted position of the boundary point of the image to be processed and the predicted category of the boundary point includes: Connecting the boundary points of the image to be processed according to the predicted positions of the boundary points of the image to be processed to obtain a closed area; Determining the boundary points of the drivable area of ​​the image to be processed and the boundary points of the prohibited drivable area of ​​the image to be processed according to the predicted categories of the boundary points; The drivable area of ​​the target vehicle is determined according to the drivable area boundary points, the prohibited drivable area boundary points and the closed area.

6. The detection method according to claim 2, characterized in that: The training method of the feature extraction model comprises: Acquire sample overhead view image data; the sample overhead view image data is annotated with a sample distance label and a sample boundary point category label; Based on a feature extraction network of a preset model, feature extraction processing is performed on the sample overhead view image data to obtain sample distance features and sample boundary point category features corresponding to the sample overhead view image data; Performing distance prediction processing on the sample distance feature based on the distance prediction network to obtain a sample distance result of the sample overhead view image data; Performing boundary point category recognition processing on the sample boundary point category features based on the boundary point category recognition network to obtain a sample boundary point category result of the overhead image data; The preset model is trained according to a first difference between the sample distance result and the sample distance label, and a second difference between the sample boundary point category result and the sample boundary point category label, and the feature extraction network in the preset model after the training is used as the feature extraction model.

7. The detection method according to claim 6, characterized in that: The method of training the preset model according to the first difference between the sample distance result and the sample distance label, and the second difference between the sample boundary point category result and the sample boundary point category label, and using the feature extraction network in the preset model after the training as the feature extraction model, comprises: determining first loss information according to a first difference between the sample distance result and the sample distance label; Determining second loss information according to a second difference between the sample boundary point classification result and the sample boundary point classification label; Determining target loss information according to the first loss information and the second loss information; According to the target loss information, the parameters of the preset model are adjusted, and the feature extraction network in the preset model after training is used as the feature extraction model.

8. A device for detecting a vehicle drivable area, characterized in that: The device comprises: The image module to be processed is used to obtain an image to be processed of a preset area corresponding to the target vehicle; the preset area is an area whose distance from the target vehicle meets a preset distance condition; A feature extraction module is used to extract features from the image to be processed to obtain distance features and boundary point category features corresponding to the boundary points; the distance features represent the distance between the boundary points of the image to be processed and the center point of the image to be processed; A prediction position module, used to determine the predicted position of the boundary point of the image to be processed according to the distance feature corresponding to the boundary point; A prediction category module, used to determine the prediction category of the boundary point according to the predicted position of the boundary point of the image to be processed and the category characteristics of the boundary point; The drivable area module is used to determine the drivable area of ​​the target vehicle according to the predicted positions of the boundary points of the image to be processed and the predicted categories of the boundary points.

9. An electronic device, characterized in that: The device includes a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or the at least one program is loaded by the processor and executes the method for detecting a vehicle drivable area as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that: The storage medium stores at least one instruction or at least one program, and the at least one instruction or at least one program is loaded and executed by the processor to implement the method for detecting the vehicle drivable area as described in any one of claims 1-7.

Images