Intelligent path recommendation method and system for dynamic generation of driving test scene

Abstract

The invention provides an intelligent path recommendation method and system for dynamic generation of a driving test scene, and relates to the technical field of path recommendation, and the method comprises the steps: obtaining driving test road condition features of a preset driving task; generating a road condition three-dimensional model according to the driving test road condition features to obtain a driving test scene model; performing vehicle practice path matching based on the driving test scene model to obtain a target vehicle practice path; and sending the target driving practice path to the user side to guide driving practice, thereby solving the technical problem of low training efficiency caused by incapability of dynamically generating a personalized practice path according to the driving test scene in the prior art, and improving the training efficiency by dynamically matching the driving test scene features and recommending the personalized training path. The technical effect of improving pertinence and efficiency of driving test simulation practice is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of route recommendation, and in particular to an intelligent route recommendation method and system for dynamically generating driving test scenarios. Background Art

[0002] With the rapid growth of the number of motor vehicles and the increasing complexity of traffic, driving tests have become an important part of ensuring road traffic safety. Traditional driving test training mainly relies on fixed venues and preset training routes. This method has the following limitations: First, fixed routes cannot fully simulate the changing real road conditions, resulting in students' lack of adaptability when encountering complex road conditions during the test or actual driving; second, the existing training methods lack personalized design and cannot adjust the training content according to the driving ability and learning progress of different students, which reduces learning efficiency; finally, the traditional path generation method cannot dynamically combine the real-time status of the venue (such as road congestion, temporary closure, etc.) to optimize the training path, resulting in low resource utilization. Summary of the invention

[0003] The present application provides an intelligent path recommendation method and system for dynamically generating driving test scenarios, which solves the technical problem in the prior art that a personalized practice path cannot be dynamically generated according to the driving test scenarios, resulting in low training efficiency.

[0004] In view of the above problems, the present application provides an intelligent path recommendation method and system for dynamic generation of driving test scenarios.

[0005] In a first aspect, the present application provides an intelligent path recommendation method for dynamically generating driving test scenarios, the method comprising: obtaining driving test road condition characteristics for a preset driving task; generating a three-dimensional road condition model based on the driving test road condition characteristics to obtain a driving test scenario model; matching driving practice paths based on the driving test scenario model to obtain a target driving practice path; and sending the target driving practice path to a user terminal to guide driving practice.

[0006] In the second aspect, the present application provides an intelligent path recommendation system for dynamically generating driving test scenarios, the system comprising: a driving test road condition feature acquisition module: obtaining the driving test road condition features of a preset driving task; a model generation module: generating a three-dimensional road condition model based on the driving test road condition features to obtain a driving test scenario model; a path matching module: matching driving practice paths based on the driving test scenario model to obtain a target driving practice path; a recommendation module: sending the target driving practice path to the user end to guide driving practice.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0008] First, the driving test road condition characteristics of the preset driving task are obtained. Next, a three-dimensional road condition model is generated based on the driving test road condition characteristics to obtain a driving test scene model. Then, the driving practice path is matched based on the driving test scene model to obtain the target driving practice path. Finally, the target driving practice path is sent to the user end to guide driving practice. The technical problem that the prior art cannot dynamically generate personalized practice paths according to the driving test scene, resulting in low training efficiency, is solved. By dynamically matching the driving test scene characteristics and recommending personalized training paths, the technical effect of improving the pertinence and efficiency of driving test simulation exercises is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 A flow chart of an intelligent path recommendation method for dynamic generation of driving test scenarios is provided for this application;

[0010] Figure 2 A flowchart of obtaining driving test road condition characteristics for a preset driving task in an intelligent path recommendation method for dynamic generation of driving test scenarios is provided for this application;

[0011] Figure 3 A structural diagram of an intelligent path recommendation system for dynamic generation of driving test scenarios is provided for this application.

[0012] Explanation of the reference numerals: driving test road condition feature acquisition module 11, model generation module 12, path matching module 13, recommendation module 14. DETAILED DESCRIPTION

[0013] The present application solves the technical problem in the prior art that a personalized practice path cannot be dynamically generated according to the driving test scenario, resulting in low training efficiency, by providing an intelligent path recommendation method and system for dynamic generation of driving test scenarios.

[0014] Embodiment 1, as Figure 1 As shown, the embodiment of the present application provides an intelligent path recommendation method for dynamic generation of driving test scenarios, the method comprising:

[0015] Step S100: Obtaining the driving test road condition characteristics of the preset driving task.

[0016] Preset driving tasks usually include specific test subjects or training goals, such as: reverse parking, parallel parking, right-angle turning, ramp starting, etc. Road condition data is collected from the test site using a variety of sensor devices (such as lidar, high-definition camera, GPS module, etc.). The collected data mainly includes the geometry of the road (width, curvature), traffic signs, ground slope, ambient lighting, etc. According to the requirements of the preset driving task, the extracted road condition features are matched with the task parameters. For example, for the reverse parking task, the focus is on extracting the parking space width and the curvature features of the entry path; for the parallel parking task, the focus is on the parking space length, the location of surrounding obstacles, etc.

[0017] Furthermore, if Figure 2 As shown, step S100 of the present application also includes:

[0018] Step S110: Collect the driving test road surface image and driving test road map of the target driving test site of the preset driving task; Step S120: Extract the driving test road height distribution information according to the driving test road map; Step S130: Perform road surface feature extraction on the driving test road surface image to obtain road surface width distribution information, road surface type distribution information and road surface turn distribution information; Step S140: Add the driving test road height distribution information, the road surface width distribution information, the road surface type distribution information and the road surface turn distribution information into the driving test road condition characteristics.

[0019] Specifically, use high-definition cameras, drones and other equipment to capture full coverage of the road surface of the target driving test site to obtain clear images of the driving test road surface. These images must include complete road markings, parking spaces, road conditions and environmental elements that may affect driving (such as obstacles, plants, etc.); obtain the driving test road map of the target driving test site through GPS or high-precision map sensing equipment, including road topology, bifurcation points, landmark locations and distribution of surrounding facilities; use geographic information system (GIS) to extract height information from the driving test road map to obtain height distribution information of the driving test road; perform image processing on the collected driving test road image, identify and measure the width of the road, and obtain road width distribution information; use image recognition technology to identify the material type of the road surface, such as asphalt, cement, gravel, etc., to obtain road type distribution information; perform geometric analysis on the road map, identify and extract the turning points of the road, and obtain road turning distribution information; integrate the extracted driving test road height distribution information, road width distribution information, road type distribution information and road turning distribution information into complete driving test road condition characteristics.

[0020] Furthermore, step S130 of the present application includes:

[0021] Step S131: Process the driving test road surface image through the road surface width extraction channel of the road surface feature extraction model to extract the road surface width distribution information; Step S132: Process the driving test road surface image through the road surface type extraction channel of the road surface feature extraction model to extract the road surface type distribution information; Step S133: Process the driving test road surface image through the road surface turn extraction channel of the road surface feature extraction model to extract the road surface turn distribution information.

[0022] Specifically, through the road width extraction channel, edge detection and boundary recognition are performed on the driving test road surface image to extract the width distribution information of each road section; through the road type extraction channel, the road material is classified using semantic segmentation and texture classification algorithms to obtain the distribution ratio of different road types such as asphalt and concrete; through the road turn extraction channel, combined with curvature analysis and fitting algorithms, turning curves are extracted from the road surface boundary, and features such as turning direction, angle and curvature radius are identified.

[0023] Furthermore, step S131 of the present application includes:

[0024] Step S1311: The driving test road surface image has an image scale mark; Step S1312: Input the image scale mark and the driving test road surface image into the road surface width extraction channel, and output the road surface width distribution information; wherein the road surface width extraction channel construction step includes: collecting road surface image recording data, image scale label data and road surface width identification data; using the road surface width identification data as supervision data, the road surface image recording data and the image scale label data as input, training a convolutional neural network, and obtaining the road surface width extraction channel.

[0025] Specifically, the driving test road surface image has a clear image scale mark. First, the image scale mark and the driving test road surface image are input into the road surface width extraction channel, and the road surface width distribution information is output after processing. Among them, the road surface width extraction channel construction steps include: collecting road surface image recording data containing different road conditions, and marking the image scale label data and road surface width identification data corresponding to the image; using the road surface width identification data as the supervision data, the road surface image recording data and the image scale label data as the model input, and establishing an extraction model that can automatically identify the road surface width distribution in the image by training the convolutional neural network, and finally forming the road surface width extraction channel.

[0026] Furthermore, the steps for constructing the road surface type extraction channel and the road surface turn extraction channel are the same as the steps for constructing the road surface width extraction channel.

[0027] The construction steps of the road surface type extraction channel and the road surface turn extraction channel are the same as the construction steps of the road surface width extraction channel, that is, the road surface type extraction channel and the road surface turn extraction channel are both trained and constructed based on the deep learning model of the convolutional neural network.

[0028] The steps for constructing a pavement type extraction channel include: collecting pavement image record data containing multiple pavement types (such as asphalt, concrete, gravel, etc.), and annotating the corresponding pavement type label data and image scale label data; then generating a pixel-level semantic segmentation label map from the annotated pavement type label data, which is used as the supervision data of the model; then, using the pavement image record data and image scale label data as input, training a convolutional neural network through a semantic segmentation model (such as U-Net or DeepLab) to achieve pixel-level accuracy in pavement type classification; finally, improving the generalization ability and extraction accuracy of the model through data enhancement and hyperparameter optimization, and deploying it as a pavement type extraction channel to extract pavement type distribution information.

[0029] The steps of constructing the road turn extraction channel include: collecting road image recording data of various turning scenes, and marking the identification data such as turning curve, direction (left turn / right turn), turning angle and curvature radius, and image scale label data; generating a label map with the marked turning features, including the turning curve path and related geometric parameters; inputting the road image recording data and the image scale label data into the convolutional neural network, and training through the improved ResNet or Mask R-CNN to ensure that the model can accurately extract the geometric features of the turning curve; finally, combining the curve fitting constraints (such as polynomial fitting or B-spline curve) to optimize the extraction results, and deploying them as the road turn extraction channel for extracting turning curves and related features.

[0030] Step S200: Generate a three-dimensional road condition model according to the driving test road condition characteristics to obtain a driving test scene model.

[0031] The collected driving test road height distribution information, road width distribution information, road type distribution information and road turn distribution information are used as input data, and a three-dimensional geometric model of the road condition is constructed through a three-dimensional modeling tool (such as OpenGL, Unity or Unreal Engine); secondly, the road height distribution information is combined to perform terrain fitting on the road surface to generate an accurate terrain contour; the road width is then adjusted according to the road width distribution information to ensure that the width changes of different road sections are in line with the actual situation; at the same time, the model is given material texture through the road type distribution information to present the appearance characteristics of different road types such as asphalt and concrete; finally, the curvature of the road curve is fitted using the turn distribution information, and the scene details are improved in combination with the road node and intersection data, such as adding scene elements such as marking lines, parking spaces, and traffic signs to generate a complete driving test scene model.

[0032] Step S300: performing driving practice path matching based on the driving test scenario model to obtain a target driving practice path.

[0033] First, the requirements of the preset driving task are analyzed, and the key points of the task (such as the starting point, the end point, and the key path nodes) are extracted. The path candidate set is initially screened in combination with the road condition feature data in the driving test scenario model (including height distribution, road width, turning curvature, etc.); secondly, the path planning algorithm (such as the Dijkstra algorithm, the A* algorithm, or a path optimization algorithm based on deep learning) is used to search for the optimal path that meets the task requirements in the three-dimensional scene model, taking into account the path conditions, road condition complexity, and the difficulty of task execution; then, according to the training objectives of the preset driving task (such as training reverse parking, side parking, or starting on a slope, etc.), the matching paths are further screened in the candidate paths, and the path planning is dynamically adjusted to adapt to the dynamic changes in the scene (such as the increase or decrease of obstacles or the update of road conditions); finally, the matched target driving training path is output, and the target driving training path covers the training elements required for the task.

[0034] Furthermore, step S300 of the present application also includes:

[0035] Step S310: Obtain the road condition characteristics of the driving practice area; Step S320: Sorting the roads in the driving practice area based on the preset driving task to obtain a first driving practice path; Step S330: Matching the first driving practice path road condition characteristics of the first driving practice path according to the road condition characteristics of the driving practice area; Step S340: Generating a three-dimensional road condition model according to the first driving practice path road condition characteristics to obtain a driving practice scene model; Step S350: When the first scene similarity between the driving test scene model and the driving practice scene model is greater than or equal to a similarity threshold, adding the first driving practice path to the target driving practice path.

[0036] Preferably, first, the road condition characteristics of the driving practice area are obtained, including road distribution information, height distribution information, road surface type information and turn distribution information of the driving practice area, so as to provide basic data for subsequent path matching; secondly, the roads in the driving practice area are sorted based on the preset driving task, and the roads that meet the conditions are screened out to form the first driving practice path according to the training focus of the task requirements (such as reversing into a garage, side parking, right-angle turns, etc.); then, according to the road condition characteristics of the driving practice area, the road condition characteristics of the first driving practice path related to the first driving practice path are extracted, including information such as width, curvature, slope and material; further, three-dimensional modeling is performed according to the road condition characteristics of the first driving practice path, and a driving practice scene model corresponding to the first driving practice path is generated; finally, the similarity between the driving test scene model and the first scene in the driving practice scene model is calculated, and when the similarity is greater than or equal to the set similarity threshold, the first driving practice path is added to the target driving practice path set as a path that meets the task requirements.

[0037] Step S400: sending the target driving practice route to the user terminal to guide driving practice.

[0038] First, the target driving practice path data is formatted, including the starting point, end point, key turning nodes and road condition characteristics (such as width, slope, turning radius, etc.) of the path, and converted into a visual data format suitable for user-side display and interaction; secondly, a three-dimensional rendering tool (such as Unity or Unreal Engine) is used to generate an intuitive display of the path on the user side, and the complete layout of the target path is rendered in real time through a three-dimensional scene, and key path nodes and their operation prompts are marked, such as turning prompts, parking points and road width restrictions; then, combined with the user-side interactive equipment (such as a driving simulator, touch screen or virtual reality device), a multi-mode simulation practice function is provided, allowing users to perform actual driving tasks according to the target path; finally, during the user simulation practice process, user operation data (such as steering wheel angle, acceleration and deceleration conditions, etc.) is collected in real time, and compared and analyzed with standard operation parameters in path planning, providing users with instant feedback and optimization suggestions to help them improve their driving skills.

[0039] Furthermore, step S400 of the present application also includes:

[0040] Step S410: Set the target driving practice path as the first-stage driving practice path and recommend it to the user end for simulation practice, and obtain first user practice feedback information; Step S420: When the first user practice feedback information includes a practice pass signal, extract the second-stage driving practice path whose similarity to the second scene of the driving test scene model is less than a similarity threshold; Step S430: Recommend the second-stage driving practice path to the user end for simulation practice, and obtain second user practice feedback information; Step S440: When the second user practice feedback information includes a practice pass signal, prompt the user to take the driving test.

[0041] Preferably, first, the target driving practice path is set as the first-stage driving practice path, and is recommended to the user end for simulation practice; after the user completes the simulation practice of the first-stage driving practice path, the system collects the user's practice feedback information in real time and analyzes the user's operation; when the first user's practice feedback information contains a practice qualification signal (for example, the operation accuracy reaches the preset standard, the key task is successfully completed, etc.), the system extracts the path with a similarity to the second scene of the driving test scene model less than the similarity threshold from the driving practice path database as the second-stage driving practice path to further improve the user's ability to cope with complex road conditions; then, the second-stage driving practice path is recommended to the user end for simulation practice, and the second user's practice feedback information is also collected; when the second user's practice feedback information shows that the user's practice in a more complex scene also meets the qualification standard, the system prompts the user to have the ability to take the formal driving test, and recommends subsequent test arrangements or related guidance. Through phased path recommendations and gradually complex simulation exercises, it is ensured that users can gradually master from basic skills to complex road conditions, effectively improving driving ability and test pass rate.

[0042] In summary, the embodiments of the present application have at least the following technical effects:

[0043] First, the driving test road condition characteristics of the preset driving task are obtained. Next, a three-dimensional road condition model is generated based on the driving test road condition characteristics to obtain a driving test scene model. Then, the driving practice path is matched based on the driving test scene model to obtain the target driving practice path. Finally, the target driving practice path is sent to the user end to guide driving practice. The technical problem that the prior art cannot dynamically generate personalized practice paths according to the driving test scene, resulting in low training efficiency, is solved. By dynamically matching the driving test scene characteristics and recommending personalized training paths, the technical effect of improving the pertinence and efficiency of driving test simulation exercises is achieved.

[0044] Embodiment 2 is based on the same inventive concept as the intelligent path recommendation method for dynamic generation of driving test scenarios in the aforementioned embodiment. Figure 3 As shown, the present application provides an intelligent path recommendation system for dynamic generation of driving test scenarios, the system comprising:

[0045] Driving test road condition feature acquisition module 11: obtains the driving test road condition features of the preset driving task; model generation module 12: generates a three-dimensional road condition model according to the driving test road condition features to obtain a driving test scene model; path matching module 13: matches the driving practice path based on the driving test scene model to obtain a target driving practice path; recommendation module 14: sends the target driving practice path to the user end to guide driving practice.

[0046] Furthermore, the driving test road condition feature acquisition module 11 is used to execute the following method:

[0047] Collect a driving test road surface image and a driving test road map of a target driving test site for the preset driving task; extract driving test road height distribution information based on the driving test road map; perform road surface feature extraction on the driving test road surface image to obtain road surface width distribution information, road surface type distribution information and road surface turn distribution information; add the driving test road height distribution information, the road surface width distribution information, the road surface type distribution information and the road surface turn distribution information into the driving test road condition characteristics.

[0048] Furthermore, the driving test road condition feature acquisition module 11 is used to execute the following method:

[0049] The driving test road surface image is processed through the road surface width extraction channel of the road surface feature extraction model to extract the road surface width distribution information; the driving test road surface image is processed through the road surface type extraction channel of the road surface feature extraction model to extract the road surface type distribution information; the driving test road surface image is processed through the road surface turn extraction channel of the road surface feature extraction model to extract the road surface turn distribution information.

[0050] Furthermore, the driving test road condition feature acquisition module 11 is used to execute the following method:

[0051] The driving test road surface image has an image scale mark; the image scale mark and the driving test road surface image are input into the road surface width extraction channel, and the road surface width distribution information is output; wherein the road surface width extraction channel construction step includes: collecting road surface image recording data, image scale label data and road surface width identification data; using the road surface width identification data as supervision data, the road surface image recording data and the image scale label data as input, training a convolutional neural network, and obtaining the road surface width extraction channel.

[0052] Furthermore, the driving test road condition feature acquisition module 11 is used to execute the following method:

[0053] The steps for constructing the road surface type extraction channel and the road surface turn extraction channel are the same as the steps for constructing the road surface width extraction channel.

[0054] Furthermore, the path matching module 13 is used to execute the following method:

[0055] Obtain road condition characteristics of a driving practice area; sort roads in the driving practice area based on the preset driving task to obtain a first driving practice path; match the first driving practice path road condition characteristics of the first driving practice path according to the road condition characteristics of the driving practice area; generate a three-dimensional road condition model according to the first driving practice path road condition characteristics to obtain a driving practice scene model; when the first scene similarity between the driving test scene model and the driving practice scene model is greater than or equal to a similarity threshold, add the first driving practice path to the target driving practice path.

[0056] Furthermore, the recommendation module 14 is used to perform the following method:

[0057] The target driving practice path is set as the first-stage driving practice path and recommended to the user end for simulation practice to obtain first user practice feedback information; when the first user practice feedback information includes a practice pass signal, extract the second-stage driving practice path whose similarity to the second scene of the driving test scene model is less than a similarity threshold; recommend the second-stage driving practice path to the user end for simulation practice to obtain second user practice feedback information; when the second user practice feedback information includes a practice pass signal, prompt the user to take the driving test.

[0058] Through the above detailed description of the intelligent path recommendation method for dynamic generation of driving test scenarios in this specification, those skilled in the art can clearly understand the intelligent path recommendation system for dynamic generation of driving test scenarios in this embodiment. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description.

[0059] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An intelligent path recommendation method for dynamic generation of driving test scenarios, characterized in that: The method comprises: Obtaining driving test road condition characteristics for a preset driving task; Generate a three-dimensional road condition model according to the driving test road condition characteristics to obtain a driving test scene model; Perform driving practice path matching based on the driving test scenario model to obtain a target driving practice path; The target driving practice path is sent to the user terminal to guide the driving practice.

2. The intelligent path recommendation method for dynamic generation of driving test scenarios according to claim 1, characterized in that: Get the driving test road condition characteristics of the preset driving task, including: Collecting a driving test road surface image and a driving test road map of a target driving test site for the preset driving task; Extracting driving test road height distribution information according to the driving test road map; Extracting road features from the driving test road image to obtain road width distribution information, road type distribution information, and road turn distribution information; The driving test road height distribution information, the road surface width distribution information, the road surface type distribution information and the road surface turning distribution information are added to the driving test road condition characteristics.

3. The intelligent path recommendation method for dynamic generation of driving test scenarios according to claim 2, characterized in that: The road surface features are extracted from the driving test road surface image to obtain road surface width distribution information, road surface type distribution information and road surface turn distribution information, including: Processing the driving test road surface image through a road surface width extraction channel of a road surface feature extraction model to extract the road surface width distribution information; Processing the driving test road surface image through a road surface type extraction channel of a road surface feature extraction model to extract the road surface type distribution information; The driving test road surface image is processed through the road surface turn extraction channel of the road surface feature extraction model to extract the road surface turn distribution information.

4. The intelligent path recommendation method for dynamic generation of driving test scenarios according to claim 3, characterized in that: The driving test road surface image is processed through a road surface width extraction channel of a road surface feature extraction model to extract the road surface width distribution information, including: The driving test road surface image has an image scale mark; Input the image scale mark and the driving test road surface image into the road surface width extraction channel, and output the road surface width distribution information; The road width extraction channel construction step includes: Collect road surface image record data, image scale label data and road surface width identification data; The road width identification data is used as supervision data, and the road image recording data and the image scale label data are used as input to train a convolutional neural network to obtain the road width extraction channel.

5. The intelligent path recommendation method for dynamic generation of driving test scenarios according to claim 4, characterized in that: The steps for constructing the road surface type extraction channel and the road surface turn extraction channel are the same as the steps for constructing the road surface width extraction channel.

6. The intelligent path recommendation method for dynamic generation of driving test scenarios according to claim 1, characterized in that: The driving practice path is matched based on the driving test scenario model to obtain a target driving practice path, including: Obtain the road condition characteristics of the driving practice area; Based on the preset driving task, roads in the driving practice area are sorted to obtain a first driving practice path; According to the road condition characteristics of the driving practice area, matching the first driving practice path road condition characteristics of the first driving practice path; Generate a three-dimensional road condition model according to the road condition characteristics of the first driving practice route to obtain a driving practice scene model; When the first scene similarity between the driving test scenario model and the driving practice scenario model is greater than or equal to a similarity threshold, the first driving practice path is added to the target driving practice path.

7. The intelligent path recommendation method for dynamic generation of driving test scenarios according to claim 6, characterized in that: The target driving practice path is sent to the user terminal to guide driving practice, including: The target driving practice route is set as the first stage driving practice route and recommended to the user terminal for simulation practice, and the first user practice feedback information is obtained; When the first user practice feedback information includes a practice pass signal, extracting a second-stage driving practice path whose similarity to the second scene of the driving test scene model is less than a similarity threshold; Recommending the second-stage driving practice route to the user terminal for simulation practice, and obtaining practice feedback information from the second user; When the second user practice feedback information includes a practice pass signal, the user is prompted to take a driving test.

8. An intelligent path recommendation system for dynamic generation of driving test scenarios, characterized in that: The system comprises: Driving test road condition feature acquisition module: obtains the driving test road condition features of the preset driving task; Model generation module: generates a three-dimensional road condition model according to the driving test road condition characteristics to obtain a driving test scene model; Path matching module: performing driving practice path matching based on the driving test scenario model to obtain a target driving practice path; Recommendation module: sends the target driving practice path to the user terminal to guide driving practice.

9. The intelligent path recommendation system for dynamic generation of driving test scenarios according to claim 8, characterized in that: The driving test road condition feature acquisition module includes: Data acquisition module: collects the driving test road surface image and driving test road map of the target driving test site of the preset driving task; Height feature extraction module: extracting the height distribution information of the driving test road according to the driving test road map; Road surface feature extraction module: extracting road surface features from the driving test road surface image to obtain road surface width distribution information, road surface type distribution information and road surface turn distribution information; Driving test road condition feature generation module: adds the driving test road height distribution information, the road surface width distribution information, the road surface type distribution information and the road surface turn distribution information into the driving test road condition features.

10. The intelligent path recommendation system for dynamic generation of driving test scenarios according to claim 9, characterized in that: The road feature extraction module comprises: The first image processing module processes the driving test road surface image through the road surface width extraction channel of the road surface feature extraction model to extract the road surface width distribution information; The second image processing module processes the driving test road surface image through the road surface type extraction channel of the road surface feature extraction model to extract the road surface type distribution information; The third image processing module processes the driving test road surface image through the road surface turn extraction channel of the road surface feature extraction model to extract the road surface turn distribution information.

Images