Driving control method and device, storage medium and electronic equipment

Abstract

The invention discloses a driving control method and device, a storage medium and electronic equipment. The method comprises the following steps: determining an initial image comprising an indication board; determining a target data enhancement strategy based on the object features corresponding to the initial image; processing the initial image by adopting a target data enhancement strategy to obtain a target image; a target model is adopted, the target image is recognized, the recognition result of the indication board is obtained, the target model is obtained through training based on an image adopting multiple data features, and the multiple data features correspond to different candidate data enhancement strategies respectively; and determining a driving control strategy of the vehicle based on the identification result. The technical problem that the quality of an image recognition result is poor and the vehicle control accuracy is reduced in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the field of image recognition and assisted driving, and in particular to a driving control method, device, storage medium and electronic equipment. Background Art

[0002] The assisted driving system obtains image information of the vehicle's surrounding environment through sensors such as cameras, and uses image recognition technology to process and analyze this information to identify key factors such as traffic signs, lane lines, pedestrians, and vehicles. This environmental perception capability is the basis for the assisted driving system to make correct decisions and take corresponding measures.

[0003] Based on the results of image recognition, the assisted driving system can provide drivers with auxiliary functions such as warning, obstacle avoidance, lane keeping, and even achieve partial autonomous driving in some cases. The realization of these functions is highly dependent on the accuracy and real-time performance of image recognition. The training of image recognition models in related technologies requires a large amount of high-quality data support, but in actual applications, due to the complexity of data collection and labeling, it is often difficult to obtain sufficient and accurate data sets, which affects the training effect and recognition ability of the model, and to a certain extent limits the performance and safety of the assisted driving system.

[0004] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0005] The embodiments of the present invention provide a driving control method, device, storage medium and electronic device to at least solve the technical problem in the related art that the quality of image recognition results decreases and the vehicle control accuracy is reduced.

[0006] According to one aspect of an embodiment of the present invention, a driving control method is provided, comprising: determining an initial image including a signboard; determining a target data enhancement strategy based on object features corresponding to the initial image; processing the initial image using the target data enhancement strategy to obtain a target image; recognizing the target image using a target model to obtain a recognition result of the signboard, wherein the target model is trained based on an image using multiple data features, and the multiple data features respectively correspond to different candidate data enhancement strategies; and determining a vehicle driving control strategy based on the recognition result.

[0007] Optionally, determining the target data enhancement strategy based on the object features corresponding to the initial image includes: performing feature extraction on the initial image to determine the object size, position distribution, and color features of the sign; and determining the target data enhancement strategy from multiple candidate data strategies based on the object size, the position distribution, and the color features.

[0008] Optionally, the method also includes: inputting the images of the multiple data features into the target model for processing, and determining the training states corresponding to the multiple data features respectively; based on the training states corresponding to the multiple data features respectively, updating the initial data enhancement strategies corresponding to the multiple data features respectively, and obtaining the multiple candidate data strategies.

[0009] Optionally, the initial data enhancement strategy includes multiple processes, the training state includes a training loss value and a model convergence speed, and based on the training states corresponding to the multiple data features, the initial data enhancement strategies corresponding to the multiple data features are updated to obtain the multiple candidate data strategies, including: determining the application probabilities corresponding to the multiple processes based on the training loss values ​​corresponding to the multiple data features; determining the application strengths corresponding to the multiple processes based on the model convergence speeds corresponding to the multiple data features; determining the multiple candidate data strategies according to the application probabilities and application strengths corresponding to the multiple processes.

[0010] Optionally, the multiple processes include at least one of the following: rotation processing, scaling processing, translation processing, color transformation processing, blur processing, and noise addition processing.

[0011] Optionally, the target data enhancement strategy is used to process the initial image to obtain a target image, including: obtaining weather information of the area where the vehicle is located; determining an image adjustment strategy corresponding to the weather information; using the target data enhancement strategy to process the initial image to obtain an intermediate image; and adjusting the intermediate image based on the image adjustment strategy to obtain the target image.

[0012] Optionally, determining the driving control strategy of the vehicle based on the recognition result includes: based on a speed limit value indicated by the recognition result, determining the driving control strategy to be: adjusting the current speed of the vehicle so that the adjusted current speed is less than the speed limit value; and setting the maximum speed of the vehicle to the speed limit value; in response to the recognition result indicating lifting the speed limit, determining the driving control strategy to be: lifting the restriction on the maximum speed.

[0013] According to another aspect of an embodiment of the present invention, a driving control device is provided, comprising: an image acquisition module for determining an initial image including a signboard; a strategy determination module for determining a target data enhancement strategy based on object features corresponding to the initial image; an image processing module for processing the initial image using the target data enhancement strategy to obtain a target image; a recognition module for recognizing the target image using a target model to obtain a recognition result of the signboard, wherein the target model is obtained by training based on images using multiple data features, and the multiple data features respectively correspond to different candidate data enhancement strategies; and a control module for determining a vehicle driving control strategy based on the recognition result.

[0014] According to another aspect of an embodiment of the present invention, a non-volatile storage medium is provided, wherein the non-volatile storage medium stores a plurality of instructions, wherein the instructions are suitable for being loaded by a processor and executing any one of the driving control methods.

[0015] According to another aspect of an embodiment of the present invention, an electronic device is provided, comprising: one or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement any one of the driving control methods described.

[0016] In an embodiment of the present invention, an initial image including a sign is determined; a target data enhancement strategy is determined based on the object features corresponding to the initial image; the target data enhancement strategy is used to process the initial image to obtain a target image; a target model is used to recognize the target image to obtain a recognition result of the sign, wherein the target model is obtained by training based on an image using multiple data features, and the multiple data features correspond to different candidate data enhancement strategies respectively; based on the recognition result, a vehicle driving control strategy is determined. The purpose of improving the image recognition accuracy and robustness is achieved, and the technical effect of improving the accuracy of assisted driving control is achieved, thereby solving the technical problem of image recognition result quality and reduced vehicle control accuracy in the related technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0018] Figure 1 is a flow chart of an optional driving control method provided according to an embodiment of the present invention;

[0019] Figure 2 is a schematic diagram of an optional driving control device provided according to an embodiment of the present invention. DETAILED DESCRIPTION

[0020] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0021] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention 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 interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

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

[0023] Figure 1 is a flow chart of an optional driving control method provided according to an embodiment of the present invention; Figure 1 As shown, the method comprises the following steps:

[0024] Step S102, determining an initial image including a sign;

[0025] It can be understood that the image containing the sign is obtained from the camera or other image acquisition device installed on the vehicle. The initial image is the basis for subsequent processing and can clearly and accurately capture the information of the sign. Through the above processing, it is ensured that there is a reliable data source for the subsequent processing process, providing the necessary input for image processing and recognition in the subsequent steps.

[0026] Optionally, a high-definition camera is used to capture street view images in real time to ensure clear image quality and strong light adaptability. The initial image can be pre-processed by denoising and contrast enhancement to facilitate feature extraction of features such as the size, position, and color of the sign.

[0027] Step S104, determining a target data enhancement strategy based on the object features corresponding to the initial image;

[0028] It can be understood that the object features in the initial image (such as the shape, color, texture, etc. of the sign) are analyzed, and the most appropriate data augmentation strategy is selected based on these features. Data augmentation strategies may include rotation, scaling, brightness adjustment, contrast adjustment, etc., aiming to improve the generalization ability and recognition accuracy of the model. Through targeted data augmentation, the model's ability to recognize signs in complex or changing environments can be improved, and the robustness of the system can be enhanced.

[0029] In an optional embodiment, a target data enhancement strategy is determined based on object features corresponding to the initial image, including: extracting features from the initial image to determine the object size, position distribution, and color features of the sign; and determining the target data enhancement strategy from multiple candidate data strategies based on the object size, position distribution, and color features.

[0030] It can be understood that in order to determine the best data enhancement strategy for the initial image, feature extraction processing is required, and the initial image is deeply analyzed using image processing technology and algorithms to accurately extract features related to the sign, including the object size, position distribution, and color features of the sign. Based on object size, position distribution, and color features, based on considerations of computing resources and satisfaction of real-time requirements, the target data enhancement strategy that best suits the current image situation is selected from multiple candidate strategies. Through the above processing, highly customized data enhancement of the initial image is achieved. Compared with traditional general data enhancement methods, this method can realize personalized data enhancement methods, which not only helps to improve the quality of the image and make it easier to be processed by the recognition model, but also can significantly improve the accuracy and robustness of the recognition model.

[0031] Optionally, the size of the signboard may be represented by its pixel area in the image, the position distribution may be determined by the center coordinates and the tilt angle of the signboard, and the color feature may be the main color tone, color saturation, and the like.

[0032] It should be noted that by enhancing the data for the features of the sign, the noise and interference factors in the image are reduced, so that the recognition model can more accurately capture the information of the sign, thereby improving the recognition accuracy. Since the data enhancement strategy is selected based on the specific features of the sign, this method can cope with image recognition needs in different scenarios and conditions, and enhance the robustness of the system. By selecting the most suitable data enhancement strategy, unnecessary waste of computing resources is avoided and the overall efficiency of the system is improved. For example: in application scenarios with high real-time requirements, this method can ensure the rapid completion of data enhancement, thereby meeting the needs of real-time recognition.

[0033] In an optional embodiment, the method also includes: inputting images of multiple data features into a target model for processing, and determining training states corresponding to the multiple data features respectively; based on the training states corresponding to the multiple data features respectively, updating initial data enhancement strategies corresponding to the multiple data features respectively, and obtaining multiple candidate data strategies.

[0034] It can be understood that a variety of images containing different data features (such as object size, position distribution, color features, etc.) are input into the target model for processing. These images are used to evaluate the performance of the model under different features. Through the processing of the input image by the model, the performance of the model in identifying each data feature, that is, the training state, is evaluated. Based on the evaluation results of the above training state, the initial data enhancement strategy corresponding to each data feature is adjusted and optimized. If a feature performs poorly in training, the data enhancement intensity of the feature may be increased or the enhancement strategy may be changed to improve the model's recognition ability of the feature. After updating, the original initial data enhancement strategy set is transformed into a candidate data strategy set containing multiple processing, which improves the recognition accuracy and robustness of the model in subsequent practical applications. By introducing this data enhancement strategy update mechanism based on training state feedback, this method achieves continuous optimization of the data enhancement strategy.

[0035] Optionally, the above training status may also include evaluation indicators such as accuracy, recall, and F1 score to quantify the recognition ability of the model under different features. Accuracy refers to the ratio of the number of samples correctly classified by the model to the total number of samples, which is used to measure the performance of the model and reflects the proportion of the overall correct prediction of the model. It can be expressed in the following way:

[0036] Accuracy = (number of correctly classified samples / total number of samples) × 100%

[0037] Recall refers to the ratio of the number of positive samples correctly predicted by the model to the number of actual positive samples. It is used to measure the proportion of samples that are actually positive and are correctly predicted by the model. It can be expressed in the following way:

[0038] Recall rate = (number of correctly predicted positive samples / number of actual positive samples) × 100%

[0039] The F1 score is the harmonic mean of precision and recall and can be expressed as follows:

[0040] F1 score = 2 × (precision × recall) / (precision + recall).

[0041] In an optional embodiment, the initial data enhancement strategy includes multiple processes, the training state includes a training loss value and a model convergence speed, and based on the training states corresponding to the multiple data features, the initial data enhancement strategies corresponding to the multiple data features are updated to obtain multiple candidate data strategies, including: determining the application probabilities corresponding to the multiple processes based on the training loss values ​​corresponding to the multiple data features; determining the application strengths corresponding to the multiple processes based on the model convergence speeds corresponding to the multiple data features; and determining multiple candidate data strategies based on the application probabilities and application strengths corresponding to the multiple processes.

[0042] It can be understood that for each data feature, the loss value generated during the training process is analyzed. The training loss value is an important indicator to measure the difference between the model prediction and the true label. A lower loss value usually means that the model performs better on the data feature. According to the effect of each processing step in reducing the training loss value of a specific data feature, an application probability is assigned to the processing. The more significant the identification effect of the processing step, the higher its application probability should be. The model convergence speed reflects how fast the model reaches a stable state during the training process. A faster convergence speed means that the model can learn the inherent laws of the data faster. Based on the model convergence speed of each data feature, an application intensity is determined for each processing step. For features with a slower convergence speed, the application intensity of the relevant processing steps can be increased to promote the rapid convergence of the model. The application probability and application intensity of each processing step are comprehensively considered to generate multiple candidate data strategies. While maintaining a certain diversity, the specific performance of the model on different data features is also optimized. In this way, the data enhancement strategy can be adjusted and optimized more flexibly to meet the needs of different data features and model performance.

[0043] It should be noted that by adjusting the data enhancement strategy in a targeted manner, the performance of the model on different data features can be improved, thereby improving the overall performance. Increasing the processing intensity of features with slower convergence speed can accelerate the model training process and shorten the training time. The generated candidate data strategies are diverse, and the most appropriate strategy can be selected according to the actual situation to improve the adaptability and flexibility of the strategy.

[0044] Optionally, multiple candidate data strategies are evaluated to select the best strategy or strategy combination. The above evaluation process can be performed based on methods such as cross-validation and model performance comparison.

[0045] Optionally, the generalization ability of the model on different datasets and tasks can be improved through dynamic adjustment and diversified enhancement methods, reducing the risk of overfitting and underfitting.

[0046] In an optional embodiment, the multiple processes include at least one of the following: rotation processing, scaling processing, translation processing, color transformation processing, blurring processing, and noise adding processing.

[0047] It can be understood that the multiple processing includes at least one of the following: rotation processing, scaling processing, translation processing, color change processing, blur processing, and noise addition processing. Scaling processing can be scaling the sign to different sizes according to the actual size of the sign in the image to simulate the shooting effect at different distances and viewing angles. For smaller signs, it can be appropriately enlarged; for larger signs, it is reduced. Translation processing can be translating the position of the traffic sign in the image to simulate its appearance in different lanes and different positions, which helps the model learn the feature representation of the sign at different positions. Rotation processing can be rotating the sign image to different angles (such as ±10° to ±45°), or flipping horizontally / vertically to simulate its placement in different directions or angles. Color transformation can be adjusting parameters such as brightness, contrast and saturation of the image to simulate the color change of the sign under different lighting conditions. For brightly colored signs (such as red, blue, etc.), focus on their color performance under different lighting conditions. Blurring processing, also known as image smoothing processing, is a basic operation in image processing, which is used to reduce noise and details in the image and make the image smoother. Blurring can be achieved through a variety of methods, including mean blur, Gaussian blur, median blur, etc. Noise addition is an enhancement technique in image processing. It increases the diversity and complexity of the image by adding noise to the image, thereby improving the performance and robustness of the image processing algorithm. The noise type can be Gaussian noise.

[0048] Optionally, a combination of multiple enhancement techniques can be used, each targeting different data features or training objectives. For example, for objects that need to be recognized in multiple directions and scales, a combination of rotation + scaling can be used; for color-sensitive tasks, a combination of color transformation + brightness adjustment can be used.

[0049] Optionally, select an initial combination of enhancement strategies for different categories of images. For example, for the animal category, use a combination of color transformation + brightness adjustment to increase color diversity. For the vehicle category, use a combination of scaling + cropping to cope with images of different scales and perspectives.

[0050] Optionally, if it is found that the model performs poorly in recognizing images of a certain category (e.g., the accuracy rate decreases), reanalyze the data features of the category. According to the new feature analysis results, adjust the combination of enhancement strategies. For example, if the recognition rate of the animal category decreases, the blur operation may be added to simulate the clarity in the actual shooting. Apply the new enhancement strategy to the next round of training data and continue to train the model.

[0051] Step S106, using a target data enhancement strategy to process the initial image to obtain a target image;

[0052] It can be understood that the target data enhancement strategy is applied to process the initial image to generate a new, enhanced image, namely the target image, which is used to improve the image quality and make it easier to be processed by the recognition model. Through data enhancement processing, the noise and interference factors in the image are reduced, the clarity and recognition of the image are improved, and higher quality input is provided for the subsequent recognition steps.

[0053] In an optional embodiment, a target data enhancement strategy is adopted to process the initial image to obtain a target image, including: obtaining weather information of the area where the vehicle is located; determining an image adjustment strategy corresponding to the weather information; adopting the target data enhancement strategy to process the initial image to obtain an intermediate image; and adjusting the intermediate image based on the image adjustment strategy to obtain the target image.

[0054] It can be understood that the real-time or recent weather information of the area where the vehicle is currently located is obtained, and the corresponding image adjustment strategy is determined based on the obtained weather information. The target data enhancement strategy is applied to process the initial image. The target data enhancement strategy may include a variety of operations such as rotation, scaling, cropping, adding noise, etc., aiming to improve the generalization ability and robustness of the model. According to the image adjustment strategy corresponding to the previously determined weather information, the intermediate image is further adjusted to maintain the data enhancement effect while simulating or compensating for the actual impact of weather on image quality, so that the model can better adapt to image recognition tasks under different weather conditions.

[0055] By introducing weather information as the basis for image adjustment, the performance of the target model under different weather conditions can be significantly improved, the impact of weather on image quality can be simulated and compensated, and the model can better adapt to image recognition tasks under different weather conditions. The target data enhancement strategy and weather information adjustment strategy are combined to improve its robustness. After double processing, the target image retains important features while reducing the interference and noise caused by weather factors, which helps to improve the recognition accuracy of the model.

[0056] Optionally, the above-mentioned image adjustment strategy can make targeted adjustments to the image according to different weather conditions to simulate or compensate for the impact of weather on image quality. For example: on sunny days, it may be necessary to fine-tune the brightness and contrast of the image to maintain image clarity and color accuracy. On rainy days, it may be necessary to increase the sharpness of the image to reduce the blurring effect caused by rain; at the same time, you can also consider adjusting the color balance to compensate for the scattering effect of rain on light. On foggy days, it may be necessary to apply a defogging algorithm to improve the visibility and clarity of the image. On snowy days, it may be necessary to increase the brightness of the image and adjust the color saturation to highlight the outlines of objects covered by snow.

[0057] Optionally, obtaining weather information can be achieved in a variety of ways, such as calling a weather API interface, querying a local weather database, or using a vehicle-mounted sensor, etc. The weather information obtained should include but not be limited to temperature, humidity, light intensity, whether there is rain, snow, fog, and other weather conditions.

[0058] Step S108, using a target model to identify the target image to obtain a recognition result of the sign, wherein the target model is obtained by training based on an image using multiple data features, and the multiple data features correspond to different candidate data enhancement strategies respectively;

[0059] It can be understood that the target image is recognized using a trained target model, which is trained based on images containing multiple data features, and these data features correspond to different candidate data enhancement strategies. The processing of the target image by the target model may involve steps such as feature extraction and classification, and finally output the recognition result of the sign. Using a well-trained target model, the sign information in the image can be accurately and quickly recognized, providing a reliable basis for subsequent vehicle control.

[0060] Optionally, the target model may include multiple modules, such as: a data feature analysis module, a training status monitoring module, an adaptive enhancement strategy, and an enhancement effect evaluation module. The data feature analysis module is used to analyze the characteristics of the training data, including the size, position distribution, color characteristics, etc. of the target object (i.e., the knowledge card). According to the data characteristics, determine the appropriate enhancement method and enhancement strength. The training status monitoring module is used to monitor the training status of the model, including training loss, accuracy change, etc. According to the training status, dynamically adjust the data enhancement strategy. For example, when the model converges slowly, increase the strength of data enhancement; when the model is overfitting, reduce the enhancement strength. The adaptive enhancement strategy is to design a set of adaptive data enhancement strategies based on data characteristics and training status, including rotation, scaling, translation, color transformation, blurring and other enhancement methods. The enhancement effect evaluation module is used to evaluate the data enhancement effect. Experiments are conducted to compare the impact of different enhancement strategies on model performance and optimize the enhancement parameters.

[0061] Optionally, the target model may be a model of the YOLO (You Only Look Once) series. The YOLO series models have good application effects in the field of target detection due to their real-time performance and high precision, for example, YOLOv5, and four different versions of models (Yolov5s, Yolov5m, Yolov5l, Yolov5x) provided by YOLOv5. The main difference lies in the two parameters of model depth and width, thereby meeting the needs of different application scenarios.

[0062] Optionally, weights may be used to adjust the application probability and strength of each enhancement method, making the enhancement process more flexible and efficient.

[0063] Step S110, determining the driving control strategy of the vehicle based on the recognition result.

[0064] It can be understood that the vehicle's driving control strategy is comprehensively judged and determined based on the recognition results (such as the content and location of the sign) and the current state of the vehicle (such as speed, location, etc.). For example, it includes acceleration, deceleration, steering, etc., to ensure that the vehicle can drive in accordance with traffic rules and safety requirements. Through timely and accurate driving control strategies, the safety and efficiency of vehicle driving can be improved, and the risk of traffic accidents caused by recognition errors or improper control can be reduced.

[0065] In an optional embodiment, based on the recognition result, a driving control strategy of the vehicle is determined, including: based on the speed limit value indicated by the recognition result, determining the driving control strategy as: adjusting the current speed of the vehicle so that the adjusted current speed is less than the speed limit value; and setting the maximum speed of the vehicle to the speed limit value; in response to the recognition result indicating that the speed limit is released, determining the driving control strategy as: releasing the limit on the maximum speed.

[0066] It can be understood that by identifying the current speed limit value, a driving control strategy is determined based on the identified speed limit value to adjust the current speed of the vehicle. For example, the speed adjustment amount can be calculated to ensure that the adjusted speed does not exceed the speed limit value. If the current speed is higher than the speed limit value, the vehicle will slow down; if the current speed is already lower than or equal to the speed limit value, further speed adjustment may not be required, but it will be ensured that the maximum speed does not exceed the speed limit value. Even if the speed increases due to some reasons during subsequent driving, the vehicle will automatically be limited to this new maximum speed to avoid speeding.

[0067] When it is detected that the speed limit condition has been lifted, such as entering a road section without speed limit, or the speed limit sign has changed, the driving control strategy will be adjusted accordingly. The previously set maximum speed limit can be automatically lifted to allow the vehicle to travel at a higher speed.

[0068] Optionally, the above recognition results can be multiple. For example, if "school ahead" (not a specific speed limit value) is recognized, the vehicle can be controlled to execute the speed limit. If a sign indicating that the road ahead is narrowing is recognized, the action can be executed in advance. If pedestrians are recognized ahead, the vehicle can be controlled to stop and avoid. Through multiple recognition results, the processing capability of assisted driving can be improved.

[0069] Through the above step S102, the initial image including the sign is determined; step S104, based on the object features corresponding to the initial image, the target data enhancement strategy is determined; step S106, the target data enhancement strategy is adopted to process the initial image to obtain the target image; step S108, the target model is adopted to recognize the target image to obtain the recognition result of the sign, wherein the target model is obtained by training based on an image using multiple data features, and the multiple data features correspond to different candidate data enhancement strategies; step S110, based on the recognition result, the driving control strategy of the vehicle is determined. The purpose of improving the image recognition accuracy and robustness is achieved, and the technical effect of improving the accuracy of assisted driving control is achieved, thereby solving the technical problems of the quality of image recognition results and reducing the vehicle control accuracy in the related technology.

[0070] Based on the above embodiments and optional embodiments, the present invention proposes an optional implementation mode, which is an intelligent system in the driving process of a vehicle. The system can collect road street scene images in real time, accurately identify traffic signs, and automatically adjust the vehicle speed or perform pedestrian avoidance operations according to the recognition results, so as to improve driving safety and intelligence level.

[0071] Use a high-definition camera to capture the road street view image (i.e., the initial image) in real time to ensure that the image quality is clear and has strong light adaptability. Perform feature extraction on the initial image after preprocessing operations such as denoising and contrast enhancement, and perform feature extraction based on the size, position, color, and other features of the sign.

[0072] The YOLOv5 model (i.e., target model) is used for real-time detection and recognition of traffic signs, and a dynamic adjustment enhancement strategy for data enhancement and model optimization is implemented. The method and intensity of data enhancement are dynamically adjusted based on the characteristics of the training data and performance indicators (such as loss value and accuracy) during model training.

[0073] By monitoring the loss and accuracy changes during training, the enhancement strategy is adjusted in real time to ensure the efficiency and stability of the training process. A variety of enhancement technologies such as rotation, scaling, translation, color transformation, blurring, and noise addition are used to enrich data diversity. According to data characteristics and model requirements, multiple enhancement methods are intelligently combined to achieve more efficient enhancement effects.

[0074] Through dynamic adjustment and diversified enhancement methods, the generalization ability of the model on different data sets and tasks is enhanced, and the risk of overfitting and underfitting is reduced. For the specific application scenario of traffic sign recognition, the enhancement strategy and model parameters are optimized to improve the detection accuracy of the YOLOv5 model.

[0075] In order to adapt to different data sets and tasks, determine the enhancement method with universality and adaptability to ensure that it can be widely used in various target detection tasks and data sets. Allow customization and optimization of enhancement strategies according to the needs of different tasks and data sets, and provide flexible solutions (multiple candidate data enhancement strategies). Through adaptive enhancement methods, maximize the use of existing training data and improve the effectiveness and utilization of data. When there is less training data, through effective data enhancement strategies, effectively expand the amount of data and improve model performance.

[0076] The above YOLOv5 model is used to process and obtain recognition results. Based on the recognition results, the content of traffic signs (such as speed limit, stop, turn, etc.) is analyzed. Instructions are issued to the vehicle control system to adjust the speed or perform avoidance operations. Dynamic adjustment and enhancement of the data set can also be set to further improve the model performance during the application process.

[0077] The above optional implementation method achieves at least the following effects: by comprehensively using dynamically adjusted data enhancement strategies, diversified enhancement technologies, and optimized YOLOv5 models, it aims to build an efficient and accurate traffic sign recognition and response system to improve the driving safety and intelligence level of vehicles.

[0078] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0079] In this embodiment, a driving control device is also provided, which is used to implement the above-mentioned embodiments and preferred implementation modes, and the descriptions that have been made will not be repeated. As used below, the terms "module" and "device" can implement a combination of software and / or hardware for a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0080] According to an embodiment of the present invention, there is also provided an embodiment of a device for implementing a driving control method. Figure 2 is a schematic diagram of a driving control device according to an embodiment of the present invention. Figure 2As shown, the driving control device includes: an image acquisition module 202, a strategy determination module 204, an image processing module 206, a recognition module 208, and a control module 210. The device is described below.

[0081] An image acquisition module 202, used to determine an initial image including a sign;

[0082] A strategy determination module 204, connected to the image acquisition module 202, is used to determine a target data enhancement strategy based on object features corresponding to the initial image;

[0083] The image processing module 206 is connected to the strategy determination module 204 and is used to process the initial image using the target data enhancement strategy to obtain the target image;

[0084] The recognition module 208 is connected to the image processing module 206 and is used to recognize the target image using a target model to obtain a recognition result of the sign, wherein the target model is obtained by training based on an image using multiple data features, and the multiple data features correspond to different candidate data enhancement strategies respectively;

[0085] The control module 210 is connected to the recognition module 208 and is used to determine the driving control strategy of the vehicle based on the recognition result.

[0086] In a driving control device provided by an embodiment of the present invention, an image acquisition module 202 is provided to determine an initial image including a sign; a strategy determination module 204 is connected to the image acquisition module 202 and is used to determine a target data enhancement strategy based on the object features corresponding to the initial image; an image processing module 206 is connected to the strategy determination module 204 and is used to process the initial image using the target data enhancement strategy to obtain a target image; a recognition module 208 is connected to the image processing module 206 and is used to recognize the target image using a target model to obtain a recognition result of the sign, wherein the target model is obtained by training based on an image using multiple data features, and the multiple data features correspond to different candidate data enhancement strategies respectively; and a control module 210 is connected to the recognition module 208 and is used to determine a driving control strategy of the vehicle based on the recognition result. The purpose of improving the image recognition accuracy and robustness is achieved, and the technical effect of improving the accuracy of assisted driving control is achieved, thereby solving the technical problem of the quality of image recognition results and the reduction of vehicle control accuracy in the related art.

[0087] It should be noted that the above modules can be implemented by software or hardware. For example, for the latter, it can be implemented in the following ways: the above modules can be located in the same processor; or the above modules can be located in different processors in any combination.

[0088] It should be noted that the above-mentioned image acquisition module 202, strategy determination module 204, image processing module 206, recognition module 208, and control module 210 correspond to steps S102 to S110 in the embodiment, and the examples and application scenarios implemented by the above-mentioned modules and corresponding steps are the same, but are not limited to the contents disclosed in the above-mentioned embodiment. It should be noted that the above-mentioned modules can be run in a computer terminal as part of the device.

[0089] It should be noted that the optional or preferred implementation of this embodiment can refer to the relevant description in the embodiment, which will not be repeated here.

[0090] The above-mentioned driving control device may also include a processor and a memory. The image acquisition module 202, the strategy determination module 204, the image processing module 206, the recognition module 208, the control module 210, etc. are all stored in the memory as program units, and the processor executes the above-mentioned program units stored in the memory to realize the corresponding functions.

[0091] The processor includes a kernel, which retrieves the corresponding program unit from the memory. There can be one or more kernels. The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one storage chip.

[0092] An embodiment of the present invention provides a non-volatile storage medium on which a program is stored. When the program is executed by a processor, a driving control method is implemented.

[0093] An embodiment of the present invention provides an electronic device, which includes a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, the following steps are implemented: determining an initial image including a sign; determining a target data enhancement strategy based on object features corresponding to the initial image; using the target data enhancement strategy, processing the initial image to obtain a target image; using a target model, recognizing the target image to obtain a recognition result of the sign, wherein the target model is obtained by training based on an image using multiple data features, and the multiple data features correspond to different candidate data enhancement strategies; based on the recognition result, determining a vehicle driving control strategy. The device in this article may be a server, a PC, etc.

[0094] The present invention also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program that is initialized with the following method steps: determining an initial image including a signboard; determining a target data enhancement strategy based on object features corresponding to the initial image; using the target data enhancement strategy to process the initial image to obtain a target image; using a target model to recognize the target image to obtain a recognition result of the signboard, wherein the target model is obtained by training based on images using multiple data features, and the multiple data features correspond to different candidate data enhancement strategies respectively; and determining a vehicle driving control strategy based on the recognition result.

[0095] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0096] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0097] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0098] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0099] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0100] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0101] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0102] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0103] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0104] The above are only embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of the claims of the present invention.

Claims

1. A driving control method, characterized in that: include: Determine the initial image including the sign; Determining a target data enhancement strategy based on object features corresponding to the initial image; Using the target data enhancement strategy, the initial image is processed to obtain a target image; Using a target model to recognize the target image, and obtaining a recognition result of the sign, wherein the target model is obtained by training based on an image using multiple data features, and the multiple data features respectively correspond to different candidate data enhancement strategies; Based on the recognition result, a driving control strategy of the vehicle is determined.

2. The method according to claim 1, characterized in that: The step of determining a target data enhancement strategy based on the object features corresponding to the initial image includes: Extracting features from the initial image to determine the object size, position distribution, and color features of the sign; The target data enhancement strategy is determined from a plurality of candidate data strategies based on the object size, the position distribution, and the color feature.

3. The method according to claim 2, characterized in that The method further comprises: Inputting the images of the multiple data features into the target model for processing, and determining the training states corresponding to the multiple data features respectively; Based on the training states respectively corresponding to the multiple data features, the initial data enhancement strategies respectively corresponding to the multiple data features are updated to obtain the multiple candidate data strategies.

4. The method according to claim 3, characterized in that The initial data enhancement strategy includes multiple processes, the training state includes a training loss value and a model convergence speed, and the initial data enhancement strategies corresponding to the multiple data features are updated based on the training states corresponding to the multiple data features to obtain the multiple candidate data strategies, including: Determining application probabilities corresponding to the multiple processings respectively based on the training loss values ​​corresponding to the multiple data features respectively; Determining application intensities corresponding to the multiple processings respectively based on model convergence speeds corresponding to the multiple data features respectively; The plurality of candidate data strategies are determined according to the application probabilities and application intensities respectively corresponding to the plurality of processes.

5. The method according to claim 4, characterized in that The multiple processes include: rotation processing, scaling processing, translation processing, color conversion processing, blur processing, and noise addition processing.

6. The method according to claim 1, characterized in that The adopting the target data enhancement strategy to process the initial image to obtain the target image includes: Obtaining weather information of the area where the vehicle is located; Determining an image adjustment strategy corresponding to the weather information; Using the target data enhancement strategy, the initial image is processed to obtain an intermediate image; Based on the image adjustment strategy, the intermediate image is adjusted to obtain the target image.

7. The method according to claim 1, characterized in that Determining a driving control strategy of the vehicle based on the recognition result includes: Based on the speed limit value indicated by the recognition result, determining the driving control strategy is: adjusting the current speed of the vehicle so that the adjusted current speed is less than the speed limit value; and setting the maximum speed of the vehicle to the speed limit value; In response to the recognition result indicating that the speed limit is lifted, the driving control strategy is determined to be: lifting the limit on the maximum vehicle speed.

8. A driving control device, characterized in that: include: An image acquisition module, used for determining an initial image including the sign; A strategy determination module, used to determine a target data enhancement strategy based on object features corresponding to the initial image; An image processing module, used to process the initial image using the target data enhancement strategy to obtain a target image; A recognition module, configured to recognize the target image using a target model to obtain a recognition result of the sign, wherein the target model is obtained by training based on an image using multiple data features, and the multiple data features correspond to different candidate data enhancement strategies respectively; The control module is used to determine the driving control strategy of the vehicle based on the recognition result.

9. A non-volatile storage medium, characterized in that: The non-volatile storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor and executing the driving control method described in any one of claims 1 to 7.

10. An electronic device, characterized in that: include: One or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the driving control method described in any one of claims 1 to 7.

Images