Detection Method, Electronic Device, Vehicle, Medium and Product of Traffic Signal Lights

Abstract

The present application discloses a detection method, an electronic device, a vehicle, a computer-readable storage medium, and a computer program product for traffic lights. The method includes: detecting the traffic light light box and the traffic light light position for the acquired vehicle driving environment information to determine the light box detection result and the light position detection result, and determining the target detection result of the traffic light according to the light box detection result and the light position detection result. In this way, the present application can detect the traffic light light box and the traffic light light position for the vehicle driving environment information, and determine the target detection result of the traffic light according to the light box detection result and the light position detection result, so that the detection of the traffic light is realized based on multi-dimensional information, avoiding misdetection and adapting to various types of traffic lights, ensuring the accuracy of the target detection result of the traffic light, and further ensuring the effectiveness of the instruction generated by the vehicle according to the target detection result.

Description

Technical Field

[0001] The present application relates to the technical field of vehicles, and particularly relates to a method for detecting traffic lights, an electronic device, a vehicle, a computer-readable storage medium, and a computer program product. Background Art

[0002] To meet the needs of users, vehicles can be equipped with intelligent driving functions, such as identifying information such as the color and shape of traffic lights to assist vehicle driving. However, in this case, the recognition accuracy of traffic lights in related technologies is relatively limited, thus affecting the execution effect of the vehicle intelligent driving function to a certain extent, and further affecting the vehicle riding experience of users. Summary of the Invention

[0003] The present application provides a method for detecting traffic lights, an electronic device, a vehicle, a computer-readable storage medium, and a computer program product.

[0004] An embodiment of the present application provides a method for detecting traffic lights, including:

[0005] Detect the traffic light light box and the traffic light position for the obtained vehicle driving environment information to determine the light box detection result and the light position detection result;

[0006] Determine the target detection result of the traffic light according to the light box detection result and the light position detection result.

[0007] In this way, in the embodiment of the present application, the traffic light light box and the traffic light position can be detected for the vehicle driving environment information, and the target detection result of the traffic light can be determined according to the light box detection result and the light position detection result. Thus, to a certain extent, the detection of the traffic light is realized based on multi-dimensional information, and misdetection can be avoided to a certain extent and various types of traffic lights can be adapted, so as to ensure the accuracy of the target detection result of the traffic light, and further ensure the effectiveness of the instruction generated by the vehicle according to the target detection result.

[0008] In some embodiments of the present application, the target detection result includes at least one of the light box position, the light position color, and the light position shape.

[0009] In this way, in the embodiment of the present application, at least one of the light box position, the light position color, and the light position shape of the traffic light can be determined through the light box detection result and the light position detection result, so as to complete the detection of the traffic light.

[0010] In some embodiments of the present application, the vehicle driving environment information includes vehicle driving environment images.

[0011] Thus, in the embodiments of the present application, it is possible to detect the traffic signal light box and the traffic signal light position for the driving environment image, thereby ensuring the detection accuracy of the traffic signal light box and the traffic signal light position to a certain extent.

[0012] In some embodiments of the present application, the detection of the traffic signal light box and the traffic signal light position for the obtained vehicle driving environment information, and determining the light box detection result and the light position detection result, includes:

[0013] Using a detection model that has been pre-trained and can be deployed in the vehicle to perform the detection on the vehicle driving environment image, and determining the light box detection result and the light position detection result, wherein the detection model is configured to detect the traffic signal light box and the traffic signal light position for the vehicle driving environment image.

[0014] Thus, in the embodiments of the present application, it is possible to use a detection model that has been pre-trained and can be deployed in the vehicle to detect the vehicle driving environment image, thereby determining the light box detection result and the light position detection result. To a certain extent, it can ensure the efficient determination of the light box detection result and the light position detection result, and further ensure the detection efficiency of the traffic signal lights.

[0015] In some embodiments of the present application, the detection model is configured to:

[0016] Extract features from the vehicle driving environment image to determine the vehicle driving environment image features;

[0017] Perform detection on the vehicle driving environment image features for the traffic signal light box and the traffic signal light position to determine the light box detection result and the light position detection result.

[0018] Thus, in the embodiments of the present application, it is possible to extract features from the vehicle driving environment image based on the detection model to determine the vehicle driving environment image features, and perform detection on the vehicle driving environment image features for the traffic signal light box and the traffic signal light position to determine the light box detection result and the light position detection result. Thereby, the light box detection result and the light position detection result can be determined based on the vehicle driving environment image features, ensuring the reliability of the light box detection result and the light position detection result to a certain extent.

[0019] In some embodiments of the present application, the detection model includes a backbone module and at least two branch modules;

[0020] The backbone module is configured to extract features from the vehicle driving environment image to determine the vehicle driving environment image features;

[0021] At least one of the branch modules is configured to detect the traffic signal light box for the vehicle driving environment image features and determine the light box detection result;

[0022] At least another one of the branch modules is configured to detect the traffic signal light position for the vehicle driving environment image features and determine the light position detection result.

[0023] In this way, in the embodiment of the present application, the detection model can be implemented based on the backbone module and at least two branch modules.

[0024] In some embodiments of the present application, the light box detection result includes at least one of the light box size, the number of light positions in the light box, the light box type, and the light box prediction probability.

[0025] In this way, in the embodiment of the present application, at least one of the light box size, the number of light positions in the light box, the light box type, and the light box prediction probability can be detected according to the vehicle driving environment information, thereby determining the light box detection result.

[0026] In some embodiments of the present application, the light position detection result includes at least one of the light position size, the light position color, the light position shape, and the light position prediction probability.

[0027] In this way, in the embodiment of the present application, at least one of the light position size, the light position shape, and the light position prediction probability can be detected according to the vehicle driving environment information, thereby determining the light position detection result.

[0028] In some embodiments of the present application, the light box detection result includes the light box prediction probabilities of multiple candidate light boxes, the light position detection result includes the light position prediction probabilities of multiple candidate light positions, and determining the target detection result of the traffic signal according to the light box detection result and the light position detection result includes:

[0029] Determining a target candidate light box and a target candidate light position according to the light box prediction probabilities of the multiple candidate light boxes and the light position prediction probabilities of the multiple candidate light positions;

[0030] Determining the target detection result according to the target candidate light box and the target candidate light position.

[0031] In this way, in the embodiment of the present application, a target candidate light box and a target candidate light position can be determined according to the light box prediction probabilities of multiple candidate light boxes and the light position prediction probabilities of multiple candidate light positions, and the target detection result of the traffic signal can be determined according to the target candidate light box and the target candidate light position, thereby determining the target detection result of the traffic signal.

[0032] In some embodiments of the present application, determining the target candidate light box and the target candidate light position according to the light box prediction probabilities of a plurality of the candidate light boxes and the light position prediction probabilities of a plurality of the candidate light positions includes:

[0033] Determining the candidate light boxes with the light box prediction probabilities greater than or equal to a first threshold as the target candidate light boxes, and determining the candidate light positions with the light position prediction probabilities greater than or equal to a second threshold as the target candidate light positions.

[0034] Thus, in the embodiments of the present application, the candidate light boxes with the light box prediction probabilities greater than or equal to the first threshold can be used as the target candidate light boxes, and the candidate light positions with the light position prediction probabilities greater than or equal to the second threshold can be used as the target candidate light positions, thereby ensuring the effectiveness and reliability of the target candidate light positions and the target candidate light boxes.

[0035] In some embodiments of the present application, each light box detection result further includes the light box sizes of a plurality of the candidate light boxes, and each light position detection result further includes the light position sizes of a plurality of the candidate light positions. Determining the target detection result according to the target candidate light boxes and the target candidate light positions includes:

[0036] In the case where there are a plurality of the target candidate light boxes and a plurality of the target candidate light positions, determining the target candidate light positions corresponding to each of the target candidate light boxes according to the ratios of the light box sizes of each of the target candidate light boxes to the light position sizes of each of the target candidate light positions;

[0037] Determining the target detection result of the traffic signal lamp corresponding to each of the target candidate light boxes according to each of the target candidate light boxes and the target candidate light positions corresponding to each of the target candidate light boxes.

[0038] Thus, in the embodiments of the present application, in the case where there are a plurality of the target candidate light boxes and a plurality of the target candidate light positions, determining the target candidate light positions corresponding to each of the target candidate light boxes according to the ratios of the light box sizes of each of the target candidate light boxes to the light position sizes of each of the target candidate light positions, and determining the target detection result of the traffic signal lamp corresponding to each of the target candidate light boxes according to each of the target candidate light boxes and the target candidate light positions corresponding to each of the target candidate light boxes, whereby the accurate detection and recognition of each traffic signal lamp in the driving environment by the vehicle can be ensured.

[0039] In some embodiments of the present application, detecting the traffic signal lamp box and the traffic signal lamp position for the obtained vehicle driving environment information to determine the light box detection result and the light position detection result includes:

[0040] Performing preprocessing on the obtained vehicle driving environment information to determine the processed environment information;

[0041] Perform the detection on the processed environmental information to determine the light box detection result and the light position detection result.

[0042] In this way, in the embodiment of the present application, the obtained vehicle driving environmental information can be preprocessed to determine the processed environmental information, and the detection is performed on the processed environmental information to determine the light box detection result and the light position detection result, so that the light box detection result and the light position detection result can be determined based on the preprocessed vehicle driving environmental information, thereby making the light box detection result and the light position detection result reliable.

[0043] The embodiment of the present application provides an electronic device, including a memory and a processor. A computer program is stored in the memory, and when the computer program is executed by the processor, the above-mentioned traffic signal lamp detection method is implemented.

[0044] The embodiment of the present application provides a vehicle, including the above-mentioned electronic device.

[0045] The embodiment of the present application provides a computer-readable storage medium, and the computer-readable storage medium stores a computer program. When the computer program is executed by one or more processors, the above-mentioned traffic signal lamp detection method is implemented.

[0046] The embodiment of the present application provides a computer program product, including a computer program / instructions. When the computer program / instructions are executed by a processor, the above-mentioned traffic signal lamp detection method is implemented.

[0047] The electronic device, vehicle, computer-readable storage medium, and computer program product provided by the embodiment of the present application can perform the detection on the vehicle driving environmental information for the traffic signal lamp box and the traffic signal lamp position, and determine the target detection result of the traffic signal lamp according to the light box detection result and the light position detection result. Thus, to a certain extent, the detection of the traffic signal lamp is realized based on multi-dimensional information, and to a certain extent, misdetection can be avoided and traffic lights of various types can be adapted, so as to ensure the accuracy of the target detection result of the traffic signal lamp, and further ensure the effectiveness of the instruction generated by the vehicle according to the target detection result.

[0048] The additional aspects and advantages of the embodiment of the present application will be partially given in the following description, partially become obvious from the following description, or be understood through the practice of the embodiment of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The above and / or additional aspects and advantages of the present application will become obvious and easy to understand from the description of the embodiments in conjunction with the following drawings, where:

[0050] Figure 1 One of the flow diagrams of the traffic signal detection method in some embodiments of the present application;

[0051] Figure 2 One of the application scenario diagrams in some embodiments of the present application;

[0052] Figure 3 Two of the application scenario diagrams in some embodiments of the present application;

[0053] Figure 4 Three of the application scenario diagrams in some embodiments of the present application;

[0054] Figure 5 Four of the application scenario diagrams in some embodiments of the present application;

[0055] Figure 6 Five of the application scenario diagrams in some embodiments of the present application;

[0056] Figure 7 Two of the flow diagrams of the traffic signal detection method in some embodiments of the present application;

[0057] Figure 8 Three of the flow diagrams of the traffic signal detection method in some embodiments of the present application;

[0058] Figure 9 Four of the flow diagrams of the traffic signal detection method in some embodiments of the present application;

[0059] Figure 10 Five of the flow diagrams of the traffic signal detection method in some embodiments of the present application;

[0060] Figure 11 Six of the flow diagrams of the traffic signal detection method in some embodiments of the present application. Detailed implementation manners

[0061] The following details the implementation manners of the present application. The examples of the implementation manners are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The implementation manners described below with reference to the accompanying drawings are exemplary only for explaining the implementation manners of the present application and should not be construed as a limitation on the implementation manners of the present application.

[0062] In the fields of autonomous driving and intelligent transportation systems, the recognition and understanding of traffic signals are one of the key factors to ensure traffic safety and smooth traffic flow. Information such as the position and type of traffic signals directly affects the speed and behavior of vehicles. Therefore, accurately detecting information such as the position and type of traffic signals is crucial for the safety of vehicles and users.

[0063] Furthermore, to improve the recognition accuracy of traffic lights, a traffic light recognition scheme based on Deep Learning has been proposed in the related art. It can be understood that significant progress has been made in the field of traffic light recognition with Deep Learning. Most of the schemes are based on deep learning technologies such as Convolutional Neural Networks (CNN) to improve the accuracy and robustness of traffic light detection. It can also be understood that these schemes mostly rely on a large-scale labeled traffic light image dataset to accurately classify the states of traffic lights by training a deep neural network. For example, in a traffic light recognition scheme based on Deep Learning, the angle, depth, and occlusion branches of the target traffic light in the features can be detected and output, and the color, shape, etc. of the traffic light can be determined based on the output results of each branch.

[0064] Furthermore, to achieve real-time object detection, some schemes are implemented based on models such as YOLO (You Only Look Once) and Faster R-CNN (Faster Region-based Convolutional Neural Network) to quickly recognize traffic lights in complex traffic scenarios.

[0065] In addition, some other schemes propose to recognize traffic lights based on multi-modal information, such as based on images and video streams, and effectively monitor the states of dynamic traffic lights through methods such as Recurrent Neural Network (RNN).

[0066] It can be understood that the above-mentioned traffic light recognition schemes based on Deep Learning have played an important role in improving traffic safety and intelligent traffic management, laying a foundation for the development of future intelligent transportation systems.

[0067] However, the traffic light recognition task faces difficulties and challenges in many aspects, mainly concentrated in the following aspects:

[0068] Diverse types: Different regions and countries may adopt different types, shapes, and colors of traffic lights. Therefore, the algorithm needs to be able to adapt to various changes in traffic lights, including red lights, yellow flashing lights, and pedestrian lights, etc.

[0069] Inaccurate labels: To train a deep learning model, a large amount of labeled training data is required. However, the process of labeling traffic lights can be quite cumbersome and time-consuming, especially when the traffic lights are occluded, appear in complex scenarios, or require fine-grained annotation.

[0070] Sample imbalance: The traffic signal data is collected from roads. The acquisition of traffic signal images is not comprehensive, and there are large sample biases. For example, the sample sizes of yellow lights and right-turn red lights are small, which poses certain difficulties for the recognition of these categories.

[0071] Small detection targets: Traffic signals are usually relatively small targets in the image. Therefore, challenges are faced in dealing with the recognition of small targets, including accurately positioning small targets in the image, handling scale changes, and maintaining high accuracy.

[0072] Requirement for real-time performance: In the traffic management system, there is a high requirement for the real-time accuracy of traffic signal states. The algorithm needs to complete rapid inference and decision-making within a short time to ensure smooth and safe traffic.

[0073] Based on the above possible problems, please refer to Figure 1 , the embodiments of the present application provide a method for detecting traffic signals, including:

[0074] 01: Detect the traffic signal light box and the traffic signal light position for the obtained vehicle driving environment information, and determine the light box detection result and the light position detection result;

[0075] 02: Determine the target detection result of the traffic signal according to the light box detection result and the light position detection result.

[0076] The embodiments of the present application also provide an electronic device, which includes a memory and a processor. The method for detecting traffic signals in the embodiments of the present application can be implemented by the electronic device in the embodiments of the present application. Specifically, a computer program is stored in the memory, and the processor is used to detect the traffic signal light box and the traffic signal light position for the obtained vehicle driving environment information, determine the light box detection result and the light position detection result, and is used to determine the target detection result of the traffic signal according to the light box detection result and the light position detection result.

[0077] Specifically, in the embodiments of the present application, a vehicle (or an electronic device) can detect the traffic signal light box and the traffic signal light position for the obtained vehicle driving environment information, so as to determine the detection result of the traffic signal light box and the detection result of the traffic signal light position, that is, the light box detection result and the light position detection result. And, the vehicle (or an electronic device) can determine the target detection result of the traffic signal according to the light box detection result and the light position detection result, such as determining the color, shape, etc. of the traffic signal.

[0078] Thus, in the embodiments of the present application, the vehicle driving environment information can be detected for the traffic signal light box and the traffic signal light position, and the target detection result of the traffic signal can be determined according to the light box detection result and the light position detection result. To a certain extent, the detection of the traffic signal is realized based on multi-dimensional information, which can avoid false detection to a certain extent and adapt to various types of traffic lights, thereby ensuring the accuracy of the target detection result of the traffic signal, and further ensuring the effectiveness of the instruction generated by the vehicle according to the target detection result.

[0079] In one example, the vehicle driving environment information is the information obtained by the vehicle through sensors such as cameras and radars mounted on the vehicle body to perceive the external environment.

[0080] In one example, the traffic signal light box refers to the light box structure installed on a traffic signal (such as a traffic light), which is usually used to accommodate lighting devices such as bulbs or LED lights and provides protection to prevent the bulbs from being damaged or affected by the external environment. Specifically, please refer to Figure 2 、 Figure 3 and Figure 4 together. Figure 2 、 Figure 3 and Figure 4 are all schematic diagrams of application scenarios in some embodiments of the present application. As shown in Figure 2 , the first light box 110 includes three light positions, and these three light positions are arranged in the first light box 110 in sequence along the horizontal direction. As shown in Figure 3 , the second light box 120 includes three light positions, and these three light positions are arranged in the second light box 120 in sequence along the vertical direction. As shown in Figure 4 , the light boxes in the embodiments of the present application include various types, such as three-light-position light boxes, four-light-position light boxes, pedestrian light boxes, multi-light light boxes, and side light boxes.

[0081] In one example, the light position refers to the position or angle of different signal lights. For example, traffic lights usually consist of red, yellow, and green lights, representing stop, warning, and passage respectively, and these lights are installed in different positions so that pedestrians and vehicles can clearly see and comply with traffic rules.

[0082] Specifically, please refer to Figure 2 、 Figure 3 、 Figure 4 and Figure 5 together. Figure 5 is a schematic diagram of an application scenario in some embodiments of the present application. As shown in Figure 2 , Figure 2 shows that the first light box 110 includes three light positions, namely the first light position 111, the second light position 112, and the third light position 113. Figure 3The second light box 120 shown includes three lamp positions, namely, the fourth lamp position 121, the fifth lamp position 122, and the sixth lamp position 123. Figure 4 There are shown a three-lamp light box, a four-lamp light box, and a pedestrian light box including two lamp positions. Figure 5 There are shown the shape types and color types of lamp positions. The color types include three types: red, green, and yellow. The shape types include circular, left-turn arrow-shaped, right-turn arrow-shaped, straight-ahead arrow-shaped, pedestrian-shaped, downward-arrow-shaped indicating passable, X-shaped indicating non-passable, and digit-shaped indicating the remaining time of the signal lamp.

[0083] It can be understood that under different lighting conditions, weather conditions, and viewing angles, the visibility of the traffic signal light box and the traffic signal lamp positions may vary. It can also be understood that the embodiments of the present application detect the traffic signal light box and the traffic signal lamp positions, thereby improving the detection accuracy of traffic signals under different lighting conditions, weather conditions, and viewing angles, and ensuring that the status information such as the color and shape of traffic signals can be accurately recognized in various situations. Moreover, different traffic lights may have different light box designs and lamp position layouts. Detecting the light box and lamp positions separately can make the detection of traffic signals more flexibly adapt to these differences in lighting conditions, weather conditions, and viewing angles, and guarantee the detection accuracy.

[0084] In one example, the target detection result of the traffic signal includes the current traffic signal indicated by the traffic signal, such as allowing straight-ahead, not allowing straight-ahead, allowing left-turn, not allowing left-turn, etc.

[0085] In some embodiments of the present application, the target detection result includes at least one of the light box position, the lamp position color, and the lamp position shape.

[0086] Specifically, in the embodiments of the present application, the vehicle can determine at least one of the light box position, the lamp position color, and the lamp position shape of the traffic signal based on the light box detection result and the lamp position detection result.

[0087] In one example, the light box position is the position where the traffic signal light box is located.

[0088] In one example, the lamp position color is the color of the lamp position in the traffic signal. For Figure 2 example, for the Figure 2 traffic signal shown, the color of the first lamp position 111, the color of the second lamp position 112, and the color of the third lamp position 113 can be determined.

[0089] In one example, the lamp position shapes include circular, left-turn arrow-shaped, right-turn arrow-shaped, straight-ahead arrow-shaped, pedestrian-shaped, downward-arrow-shaped indicating passable, X-shaped indicating non-passable, and digit-shaped indicating the remaining time of the signal lamp.

[0090] Thus, in the embodiments of the present application, at least one of the light box position, the light position color, and the light position shape of the traffic signal can be determined based on the light box detection result and the light position detection result, thereby completing the detection of the traffic signal.

[0091] In some embodiments of the present application, the driving environment information includes driving environment images.

[0092] Specifically, in the embodiments of the present application, the vehicle detects the traffic signals in the current driving road of the vehicle through the images captured by the cameras mounted on the vehicle body, that is, the driving environment images. In other words, the vehicle in the embodiments of the present application can perform image processing on the driving environment image information, thereby realizing the detection of the traffic signal light box and the traffic signal light position in the driving environment image information.

[0093] Thus, in the embodiments of the present application, the driving environment image can be detected for the traffic signal light box and the traffic signal light position, thereby ensuring the detection accuracy of the traffic signal light box and the traffic signal light position to a certain extent.

[0094] In some embodiments of the present application, step 01 includes:

[0095] Detecting the vehicle driving environment image according to a detection model that has been pre-trained and can be deployed in the vehicle to determine the light box detection result and the light position detection result, wherein the detection model is configured to detect the traffic signal light box and the traffic signal light position in the vehicle driving environment image.

[0096] The processor in the embodiments of the present application is further configured to detect the vehicle driving environment image according to a detection model that has been pre-trained and can be deployed in the vehicle to determine the light box detection result and the light position detection result, wherein the detection model is configured to detect the traffic signal light box and the traffic signal light position in the vehicle driving environment image.

[0097] Specifically, in the embodiments of the present application, the vehicle can detect the driving environment image based on a detection model that has been pre-trained and can be deployed in the vehicle, that is, detect the traffic signal light box and the traffic signal light position in the vehicle driving environment image, and output the light box detection result and the light position detection result.

[0098] In one example, to ensure that the vehicle can efficiently complete the detection of traffic signals, and to realize the real-time detection and recognition of traffic light states, and improve the real-time response speed and driving efficiency of autonomous vehicles, the detection model is a lightweight model with a small number of parameters, or, after the training of the original detection model is completed, the original detection model is processed through one or more of quantization, pruning, distillation, etc. to obtain the detection model that can be deployed in the vehicle in the embodiments of the present application.

[0099] In one example, to improve the detection speed and optimize the operation efficiency of the model, the detection model is an end-to-end model. In other words, the positioning, segmentation, and classification of traffic lights are integrated in one detection model.

[0100] Thus, in the embodiments of the present application, the detection model that has been pre-trained and can be deployed in a vehicle can be used to detect the image of the vehicle driving environment, so as to determine the detection result of the light box and the detection result of the light position, which can ensure the efficient determination of the detection result of the light box and the detection result of the light position to a certain extent, and further ensure the detection efficiency of traffic lights.

[0101] In some embodiments of the present application, the detection model is configured to:

[0102] Extract features from the image of the vehicle driving environment to determine the features of the image of the vehicle driving environment;

[0103] Detect the traffic light light box and the traffic light light position from the features of the image of the vehicle driving environment to determine the detection result of the light box and the detection result of the light position.

[0104] Specifically, to ensure the efficient determination of the detection result of the light box and the detection result of the light position, in the embodiments of the present application, the detection model can first extract features from the image of the vehicle driving environment to determine the features of the image of the vehicle driving environment. Then, detect the traffic light light box and the traffic light light position from the features of the image of the vehicle driving environment to determine the detection result of the light box and the detection result of the light position.

[0105] In one example, the detection model can perform downsampling on the image of the vehicle driving environment and use the downsampling result of the image of the vehicle driving environment as the features of the image of the vehicle driving environment.

[0106] In one example, the detection model includes a feature extraction module. Furthermore, the detection module can extract features from the image of the vehicle driving environment based on the feature extraction module to determine the features of the image of the vehicle driving environment.

[0107] In one example, the feature extraction module is constructed by convolutional layers.

[0108] Thus, in the embodiments of the present application, the features of the image of the vehicle driving environment can be extracted based on the detection model to determine the features of the image of the vehicle driving environment, and the traffic light light box and the traffic light light position can be detected from the features of the image of the vehicle driving environment to determine the detection result of the light box and the detection result of the light position, so that the detection result of the light box and the detection result of the light position can be determined based on the features of the image of the vehicle driving environment, which ensures the reliability of the detection result of the light box and the detection result of the light position to a certain extent.

[0109] In some embodiments of the present application, the detection model includes a backbone module and at least two branch modules. The backbone module is configured to extract features from the vehicle driving environment image and determine the features of the vehicle driving environment image. At least one branch module is configured to detect the traffic signal light box for the features of the vehicle driving environment image and determine the light box detection result. At least another branch module is configured to detect the traffic signal light position for the features of the vehicle driving environment image and determine the light position detection result.

[0110] Specifically, in the embodiments of the present application, the detection model may include two parts. One is the backbone module for feature extraction, and the other is the branch module for implementing the "detection of traffic signal light boxes and traffic signal light positions".

[0111] In one example, specifically refer to Figure 6 , Figure 6 which is the application scenario diagram in some embodiments of the present application, that is, as Figure 6 shown, the detection model may include a backbone module 210, a first branch module 220, and a second branch module 230. Among them, the backbone module 210 can extract features from the vehicle driving environment image and output the features of the vehicle driving environment image. The first branch module 220 can detect the traffic signal light box for the features of the vehicle driving environment image and output the light box detection result. The second branch module 230 can detect the traffic signal light position for the features of the vehicle driving environment image and output the light position detection result.

[0112] In one example, the backbone module 210 is a convolutional neural network model.

[0113] Thus, in the embodiments of the present application, the detection model can be implemented based on the backbone module and at least two branch modules.

[0114] In some embodiments of the present application, the light box detection result includes at least one of the light box size, the number of light positions in the light box, the light box type, and the light box prediction probability.

[0115] Specifically, in the embodiments of the present application, when detecting the traffic signal light box for the vehicle driving environment information, the vehicle can predict the vehicle driving environment information to predict the "light box that may be the 'traffic signal light box in the real physical space' in the vehicle driving environment information", and output at least one of the light box size, the number of light positions in the light box, the light box type, and the light box prediction probability of the light box.

[0116] In one example, the light box size may be the height and width, such as Figure 2 and Figure 3The height and width of the light box shown are the light box dimensions.

[0117] In one example, the number of light positions in the light box refers to the number of light positions in the light box. For example, Figure 2 the first light box in Figure 3 and the second light box in

[0118] both include 3 light positions. Figure 2 In one example, the light box type refers to the category of the light box, such as a three-light-position light box, a four-light-position light box, a pedestrian light light box, a multi-light light box, and a side light box. Figure 3 the first light box in

[0119] and the second light box in

[0120] are both three-light-position light boxes. Figure 6 In one example, the light box prediction probability can be understood as the confidence level of the vehicle for the light box detection result. The higher the light box prediction probability, the higher the confidence level of the vehicle for the light box prediction result, and vice versa.

[0121] In the embodiments of the present application, at least one of the light box dimensions, the number of light positions in the light box, the light box type, and the light box prediction probability can be detected according to the vehicle driving environment information, thereby determining the light box detection result.

[0122] In some embodiments of the present application, the light position detection result includes at least one of the light position dimensions, the light position color, the light position shape, and the light position prediction probability.

[0123] Specifically, in the embodiments of the present application, when detecting the traffic signal light positions in the vehicle driving environment information, the vehicle can predict the vehicle driving environment information to predict the "light positions in the vehicle driving environment information that may be 'traffic signal light boxes in the real physical space'", and output at least one of the light position dimensions, the light position shape, and the light position prediction probability of the light positions.

[0124] In one example, the light position dimensions can be the height and width. For example, Figure 3 the height and width of the light position shown are the light position dimensions.

[0125] In one example, the light position color can be one of yellow, red, and green.

[0126] In one example, the lamp position prediction probability can be understood as the confidence level of the vehicle in the lamp position detection result. The higher the lamp position prediction probability, the higher the confidence level of the vehicle in the lamp position detection result, and vice versa.

[0127] In an example as Figure 6 shown, the "lamp position score" output by the second branch module 230 can be understood as the above-mentioned lamp position prediction probability.

[0128] Thus, in the embodiment of the present application, at least one of the lamp position size, lamp position shape, and lamp position prediction probability can be detected according to the vehicle driving environment information, thereby determining the lamp position detection result.

[0129] Please refer to Figure 7 , in some embodiments of the present application, the lamp box detection result includes the lamp box prediction probabilities of multiple candidate lamp boxes, and the lamp position detection result includes the lamp position prediction probabilities of multiple candidate lamp positions. According to the lamp box detection result and the lamp position detection result, furthermore, step 02 includes:

[0130] 020: Determine the target candidate lamp box and the target candidate lamp position according to the lamp box prediction probabilities of multiple candidate lamp boxes and the lamp position prediction probabilities of multiple candidate lamp positions;

[0131] 021: Determine the target detection result according to the target candidate lamp box and the target candidate lamp position.

[0132] The processor in the embodiment of the present application is further configured to determine the target candidate lamp box and the target candidate lamp position according to the lamp box prediction probabilities of multiple candidate lamp boxes and the lamp position prediction probabilities of multiple candidate lamp positions, and to determine the target detection result according to the target candidate lamp box and the target candidate lamp position.

[0133] Specifically, in the embodiment of the present application, when the vehicle detects the traffic signal lamp box and the traffic signal lamp position, and outputs the lamp box prediction probabilities of multiple candidate lamp boxes and the lamp position prediction probabilities of multiple candidate lamp positions, then the vehicle can determine the target candidate lamp box and the target candidate lamp position according to the lamp box prediction probabilities of multiple candidate lamp boxes and the lamp position prediction probabilities of multiple candidate lamp positions, and output the target detection result of the traffic signal lamp according to the target candidate lamp box and the target candidate lamp position.

[0134] In one example, a candidate lamp box refers to "a lamp box in the vehicle driving environment information that may be the 'traffic signal lamp box in the real physical space'", and the lamp box prediction probability of the candidate lamp box can be understood as the confidence level of the vehicle in the candidate lamp box. The higher the lamp box prediction probability, the higher the confidence level of the vehicle in the candidate lamp box, and vice versa.

[0135] In one example, a candidate traffic light position refers to "a position in the vehicle driving environment information that may be the position of a traffic light in the real physical space". The traffic light position prediction probability of a candidate traffic light position can be understood as the confidence level of the vehicle in the candidate traffic light position. The higher the traffic light position prediction probability, the higher the confidence level of the vehicle in the candidate traffic light position, and vice versa.

[0136] In one example, the vehicle may determine a candidate light box with a light box prediction probability higher than a threshold as a target candidate light box. For example, a candidate light box with a light box prediction probability higher than 95% is determined as a target candidate light box.

[0137] In one example, the vehicle may determine a candidate traffic light position with a traffic light position prediction probability higher than a threshold as a target candidate traffic light position. For example, a candidate traffic light box with a traffic light position prediction probability higher than 95% is determined as a target candidate traffic light position.

[0138] In this way, in the embodiments of the present application, the target candidate light box and the target candidate traffic light position can be determined according to the traffic light box prediction probabilities of multiple candidate light boxes and the traffic light position prediction probabilities of multiple candidate traffic light positions, and the target detection result can be determined according to the target candidate light box and the target candidate traffic light position, thereby realizing the determination of the target detection result of the traffic light.

[0139] In some embodiments of the present application, step 020 includes:

[0140] Determine a candidate light box with a light box prediction probability greater than or equal to a first threshold as a target candidate light box, and determine a candidate traffic light position with a traffic light position prediction probability greater than or equal to a second threshold as a target candidate traffic light position.

[0141] The processor in the embodiments of the present application is further configured to determine a candidate light box with a light box prediction probability greater than or equal to a first threshold as a target candidate light box, and determine a candidate traffic light position with a traffic light position prediction probability greater than or equal to a second threshold as a target candidate traffic light position.

[0142] Specifically, in the embodiments of the present application, after the vehicle detects the traffic light box and the traffic light position of the traffic light and outputs the traffic light box prediction probabilities of multiple candidate light boxes and the traffic light position prediction probabilities of multiple candidate traffic light positions, the vehicle may use, as target candidate light boxes, the candidate light boxes among the multiple candidate light boxes whose traffic light box prediction probabilities are greater than or equal to a preset first threshold. Similarly, the vehicle may use, as target candidate traffic light positions, the candidate traffic light positions among the multiple candidate traffic light positions whose traffic light position prediction probabilities are greater than or equal to a preset second threshold.

[0143] In one example, the light box detection result includes the number of light positions inside the light box, the light box type, and the light box prediction probability of multiple candidate light boxes. The light position detection result includes the light position shape and the light position prediction probability of multiple candidate light positions. Furthermore, after determining the target candidate light box and the target candidate light position, the vehicle can output the target detection result of the traffic signal according to the number of light positions inside the target candidate light box, the light box type, and in combination with the light position shape and light position color of the target candidate light position.

[0144] For example, when the number of target candidate light boxes is 1, the number of light positions inside the target candidate light box is 3, the light box type of the target candidate light box is the three-light-position type, the number of target candidate light positions is 3, the light position shape and light position color of the first target candidate light position are circular and green respectively, the light position shape and light position color of the second target candidate light position are circular and black (or unactivated color) in sequence, and the light position shape and light position color of the third target candidate light position are circular and black (or unactivated color) respectively, then the target detection result of the traffic signal is: the three-light-position type, the light position shape and light position color of the first target candidate light position are circular and green respectively, the light position shape and light position color of the second target candidate light position are circular and black respectively, and the light position shape and light position color of the second target candidate light position are circular and black respectively.

[0145] It can be understood that when the light position shape and light position color of the first target candidate light position are circular and green respectively, and the light position shape and light position color of the second target candidate light position are circular and black respectively, it indicates that the current driving section allows passage.

[0146] In this way, in the embodiment of the present application, the candidate light box with a light box prediction probability greater than or equal to the first threshold can be used as the target candidate light box, and the candidate light position with a light position prediction probability greater than or equal to the second threshold can be used as the target candidate light position, so as to ensure the effectiveness and reliability of the target candidate light position and the target candidate light box.

[0147] Please refer to Figure 8 , in some embodiments of the present application, each light box detection result further includes the light box size of multiple candidate light boxes, and each light position detection result further includes the light position size of multiple candidate light positions. Furthermore, step 021 above includes:

[0148] 0210: When there are multiple target candidate light boxes and multiple target candidate light positions, determine the target candidate light positions corresponding to each target candidate light box according to the ratio of the light box size of each target candidate light box to the light position size of each target candidate light position;

[0149] 0211: Determine the target detection result of the traffic signal corresponding to each target candidate light box according to each target candidate light box and the target candidate light positions corresponding to each target candidate light box.

[0150] The processor according to the embodiment of the present application is further configured to, when there are multiple target candidate light boxes and multiple target candidate light positions, determine the target candidate light positions corresponding to each target candidate light box according to the ratio of the light box size of each target candidate light box to the light position size of each target candidate light position, and determine the target detection result of the traffic signal light corresponding to each target candidate light box according to each target candidate light box and the target candidate light position corresponding to each target candidate light box.

[0151] Specifically, there may be multiple traffic signal lights deployed on the vehicle driving road at the same time. For example, when the vehicle passes through an intersection, there may be traffic lights with three light positions in the light box for indicating the passage of vehicles at each intersection, and there may also be pedestrian lights for indicating the zebra crossing for pedestrians. Therefore, after the vehicle identifies the traffic signal light positions and traffic signal light boxes according to the driving environment information and determines that there are multiple candidate light positions and multiple candidate light boxes, multiple of these multiple candidate light positions may be determined as target candidate light positions, and multiple of these multiple candidate light boxes may be determined as target candidate light boxes at the same time.

[0152] Furthermore, to ensure that the vehicle can accurately identify the traffic signals indicated by each traffic signal light on the road, after the vehicle determines multiple target candidate light boxes and multiple target candidate light positions, the vehicle can determine the attribution relationship between each target candidate light box and each target candidate light position.

[0153] For example, taking Figure 2 and Figure 3 as an example, if the target candidate light positions determined by the vehicle are the first light position 111, the second light position 112, the third light position 113, the fourth light position 121, the fifth light position 122, and the sixth light position 123 in Figure 2 and Figure 3 , and the target candidate light boxes are the first light box 110 and the second light box 120 in Figure 2 and Figure 3 , then the vehicle can determine the light positions among the first light position 111, the second light position 112, the third light position 113, the fourth light position 121, the fifth light position 122, and the sixth light position 123 that belong to or are set in the first light box 110, and determine the light positions among the first light position 111, the second light position 112, the third light position 113, the fourth light position 121, the fifth light position 122, and the sixth light position 123 that belong to or are set in the second light box 120. It can be understood that, as shown in Figure 2 and Figure 3 , the first light position 111, the second light position 112, and the third light position 113 all belong to or are set in the first light box 110, and the fourth light position 121, the fifth light position 122, and the sixth light position 123 all belong to or are set in the second light box 120.

[0154] Based on this, in the embodiments of the present application, after determining multiple target candidate light boxes and multiple target candidate light positions, the vehicle can calculate the ratio between each target candidate light box and each target candidate light position according to the size of each target candidate light box and the size of each target candidate light position, and determine the attribution relationship between each target candidate light box and each target candidate light position according to this ratio, that is, determine the target candidate light position corresponding to the target candidate light box.

[0155] For example, in an example as Figure 3 shown, the target candidate light box is the second light box 120, and the target candidate light positions include the fourth light position 121, the fifth light position 122, and the sixth light position 123. The size of the second light box 120 is the product of the 'light box width' in Figure 3 and the 'light box height' in Figure 3 . The sizes of the fourth light position 121, the fifth light position 122, and the sixth light position 123 are all the product of the 'light position width' in Figure 3 and the 'light position height' in Figure 3 .

[0156] Further, let the product of the 'light box width' in Figure 3 and the 'light box height' in Figure 3 be S1, and let the product of the 'light position width' in Figure 3 and the 'light position height' in Figure 3 be S2. Then the vehicle can determine the attribution relationship between each target candidate light box and each target candidate light position according to the ratio of S1 to S2, that is, determine the target candidate light position corresponding to the target candidate light box.

[0157] In an example, after the vehicle determines the ratio of the size of each target candidate light box to the size of each target candidate light position, the vehicle can determine the corresponding relationship between the target candidate light box and the target candidate light position according to the matrix corresponding to this ratio and the Hungarian algorithm. For example, if the number of target candidate light boxes is N and the number of target candidate light positions is M, then the vehicle can determine the ratio of the size of N target candidate light boxes to the size of M target candidate light positions, that is, N×M ratios, and thus obtain a matrix of size N×M.

[0158] In one example, the light box detection result further includes the number of light positions inside the candidate light box and / or the type of the light box. It can be understood that both the number of light positions inside the target candidate light box and the type of the light box can constrain the "ratio of the size of the target candidate light box to the size of the target candidate light position". For example, if the number of light boxes inside the light box is 3 or the type of the light box is a three-light-position light box, then the ratio of the "size of the target candidate light position set inside the target candidate light box" to the size of the target candidate light box is 1 / 3. Another example is that if the number of light boxes inside the light box is 2 or the type of the light box is a pedestrian light box, then the ratio of the "size of the target candidate light position set inside the target candidate light box" to the size of the target candidate light box is 1 / 2.

[0159] Thus, the vehicle can determine the attribution relationship between each target candidate light box and each target candidate light position based on the number of light positions inside the target candidate light box and / or the type of the light box, the ratio of the size of each target candidate light box to the size of each target candidate light position, and the program code for implementing the Hungarian algorithm, that is, determine the target candidate light position corresponding to the target candidate light box.

[0160] Furthermore, after the vehicle determines the target candidate light position corresponding to each target candidate light box, the vehicle can determine the target detection result of the traffic signal corresponding to each target candidate light box based on each target candidate light box and the target candidate light position corresponding to each target candidate light box, that is, the target detection result of each traffic signal on the vehicle's driving road. For example, when passing through an intersection, it is determined that the color of the pedestrian light corresponding to the pedestrian light box is red, and it is determined that the left-turn light position of the red-green traffic signal corresponding to the three-light-position light box is red and the straight-ahead light position is green. Furthermore, the vehicle can determine that it can go straight at the current moment.

[0161] Please refer to Figure 9 , Figure 9 FIG. 9 is a schematic flowchart of a method for detecting traffic signals in some embodiments of the present application. That is, in one example as shown in FIG. 9, the vehicle can output the target detection result of the traffic signal based on the detection model. Specifically, first, the vehicle inputs the driving environment image into a detection model. The detection model includes a backbone module, a branch module for light box detection, and a branch module for light position detection. Among them, the backbone module is used to extract the features of the driving environment image. The two branch modules perform light box detection and light position detection through the traffic light feature information extracted by the backbone module. Among them, the light box detection result includes the position of the traffic light target, the number of light positions, the traffic light box category, and the light box score. The light position detection result includes the color of the traffic light position, the shape of the light position, and the light position score.

[0162] Then, after screening out the light box detection results and light position detection results with scores higher than the threshold, the area S1 of the traffic light light box is calculated by multiplying the width of the light box by the height of the light box, and the area S2 of the traffic light light position is calculated by multiplying the width of the light position by the height of the light position. The ratio of each light box area S1 to each light position area S2 is calculated to obtain a ratio matrix. After matching the light boxes and light positions through the Hungarian algorithm, the obtained light box position information and light position information are output as the detected traffic light targets that are matched.

[0163] In this way, in the embodiment of the present application, when there are multiple target candidate light boxes and multiple target candidate light positions, according to the ratio of the light position sizes of each target candidate light box to the light position sizes of each target candidate light position, the target candidate light positions corresponding to each target candidate light box are determined, and according to each target candidate light box and the target candidate light positions corresponding to each target candidate light box, the target detection results of the traffic signals corresponding to each target candidate light box are determined, thereby ensuring the accurate detection and recognition of each traffic signal in the driving environment by the vehicle.

[0164] Please refer to Figure 10 , in some embodiments of the present application, step 01 includes:

[0165] 010: Preprocess the obtained vehicle driving environment information to determine the processed environment information;

[0166] 011: Detect the processed environment information to determine the light box detection results and light position detection results.

[0167] The processor in the embodiment of the present application is also used to preprocess the obtained vehicle driving environment information to determine the processed environment information, and to detect the processed environment information to determine the light box detection results and light position detection results.

[0168] Specifically, to ensure the accurate recognition of the light box detection results and light position detection results, in the embodiment of the present application, the vehicle can also preprocess the obtained vehicle driving environment information to improve the quality of the vehicle driving environment information, thereby determining the processed environment information. Then, the vehicle can detect the processed environment information for the traffic signal light box and the traffic signal light position, thereby determining the light box detection results and light position detection results.

[0169] In one example, the vehicle driving environment information includes the vehicle driving environment image. Furthermore, the vehicle can perform operations such as cropping, resizing, and denoising on the vehicle driving environment image, thereby completing the preprocessing of the vehicle driving environment image and improving the quality of the vehicle driving environment image.

[0170] Thus, in the embodiments of the present application, the acquired vehicle driving environment information can be preprocessed to determine the processed environment information, and the processed environment information can be detected to determine the light box detection result and the light position detection result, so that the light box detection result and the light position detection result can be determined based on the preprocessed vehicle driving environment information, thereby ensuring the reliability of the light box detection result and the light position detection result.

[0171] For a clearer illustration of the embodiments of the present application, please refer to Figure 6 , Figure 9 and Figure 11 , Figure 11 which is a schematic flowchart of the traffic signal detection method in some embodiments of the present application. Specifically, in the embodiments of the present application, the vehicle captures image or video data through a camera, that is, the vehicle driving environment image.

[0172] Then, the vehicle driving environment image is preprocessed, including operations such as cropping, resizing, and denoising, to improve the quality of the vehicle driving environment image, thereby improving the efficiency and accuracy of subsequent processing.

[0173] Then, as Figure 6 and Figure 9 shown, the vehicle inputs the preprocessed driving environment image into the detection model. The backbone module in the detection model extracts the features of the driving environment image. The two branch modules in the detection model perform light box detection and light position detection based on the traffic light feature information extracted by the backbone module. Among them, the light box detection result includes the position of the traffic light target, the number of light positions, the traffic light light box category, and the light box score, and the light position detection result includes the traffic light light position color, the light position shape, and the light position score.

[0174] Then, after screening out the light box detection results and light position detection results with scores higher than the threshold, taking the width of the light box multiplied by the height of the light box as the area S1 of the traffic light light box, and the width of the light position multiplied by the height of the light position as the area S2 of the traffic light light position, calculate the area ratio of all light box areas S1 and all light position areas S2 to obtain the area ratio matrix, and after matching the light box and the light position through the Hungarian algorithm for this matrix, output the matched light box position information and light position information as the detected traffic light target that is matched.

[0175] Based on this, in the embodiments of the present application, for the possible problems encountered, through the traffic light data collected on the road, using an end-to-end fine-grained traffic light detection model, detecting the light box of the traffic light and the light position information in the light box through one detection model, screening out the high-confidence traffic light light box and light position targets output by the model, calculating the overlapping area of the light box and the light position and calculating the matching of the light box and the light position, and outputting the matched light box position and light position category information as the output result of the final traffic light detection.

[0176] An embodiment of the present application further provides a vehicle, which includes the above-mentioned electronic device.

[0177] An embodiment of the present application further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by one or more processors, the above-mentioned traffic signal detection method is implemented.

[0178] An embodiment of the present application further provides a computer program product, including a computer program / instructions. When the computer program / instructions are executed by a processor, the above-mentioned traffic signal detection method is implemented.

[0179] In the description of this specification, the descriptions referring to terms such as "specifically", "further", "specially", "understandably", etc. mean that the specific features, structures, materials or characteristics described in connection with the embodiments or examples are included in at least one embodiment or example of the present application. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in a suitable manner in any one or more embodiments or examples. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples.

[0180] Any process or method description shown in the flowchart or described in other ways herein can be understood to represent a module, segment or part of code including one or more executable instructions for implementing a specific logical function or process. The scope of the preferred embodiments of the present application includes additional implementations, where the functions can be executed in a substantially simultaneous manner or in a reverse order according to the functions involved, rather than in the order shown or discussed. This should be understood by those skilled in the art to which the embodiments of the present application belong.

[0181] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present application. Those of ordinary skill in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present application.

Claims

1. A detection method for traffic lights, characterized in that, Including: Detecting the obtained vehicle driving environment information for the traffic signal light box and the traffic signal light position to determine the light box detection result and the light position detection result, where the light box detection result includes the light box prediction probabilities of multiple candidate light boxes, and the light position detection result includes the light position prediction probabilities of multiple candidate light positions; Determining a target candidate light box and a target candidate light position according to the light box prediction probabilities of the multiple candidate light boxes and the light position prediction probabilities of the multiple candidate light positions; Determining a target detection result according to the target candidate light box and the target candidate light position.

2. The method according to claim 1, wherein The target detection result includes at least one of the light box position, the light position color, and the light position shape.

3. The method according to claim 1, wherein The vehicle driving environment information includes a vehicle driving environment image.

4. The method according to claim 3, wherein The detecting the obtained vehicle driving environment information for the traffic signal light box and the traffic signal light position to determine the light box detection result and the light position detection result includes: Detecting the vehicle driving environment image according to a detection model that has been pre-trained and can be deployed in the vehicle to determine the light box detection result and the light position detection result, where the detection model is configured to detect the vehicle driving environment image for the traffic signal light box and the traffic signal light position.

5. The method according to claim 4, characterized in that The detection model is configured as: Extracting features from the vehicle driving environment image to determine vehicle driving environment image features; Detecting the vehicle driving environment image features for the traffic signal light box and the traffic signal light position to determine the light box detection result and the light position detection result.

6. The method according to claim 5, wherein The detection model includes a backbone module and at least two branch modules; The backbone module is configured to extract features from the vehicle driving environment image to determine vehicle driving environment image features; At least one of the branch modules is configured to detect the vehicle driving environment image features for the traffic signal light box to determine the light box detection result; At least another branch module is configured to detect the vehicle driving environment image features for the traffic signal light position to determine the light position detection result.

7. The method according to claim 1, wherein The light box detection result includes at least one of the light box size, the number of light positions in the light box, the light box type, and the light box prediction probability.

8. The method according to claim 1, wherein The light position detection result includes at least one of the light position size, the light position color, the light position shape, and the light position prediction probability.

9. The method according to claim 1, wherein The determining a target candidate light box and a target candidate light position according to the light box prediction probabilities of the multiple candidate light boxes and the light position prediction probabilities of the multiple candidate light positions includes: Determining the candidate light boxes with the light box prediction probabilities greater than or equal to a first threshold as the target candidate light boxes, and determining the candidate light positions with the light position prediction probabilities greater than or equal to a second threshold as the target candidate light positions.

10. The method according to claim 1, characterized in that, Each light box detection result further includes the light box sizes of the multiple candidate light boxes, and each light position detection result further includes the light position sizes of the multiple candidate light positions. The determining the target detection result according to the target candidate light box and the target candidate light position includes: In the case where there are multiple target candidate light boxes and multiple target candidate light positions, determine the target candidate light position corresponding to each target candidate light box according to the ratio of the light box size of each target candidate light box to the light position size of each target candidate light position; According to each target candidate light box and the target candidate light position corresponding to each target candidate light box, determine the target detection result of the traffic signal corresponding to each target candidate light box.

11. The method according to claim 1, wherein The detecting the acquired vehicle driving environment information for the traffic signal light box and the traffic signal light position to determine the light box detection result and the light position detection result includes: Preprocess the acquired vehicle driving environment information to determine the processed environment information; Perform the detection on the processed environment information to determine the light box detection result and the light position detection result.

12. An electronic device, characterized in that, It includes a memory and a processor. A computer program is stored in the memory. When the computer program is executed by the processor, the method according to any one of claims 1-11 is implemented.

13. A vehicle, characterized in that, The vehicle includes the device according to claim 12.

14. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program. When the computer program is executed by one or more processors, the method according to any one of claims 1-11 is implemented.

15. A computer program product, comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, the method according to any one of claims 1-11 is implemented.

Images