Traffic light state perception method and device based on automatic learning between nodes

Abstract

The application discloses a traffic light state perception method and device based on automatic learning between nodes, and belongs to the field of automatic driving. The method comprises the following steps: acquiring time sequence images collected by a vehicle-mounted camera within a preset time period; detecting the light group frame of each group of traffic lights in each frame of image in the time sequence images; judging the state of each detected light group frame to obtain state information of the light group frame, wherein the state information comprises whether the light group frame is on or off, and the color when the light group frame is on; and using the state information of each light group frame in each frame of image in the time sequence images as a node, automatically learning the corresponding multiple nodes, obtaining weights representing the correlation degree between different nodes, and thus obtaining the time sequence state of each light group frame within the preset time period. The application has real-time performance and high accuracy in the aspect of traffic light state perception.

Description

Technical Field

[0001] This application relates to the field of autonomous driving, and in particular to a method and device for traffic light state perception based on automatic learning between nodes. Background Technology

[0002] With the rise of artificial intelligence, autonomous driving has become a hot technology in the automotive industry. Just as humans pay attention to traffic light information while driving, autonomous driving relies even more heavily on traffic light information for vehicles to better interact with the road. Therefore, designing a reliable traffic light perception system is crucial.

[0003] Traditional autonomous driving solutions for traffic light perception rely on panoramic images captured by onboard cameras. This approach has a high false detection rate, poor real-time performance, and cannot accurately match traffic lights with actual lanes, affecting the vehicle's decision-making at the current intersection during autonomous driving. Summary of the Invention

[0004] To address the issues of high false detection rate and poor real-time performance in traditional autonomous driving traffic light perception technologies, this application provides a traffic light state perception method and device based on automatic learning between nodes.

[0005] One technical solution adopted in this application is: providing a method for traffic light status perception based on automatic learning between nodes, which includes:

[0006] Acquire time-series images captured by the vehicle-mounted camera within a preset time period;

[0007] Detect the bounding boxes of each group of traffic lights in each frame of a time-series image;

[0008] The state of each detected light group frame is determined to obtain the state information of the light group frame, including whether the light group frame is on or off, and the color when the light group frame is on; and

[0009] By using the state information of each light group frame in each frame of the time-series image as a node, and the corresponding multiple nodes, the weights representing the degree of correlation between different nodes are automatically learned, thereby obtaining the time-series state of each light group frame within a preset time period.

[0010] Another technical solution adopted in this application is: providing a traffic light status sensing device based on automatic learning between nodes, which includes:

[0011] Module used to acquire time-series images captured by vehicle-mounted cameras within a preset time period;

[0012] This module is used to detect the light group frames of each group of traffic lights in each frame of a time-series image.

[0013] This module is used to determine the state of each detected light group frame and obtain the state information of the light group frame, including whether the light group frame is on or off, and the color when the light group frame is on; and

[0014] This module is used to automatically learn the state information of each light group frame in each frame of a time-series image as a node, and the corresponding multiple nodes to obtain the weights representing the degree of correlation between different nodes, thereby obtaining the time-series state of each light group frame within a preset time period.

[0015] Another technical solution adopted in this application is: providing a method for determining the current intersection driving state in autonomous driving, which includes:

[0016] Acquire time-series images captured by the vehicle-mounted camera within a preset time period;

[0017] Detect the light group frames of each group of traffic lights in each frame of the time sequence image, and mark the light group frame identifier frames at the corresponding positions on the time sequence image;

[0018] The state of each detected light group frame is judged to obtain the state information of the light group frame, which includes whether the light group frame is on or off, and the color when the light group frame is on.

[0019] By using the state information of each light group frame in each frame of the time-series image as a node, and the corresponding multiple nodes, automatic learning is performed to obtain the weights representing the degree of correlation between different nodes, thereby obtaining the time-series state of each light group frame within a preset time period.

[0020] The directional information of the light group frame marker frame matched with the location of the light group frame in the high-precision map is loaded into the status information of the corresponding light group frame marker frame to obtain the current traffic status of the intersection; and

[0021] Decisions are made regarding the driving status of autonomous vehicles based on traffic conditions.

[0022] Another technical solution adopted in this application is to provide a computer-readable storage medium storing computer instructions that are operated to execute the traffic light state perception method based on automatic learning between nodes in Solution 1.

[0023] Another technical solution adopted in this application is: providing a computer device, which includes a processor and a memory, the memory storing computer instructions, which are operated to execute the traffic light state perception method based on automatic learning between nodes in Solution 1.

[0024] By adopting the above technical solution, this application achieves the following technical effects: This application designs a traffic light state perception method and device based on automatic learning between nodes. This method uses the detection results of images within a preset time period as nodes for automatic learning to enable autonomous vehicles to output perception results of traffic lights and make correct decisions when crossing traffic light intersections, exhibiting real-time performance and high accuracy. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 This is a schematic diagram of a specific implementation of a traffic light state perception method based on automatic learning between nodes according to this application;

[0027] Figure 2 This is a schematic diagram of a specific implementation of a traffic light status sensing device based on automatic learning between nodes, as described in this application.

[0028] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0029] The preferred embodiments of this application will now be described in detail with reference to the accompanying drawings, so that the advantages and features of this application can be more easily understood by those skilled in the art, thereby providing a clearer and more definite definition of the scope of protection of this application.

[0030] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.

[0031] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0032] Figure 1 This paper illustrates a specific implementation of a traffic light state perception method based on automatic learning between nodes, as described in this application. Figure 1 In the specific implementation shown, the traffic light state perception method based on automatic learning between nodes includes:

[0033] Step S101: Acquire time-series images captured by the vehicle-mounted camera within a preset time period.

[0034] In this embodiment, the preset time period ranges from 1.5 seconds to 2 seconds. A time-series image is an image acquired sequentially over time. The acquisition frequency, measured in Hertz, is converted to a value in milliseconds per frame or seconds per frame. The number of time-series images acquired within the preset time period is equal to the preset time period divided by the value in milliseconds per frame or seconds per frame.

[0035] In one example of this application, the vehicle-mounted camera's acquisition frequency is 20Hz, meaning it acquires one frame of image every 50 milliseconds. In this case, the vehicle-mounted camera can acquire 30 to 40 frames of images within a preset time period. These images are arranged chronologically according to the order in which they were captured to obtain a time-series image. Each image in the time-series image set is considered a time-series image. During autonomous driving, traffic lights on actual roads often flash, with an on / off cycle of approximately one second. To detect the flashing, the image acquisition time needs to exceed one second. However, excessively long acquisition times lead to significant delays. Therefore, the image acquisition time, i.e., the preset time period, ranges from 1.5 to 2 seconds.

[0036] For example, if the number of frames captured by the vehicle-mounted camera is too small, it may not even capture a complete on / off state of a traffic light. In this case, it is difficult to determine the flashing state of the traffic light; it can only determine that the current traffic light may be red, green, or yellow. For instance, when the light is green, there are two states: constant green and flashing green. If the time-series image is not fully captured, it can only determine whether the current green light is green or off, but it is impossible to obtain the on / off state of the current green light over a period of time.

[0037] It should be noted that this application can also use cameras with other acquisition frequencies, and there are no restrictions on the acquisition frequency.

[0038] exist Figure 1 In the specific implementation shown, the traffic light state perception method based on automatic learning between nodes further includes:

[0039] Step S102: Detect the light group frames of each group of traffic lights in each frame of the time sequence image.

[0040] In this embodiment, a group of traffic lights corresponds to a light group frame. There may be multiple traffic lights in each frame of the time sequence image, or there may be only one, or there may be none. It is necessary to detect the light group frames in each frame of the image to prevent missed detection.

[0041] In an optional embodiment of this application, detecting the light group frames of each group of traffic lights in each frame of a time-series image includes: marking a light group frame identifier frame at a corresponding position on the time-series image to indicate the light group frame.

[0042] In this embodiment, the present application can use a target detection model to detect the light group bounding boxes of each group of traffic lights in each frame of the time-series image. The number of light group bounding box identifiers corresponds to the number of traffic lights present. The size of the light group bounding box identifiers generally needs to strictly match the size of the actual traffic light group bounding box; that is, the light group bounding box identifier must completely enclose the light group bounding box. The light group bounding box identifiers not only facilitate target tracking in subsequent steps but also ensure the classification of the light group bounding boxes.

[0043] exist Figure 1 In the specific implementation shown, the traffic light state perception method based on automatic learning between nodes further includes:

[0044] Step S103: Determine the state of each detected light group frame to obtain the state information of the light group frame, wherein the state information includes whether the light group frame is on or off, and the color when the light group frame is on; and determine the red and green light state of the light group frame to obtain the state information of the light group frame, wherein the state information includes the color information and on / off information corresponding to each light group frame.

[0045] In this implementation, in each frame of the time-series image, it is necessary to determine whether each light in each light group frame is on or off, and if it is on, whether its display state is red, green, or yellow. Determining the color information and on / off information of the light group frame facilitates further learning of the traffic light states by the subsequent model.

[0046] exist Figure 1 In the specific implementation shown, the traffic light state perception method based on automatic learning between nodes further includes:

[0047] Step S104: Using the state information of each light group frame in each frame of the time sequence image as a node, and the corresponding multiple nodes, automatic learning is performed to obtain the weights representing the degree of correlation between different nodes, thereby obtaining the time sequence state of each light group frame within a preset time period.

[0048] In this embodiment, the state information includes the spatial position information of the light group frame and the temporal sequence information. By learning from each other among multiple nodes corresponding to the light group frames in the temporal images, the correlation degree between the light group frames in two adjacent frames is obtained, thereby predicting the temporal state of each group of light group frames, where the temporal state includes constant red, constant yellow, yellow flashing, constant green, green flashing, or off.

[0049] This application can train a time-series state prediction model through automatic learning between different nodes, which can be used to predict the time-series state of the light group frame.

[0050] In an optional embodiment of this application, before using the state information of each light group frame in each frame of the time-series image as a node, and automatically learning the corresponding multiple nodes to obtain weights representing the degree of correlation between different nodes, thereby obtaining the time-series state of each light group frame within a preset time period, the process includes: performing time-series tracking on each detected light group frame to obtain the state vector of each light group frame within a predetermined time period. The state vector is a time-series sequence of the state information of the corresponding light group frame within the predetermined time period, wherein the state information of each light group frame in each frame of the time-series image is the state information of the corresponding light group frame within the state vector.

[0051] In this embodiment, the present application can find the light group frames belonging to the same group of traffic lights in each frame of time sequence image according to the ID corresponding to the light group frame identifier frame, and record the state information of the light group frames in the corresponding time sequence according to the ID, and combine the state information of the light group frames belonging to the same group of traffic lights in each frame of time sequence image into a state vector.

[0052] In one example of this application, different numerical symbols are used to represent different state information, resulting in a state vector represented by numerical symbols. Each state vector corresponds to one or two of the different numbers. Color information and on / off information are represented numerically to facilitate recording the state information of the lamp group frame. Each state information corresponds to one of four different numerical values, and the state vector corresponding to the same lamp group frame corresponds to one or two of these four different numerical values.

[0053] For example, a red light can be represented by 1, a green light by 2, a yellow light by 3, and an off light by 4. If all 12 frames in the time sequence image set contain the same light group frame, then if all the frames in the same light group are red, the state vector is [111111111111]; if the frames in the same light group are flashing green, the state vector is [222444222444].

[0054] In an optional embodiment of this application, the state information of each light group frame in each frame of the time-series image is used as a node, and the corresponding multiple nodes are automatically learned to obtain weights representing the degree of correlation between different nodes, thereby obtaining the time-series state of each light group frame within a preset time period, including: when the weight between a node in a frame image and another node in an adjacent image is greater than a predetermined weight threshold, the light group frames corresponding to the two nodes are determined as the same light group frame.

[0055] In this embodiment, the weights between different nodes are learned, and implicit correlations between time-series images of different frames are made through the weights. This has a strong ability to avoid errors and the amount of data processed is relatively small.

[0056] In one example of this application, assuming there are 35 time-series images arranged chronologically, all state information of the same light group bounding boxes detected in the 35 time-series images is used as different nodes and input into a time-series state prediction model. The time-series state prediction model can automatically learn the degree of correlation between different nodes based on its own attention mechanism. When the weight between two nodes is greater than a predetermined weight threshold, it indicates that the light group bounding boxes in the two time-series images are related. Generally, a weight greater than zero indicates correlation. Each node contains both temporal and spatial information, including the time and position of the light group bounding box. For example, the position of the light group bounding box in the time-series image is spatial information. The weights actually include some target tracking information. For example, between two consecutive time-series images, nodes with closer light group bounding box markers will have higher weights, indicating some implicit correlation between the two time-series images.

[0057] In one optional embodiment of this application, the state information of each light group frame in each frame of the time-series image is used as a node, and the corresponding multiple nodes are automatically learned to obtain weights representing the degree of correlation between different nodes, thereby obtaining the time-series state of each light group frame within a preset time period. The method further includes: obtaining the time-series state of the light group frame based on the time-series and state information of the same light group frame in different frame images.

[0058] In this embodiment, thirty to forty frames of timing images can be collected within a preset time period. The timing state of the same light group frame is obtained by finding the same light group frame in the timing images based on the timing information and status information.

[0059] In one example of this application, if there are three sets of traffic lights guiding vehicles through the current intersection, then after a single acquisition and processing of time-series images, three sets of state vectors can be obtained. These three sets of state vectors can be simultaneously input into the time-series state prediction model, which can process the three sets of state vectors separately at the same time. The state vectors contain a lot of information, including the time information, location information, and ID information of the same light group frame, thus ensuring that the time-series state model does not confuse the category of light group frames when processing multiple sets of state vectors.

[0060] In one example of this application, each state information in the state vector is treated as a node and input into the temporal state prediction model. The node includes the time information and position information of its corresponding light group frame. The temporal state prediction model processes the state vector to obtain the association weight between two adjacent nodes in the state vector. When the association weight is greater than the weight threshold, it is determined that the light group frame corresponding to the next node is associated with the light group frame corresponding to the previous node and is determined to be the same light group frame. The temporal state prediction model predicts the state vector of the same light group frame to obtain the temporal state of the same light group frame.

[0061] In one optional embodiment of this application, the directional information of the light group frame that matches the position of the light group frame in the high-precision map is loaded onto the status information of the light group frame corresponding to the light group frame to obtain the current traffic status of the intersection.

[0062] In this embodiment, based on the position and direction information of different types of traffic light frames at the current intersection in the high-precision map, the calibration parameters of the vehicle-mounted camera are used to match the same traffic light frame in the latest frame of the time-series image, ultimately obtaining the temporal state of different types of traffic light frames at the current intersection under the corresponding time sequence. By combining the traffic light information in the high-precision map about real-world traffic lights with the traffic lights in the time-series image, the states of different types of traffic lights under different time sequences can be obtained, facilitating vehicle decision-making at intersections and improving accuracy.

[0063] This application achieves the perception results of autonomous vehicles for traffic lights by using the detection results of images within a preset time period as nodes for automatic learning, and makes correct decisions when passing through traffic light intersections, with real-time performance and high accuracy.

[0064] Figure 2 This paper illustrates a specific implementation of a traffic light state sensing device based on automatic learning between nodes, as described in this application. Figure 2 In the specific implementation shown, the traffic light status sensing device based on automatic learning between nodes mainly includes:

[0065] Module 201 is a module used to acquire time-series images captured by a vehicle-mounted camera within a preset time period;

[0066] Module 202 is used to detect the light group frames of each group of traffic lights in each frame of a time-series image.

[0067] Module 203 is used to determine the state of each detected light group frame and obtain the state information of the light group frame, wherein the state information includes whether the light group frame is on or off, and the color when the light group frame is on; and

[0068] Module 204 is used to automatically learn the state information of each light group frame in each frame of the time-series image as a node, and the corresponding multiple nodes to obtain the weights representing the degree of correlation between different nodes, thereby obtaining the time-series state of each light group frame within a preset time period.

[0069] In this embodiment, each input state vector of the time-series state model represents complete state information within the time series. The number of state vectors input at one time corresponds to the number of light group frames belonging to the same traffic light in the current time-series image set. This application improves accuracy and real-time performance by automatically learning the correlation between nodes to perceive the state of traffic lights in the time-series images.

[0070] The traffic light state sensing device based on automatic learning between nodes provided in this application can be used to execute the traffic light state sensing method based on automatic learning between nodes described in any of the above embodiments. Its implementation principle and technical effect are similar, and will not be repeated here.

[0071] In one specific embodiment of this application, the functional modules of the traffic light status sensing device based on automatic learning between nodes can be directly in hardware, in software modules executed by a processor, or in a combination of both.

[0072] Software modules may reside in RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, hard disks, removable disks, CD-ROMs, or any other form of storage medium known in this art. An exemplary storage medium is coupled to the processor, enabling the processor to read information from and write information to the storage medium.

[0073] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic, discrete hardware components, or any combination thereof. A general-purpose processor can be a microprocessor, but alternatively, it can be any conventional processor, controller, microcontroller, or state machine. The processor can also be implemented as a combination of computing devices, such as a combination of a DSP and a microprocessor, multiple microprocessors, one or more microprocessors incorporating a DSP core, or any other such configuration. Alternatively, the storage medium can be integrated with the processor. The processor and storage medium can reside in an ASIC. The ASIC can reside in the user terminal. Alternatively, the processor and storage medium can reside as discrete components in the user terminal.

[0074] In another specific embodiment of this application, a method for determining the current intersection driving state in autonomous driving includes:

[0075] Acquire time-series images captured by the vehicle-mounted camera within a preset time period;

[0076] Detect the light group frames of each group of traffic lights in each frame of the time sequence image, and mark the light group frame identifier frames at the corresponding positions on the time sequence image;

[0077] The state of each detected light group frame is judged to obtain the state information of the light group frame, which includes whether the light group frame is on or off, and the color when the light group frame is on.

[0078] By using the state information of each light group frame in each frame of the time-series image as a node, and the corresponding multiple nodes, automatic learning is performed to obtain the weights representing the degree of correlation between different nodes, thereby obtaining the time-series state of each light group frame within a preset time period.

[0079] The directional information of the light group frame marker frame matched with the location of the light group frame in the high-precision map is loaded into the status information of the corresponding light group frame marker frame to obtain the current traffic status of the intersection; and

[0080] Decisions are made regarding the driving status of autonomous vehicles based on traffic conditions.

[0081] In this embodiment, the state of traffic lights in the time-series image is obtained through automatic learning between different nodes. Combined with the traffic light information in the high-precision map, the state of traffic lights at the actual intersection is matched. Based on the vehicle's own planned path, the driving status of the vehicle is judged at the intersection according to the time-series state of the traffic lights, thereby improving the safety of autonomous vehicles.

[0082] In another specific embodiment of this application, a computer-readable storage medium is provided, which stores computer instructions that are operated to perform the traffic light state perception method based on automatic learning between nodes in any embodiment.

[0083] In another specific embodiment of this application, a computer device includes a processor and a memory, the memory storing computer instructions that are operated to perform the traffic light state perception method based on automatic learning between nodes in any embodiment.

[0084] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0085] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0086] The above description is merely an embodiment of this application and does not limit the patent scope of this application. Any equivalent structural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A traffic light state perception method based on automatic learning between nodes, characterized in that, include: Acquire time-series images captured by the vehicle-mounted camera within a preset time period; Detect the light group frames of each group of traffic lights in each frame of the time sequence image; The state of each detected light group frame is determined to obtain the state information of the light group frame, wherein the state information includes whether the light group frame is on or off, and the color when the light group frame is on. as well as Using the state information of each light group frame in each frame of the time-series image as a node, and corresponding to multiple nodes, automatic learning is performed to obtain weights representing the degree of correlation between different nodes, thereby obtaining the temporal state of each light group frame within the preset time period. The process of using the state information of each light group frame in each frame of the time-series image as a node, and corresponding to multiple nodes, to perform automatic learning to obtain weights representing the degree of correlation between different nodes, thereby obtaining the temporal state of each light group frame within the preset time period includes: When the weight between a node in a frame and another node in an adjacent frame is greater than a predetermined weight threshold, the light group frames corresponding to the two nodes are determined to be the same light group frame. The step of using the state information of each light group frame in each frame of the time-series image as a node, and automatically learning the corresponding multiple nodes to obtain weights representing the degree of correlation between different nodes, thereby obtaining the time-series state of each light group frame within the preset time period, further includes: The timing state of the same light group frame is obtained based on the timing and state information in different frame images.

2. The traffic light state perception method based on automatic learning between nodes as described in claim 1, characterized in that, The detection of the light group frames of each group of traffic lights in each frame of the time-series image includes: Mark the lamp group frame identifier box at the corresponding position on the time sequence image to represent the lamp group frame.

3. The traffic light state perception method based on automatic learning between nodes as described in claim 2, characterized in that, Also includes: The directional information of the light group frame identifier and the light group frame whose position is matched in the high-precision map is loaded into the status information of the light group frame corresponding to the light group frame identifier to obtain the current traffic status of the intersection.

4. The traffic light state perception method based on automatic learning between nodes as described in claim 1, characterized in that, Before using the state information of each light group frame in each frame of the time-series image as a node, and automatically learning the corresponding multiple nodes to obtain weights representing the degree of correlation between different nodes, thereby obtaining the time-series state of each light group frame within the preset time period, the process includes: Timing is performed on each detected light group frame to obtain the state vector of each light group frame within a predetermined time period. The state vector is a temporal sequence of the state information of the corresponding light group frame within the predetermined time period. In the time-series image, the state information of each light group frame in each frame is the state information of the corresponding light group frame in the state vector.

5. A traffic light status sensing device based on automatic learning between nodes, characterized in that, include: Module used to acquire time-series images captured by vehicle-mounted cameras within a preset time period; A module used to detect the light group frames of each group of traffic lights in each frame of the time-series image; A module for judging the state of each detected light group frame and obtaining the state information of the light group frame, wherein the state information includes whether the light group frame is on or off, and the color when the light group frame is on. as well as The module is used to automatically learn multiple nodes corresponding to each light group frame in each frame of the time-series image as a node, thereby obtaining weights representing the degree of correlation between different nodes, and thus obtaining the temporal state of each light group frame within the preset time period. Specifically, the step of automatically learning multiple nodes corresponding to each light group frame in each frame of the time-series image as a node, thereby obtaining weights representing the degree of correlation between different nodes, and thus obtaining the temporal state of each light group frame within the preset time period includes: when the weight between a node in one frame and another node in an adjacent image is greater than a predetermined weight threshold, determining the light group frames corresponding to the two nodes as the same light group frame; the step of automatically learning multiple nodes corresponding to each light group frame in each frame of the time-series image as a node, thereby obtaining weights representing the degree of correlation between different nodes, and thus obtaining the temporal state of each light group frame within the preset time period, further includes: obtaining the temporal state of the same light group frame based on the temporal sequence and state information of the same light group frame in different frames.

6. A method for determining the current intersection driving state in autonomous driving, characterized in that, include: Acquire time-series images captured by the vehicle-mounted camera within a preset time period; Detect the light group frame of each group of traffic lights in each frame of the time sequence image, and mark the light group frame identifier frame at the corresponding position on the time sequence image to represent the light group frame. The state of each detected light group frame is determined to obtain the state information of the light group frame, wherein the state information includes whether the light group frame is on or off, and the color when the light group frame is on. Using the state information of each light group frame in each frame of the time-series image as a node, and corresponding to multiple nodes, automatic learning is performed to obtain weights representing the degree of correlation between different nodes, thereby obtaining the temporal state of each light group frame within the preset time period. Specifically, this includes: when the weight between a node in one frame and another node in an adjacent image is greater than a predetermined weight threshold, the light group frames corresponding to the two nodes are identified as the same light group frame; further, this also includes: obtaining the temporal state of the same light group frame based on the temporal sequence and state information of the same light group frame in different frames. The directional information of the light group frame identifier and the light group frame whose location is matched in the high-precision map is loaded into the status information of the light group frame corresponding to the light group frame identifier to obtain the current traffic status of the intersection; and Based on the traffic conditions, decisions are made regarding the driving status of the autonomous vehicles.

7. A computer-readable storage medium storing computer instructions, characterized in that, The computer instructions are operated to execute the traffic light state perception method based on automatic learning between nodes as described in any one of claims 1-4.

8. A computer device comprising a processor and a memory storing computer instructions, wherein the processor operates the computer instructions to perform the traffic light state perception method based on automatic learning between nodes as described in any one of claims 1-4.

Images