Traffic light total period mining method and device
The method and device use image recognition to convert trajectory data into images, training models to predict traffic light countdowns, addressing user uncertainty and enhancing navigation efficiency.
Patent Information
- Application Number
- CN202411203686.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-29
- Publication Date
- 2024-12-17
AI Technical Summary
Users navigating with GPS often encounter uncertainty about traffic light wait times due to blocked views or lack of countdown information, leading to inefficiencies in urban traffic.
A method and device for mining the total cycle of traffic lights using image recognition models, converting trajectory data into images, and training models to predict countdowns based on vehicle positions and times.
Accurately provides traffic light countdowns, improving navigation efficiency and reducing user anxiety by seamlessly integrating with image processing models for efficient and robust cycle mining.
Smart Images
Figure CN119152709A_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to the field of big data technology, and in particular to the field of map navigation technology. Background Art
[0002] When users use navigation to drive, they will pass by traffic lights. When they encounter a red light, they need to stop and wait. In many cases, users do not know how long they need to wait at the traffic light because the car in front blocks the view or the traffic light in the real world does not display the traffic light countdown information. This requires that the traffic light countdown information be accurately displayed on the navigation page to alleviate the anxiety of users waiting for cars at the traffic lights. At the same time, when the red light is about to turn green, remind the user to get ready to start. This not only saves users' time, but also improves the efficiency of urban road traffic.
[0003] To realize the countdown function of traffic lights, it is necessary to accurately mine the total cycle of traffic lights. At present, the main method of mining the total cycle of traffic lights is to verify the target by manual driving, that is, to obtain accurate traffic light cycle data by manually waiting for a complete round of traffic light cycles. Summary of the invention
[0004] The embodiments of the present disclosure provide a method, device, equipment, storage medium and program product for mining the total cycle of traffic lights.
[0005] In a first aspect, an embodiment of the present disclosure proposes a method for mining a total cycle of a traffic light, comprising: converting target trajectory data into a target image, wherein the target trajectory data includes trajectory data of vehicles passing through a traffic light within a target time period, and the target image represents information of at least one type of trajectory points in the target trajectory data; converting reference trajectory data into a reference image, wherein the reference trajectory data includes trajectory data of vehicles passing through a traffic light within a reference total cycle before the target time period, and the reference image represents information of at least one type of trajectory points in the reference trajectory data; inputting the target image and the reference image into an image recognition model to obtain the total cycle of the traffic light.
[0006] In the second aspect, an embodiment of the present disclosure proposes a vehicle navigation method, including: obtaining the current position and current time of a vehicle that is about to pass a traffic light; based on the current position, current time and the total cycle of the traffic light, predicting the countdown of the traffic light, wherein the total cycle of the traffic light is mined using the method described in the first aspect.
[0007] In a third aspect, an embodiment of the present disclosure proposes a cycle mining model training method, comprising: converting sample reference trajectory data into a sample reference image, wherein the sample reference trajectory data includes trajectory data of vehicles passing through the sample traffic light within a sample total cycle of the sample traffic light, and the sample reference image represents information of at least one type of trajectory points in the sample reference trajectory data; inputting the sample reference image into an initial model to obtain a predicted total cycle of the sample traffic light; calculating the loss based on the predicted total cycle of the traffic light and the sample total cycle; and adjusting the parameters of the initial model based on the loss to obtain a cycle mining model.
[0008] In a fourth aspect, an embodiment of the present disclosure proposes a traffic light total cycle mining device, comprising: a first conversion module, configured to convert target trajectory data into a target image, wherein the target trajectory data includes trajectory data of vehicles passing through the traffic light within a target time period, and the target image represents information of at least one type of trajectory points in the target trajectory data; a second conversion module, configured to convert reference trajectory data into a reference image, wherein the reference trajectory data includes trajectory data of vehicles passing through the traffic light within a reference total cycle before the target time period, and the reference image represents information of at least one type of trajectory points in the reference trajectory data; and a recognition module, configured to input the target image and the reference image into an image recognition model to obtain the total cycle of the traffic light.
[0009] In the fifth aspect, an embodiment of the present disclosure proposes a vehicle navigation device, comprising: an acquisition module, configured to acquire the current position and current time of a vehicle that is about to pass a traffic light; a prediction module, configured to predict the countdown of the traffic light based on the current position, current time and the total cycle of the traffic light, wherein the total cycle of the traffic light is mined using the device described in the fourth aspect.
[0010] In the sixth aspect, an embodiment of the present disclosure proposes a cycle mining model training device, comprising: a conversion module, configured to convert sample reference trajectory data into a sample reference image, wherein the sample reference trajectory data includes trajectory data of vehicles passing through the sample traffic light within the sample total cycle of the sample traffic light, and the sample reference image represents information of at least one type of trajectory points in the sample reference trajectory data; a prediction module, configured to input the sample reference image into an initial model to obtain a predicted total cycle of the sample traffic light; a calculation module, configured to calculate a loss based on the predicted total cycle of the traffic light and the sample total cycle; an adjustment module, configured to adjust the parameters of the initial model based on the loss to obtain a cycle mining model.
[0011] In the seventh aspect, an embodiment of the present disclosure proposes an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method described in the first aspect or the second aspect or the first aspect.
[0012] In an eighth aspect, an embodiment of the present disclosure proposes a non-transitory computer-readable storage medium storing computer instructions, where the computer instructions are used to enable a computer to execute the method described in the first aspect, the second aspect, or the third aspect.
[0013] In a ninth aspect, an embodiment of the present disclosure proposes a computer program product, including a computer program, which implements the method described in the first aspect, the second aspect, or the third aspect when executed by a processor.
[0014] The key or important features of the embodiments of the present disclosure are not intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS
[0015] Other features, purposes and advantages of the present disclosure will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings. The drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure. Among them:
[0016] Figure 1 is a flow chart of an embodiment of a periodic mining model training method according to the present disclosure;
[0017] Figure 2 is a flow chart of an embodiment of a traffic light total cycle mining method according to the present disclosure;
[0018] Figure 3 is a flow chart of another embodiment of a traffic light total cycle mining method according to the present disclosure;
[0019] Figure 4 It is a structural diagram of the image recognition model;
[0020] Figure 5 is a flow chart of an embodiment of a vehicle navigation method according to the present disclosure;
[0021] Figure 6 is a structural schematic diagram of an embodiment of a periodic mining model training device according to the present disclosure;
[0022] Figure 7 is a structural schematic diagram of an embodiment of a traffic light total cycle mining device according to the present disclosure;
[0023] Figure 8 is a structural schematic diagram of an embodiment of a vehicle navigation device according to the present disclosure;
[0024] Fig. 9 It is a block diagram of an electronic device used to implement the traffic light total cycle mining method according to the embodiment of the present disclosure. DETAILED DESCRIPTION
[0025] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.
[0026] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present disclosure may be combined with each other. The present disclosure will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.
[0027] Figure 1 A process 100 of an embodiment of a cycle mining model training method according to the present disclosure is shown. The cycle mining model training method comprises the following steps:
[0028] Step 101: convert sample reference trajectory data into a sample reference image.
[0029] In this embodiment, the execution subject of the periodic mining model training method can convert the sample reference trajectory data into a sample reference image.
[0030] The execution subject of the periodic mining model training method is usually a server. The server can be hardware or software. When the server is hardware, it can be implemented as a distributed server cluster consisting of multiple servers, or it can be implemented as a single server. When the server is software, it can be implemented as multiple software or software modules (for example, used to provide distributed services), or it can be implemented as a single software or software module. No specific limitation is made here.
[0031] An intersection may be an intersection of at least two roads, such as a crossroads, a T-junction, etc. At least one traffic light may be provided at the intersection. Vehicles passing through the intersection may travel according to the instructions of the corresponding traffic light. For example, at a crossroads, traffic lights may be provided on all four sides. A left-turn traffic light and a straight-ahead traffic light may be provided on each side. In some cases, a right-turn traffic light may also be provided at the same time. The total traffic light cycle of different traffic lights at the same intersection is the same. The total traffic light cycle may be the time interval between the last time a red light switched to green and the current time a red light switched to green.
[0032] The sample reference trajectory data may include a large amount of trajectory data of vehicles passing through the sample traffic light within the sample total cycle of the sample traffic light. Generally, the similarity between trajectory data within the same sample total cycle is high, and the similarity between trajectory data within different sample total cycles is low.
[0033] The sample reference image can characterize the information of at least one type of trajectory point in the sample reference trajectory data. By converting the trajectory data into an image, the image processing model can be seamlessly connected, thereby significantly improving the efficiency of periodic mining. For the sample reference trajectory data, the number of at least one type of sample trajectory points in each preset time period in the sample reference trajectory data can be counted, and the number of at least one type of sample trajectory points can be converted into the value of the sample pixel point under at least one channel to generate a sample reference image. Among them, the trajectory point may include but is not limited to: green light pass point, trajectory start point, trajectory stop point and trajectory start stop point. For example, the number of green light pass points, trajectory start points, trajectory stop points and trajectory start stop points per second in the sample reference trajectory data is counted, the number of green light pass points is normalized to the value of the sample pixel point under the first channel, the number of trajectory start points is normalized to the value of the sample pixel point under the second channel, the number of trajectory stop points is normalized to the value of the sample pixel point under the third channel, and the number of trajectory start stop points is normalized to the value of the sample pixel point under the fourth channel.
[0034] Step 102, input the sample reference image into the initial model to obtain the predicted total period of the sample traffic light.
[0035] In this embodiment, the execution entity may input the sample reference image into the initial model to obtain the predicted total cycle of the sample traffic light.
[0036] The initial model can be a parameter-initialized model. Usually, the various parameters of the model (e.g., weight parameters and bias parameters) can be initialized with some different small random numbers. "Small random numbers" are used to ensure that the model will not enter a saturated state due to excessive weights, thereby causing training failure, and "different" is used to ensure that the model can learn normally. The initial model can extract the features of the sample reference image and predict the total cycle of the sample traffic light based on the features of the sample reference image.
[0037] Step 103, calculating the loss based on the predicted total traffic light cycle and the total sample cycle.
[0038] In this embodiment, the execution entity may calculate the loss based on the predicted total traffic light cycle and the total sample cycle.
[0039] Here, you can choose a suitable loss function. Input the predicted traffic light total cycle and the sample total cycle into the loss function to calculate the loss. The loss can be used to characterize the difference between the predicted traffic light total cycle and the sample total cycle. The smaller the difference, the better the effect of the initial model; the larger the difference, the worse the effect of the initial model.
[0040] Step 104, based on the loss, adjust the parameters of the initial model to obtain a period mining model.
[0041] In this embodiment, the execution entity may adjust the parameters of the initial model based on the loss to obtain a period mining model.
[0042] The parameters of the initial model can be continuously updated iteratively during the training process until the loss is small enough and the initial model converges, and then a periodic mining model can be obtained. Usually, BP (Back Propagation) can be used for training, and the GD (Gradient Descent) algorithm can be used to update the parameters of the initial model.
[0043] The disclosed embodiment provides a cycle mining model training method, which can seamlessly connect to the image processing model by converting massive trajectory data into images, and then train a cycle mining model, thereby significantly improving the efficiency of cycle mining.
[0044] Figure 2 A flow chart 200 of an embodiment of a traffic light total cycle mining method according to the present disclosure is shown. The traffic light total cycle mining method comprises the following steps:
[0045] Step 201, converting target trajectory data into a target image.
[0046] In this embodiment, the execution subject of the traffic light total cycle mining method can convert the target trajectory data into a target image.
[0047] The execution subject of the traffic light total cycle mining method is usually a server. The server can be hardware or software. When the server is hardware, it can be implemented as a distributed server cluster consisting of multiple servers, or it can be implemented as a single server. When the server is software, it can be implemented as multiple software or software modules (for example, used to provide distributed services), or it can be implemented as a single software or software module. No specific limitation is made here.
[0048] An intersection may be an intersection of at least two roads, such as a crossroads, a T-junction, etc. At least one traffic light may be provided at the intersection. Vehicles passing through the intersection may travel according to the instructions of the corresponding traffic light. For example, at a crossroads, traffic lights may be provided on all four sides. A left-turn traffic light and a straight-ahead traffic light may be provided on each side. In some cases, a right-turn traffic light may also be provided at the same time. The total traffic light cycle of different traffic lights at the same intersection is the same. The total traffic light cycle may be the time interval between the last time a red light switched to green and the current time a red light switched to green.
[0049] The target trajectory data may include the trajectory data of vehicles passing through the traffic light within the target time period (such as within 1 minute). The target image may characterize the information of at least one type of trajectory point in the target trajectory data. By converting the trajectory data into the form of an image, the image processing model can be seamlessly connected, thereby significantly improving the efficiency of periodic mining. For the target trajectory data, the number of at least one type of trajectory points in each preset time period in the target trajectory data can be counted, and the number of at least one type of trajectory points can be converted into the value of the pixel point under at least one channel to generate a target image. Among them, the trajectory points may include but are not limited to: green light passing points, trajectory starting points, trajectory stopping points, and trajectory starting stopping points. For example, the number of green light passing points, trajectory starting points, trajectory stopping points, and trajectory starting stopping points per second in the target trajectory data is counted, and the number of green light passing points is normalized to the value of the pixel point under the first channel, the number of trajectory starting points is normalized to the value of the pixel point under the second channel, the number of trajectory stopping points is normalized to the value of the pixel point under the third channel, and the number of trajectory starting stopping points is normalized to the value of the pixel point under the fourth channel.
[0050] Step 202: convert the reference trajectory data into a reference image.
[0051] In this embodiment, the execution entity may convert the reference trajectory data into a reference image.
[0052] The reference trajectory data may include trajectory data of vehicles passing through the traffic light within a reference total cycle before the target time period. The reference total cycle may be a value set based on experience. In order to ensure that the reference trajectory data includes trajectory data within a complete cycle of the traffic light, it is necessary to ensure that the reference total cycle is greater than the traffic light total cycle. For example, by taking statistics on the total cycle of traffic lights in real life, it is known that the total cycle of traffic lights generally does not exceed 5 minutes. Therefore, 300 seconds can be set as the reference total cycle.
[0053] It should be noted that the reference image can represent information of at least one type of trajectory point in the reference trajectory data. The reference trajectory data can be converted into an image by referring to the specific implementation method of converting the target trajectory data into an image, which will not be described in detail here.
[0054] Step 203, input the target image and the reference image into the image recognition model to obtain the total cycle of the traffic light.
[0055] In this embodiment, the execution entity may input the target image and the reference image into an image recognition model to obtain the total cycle of the traffic light.
[0056] By converting the trajectory data into images and applying the image recognition model, the total cycle of traffic lights can be quickly mined. In addition, since the target trajectory data may have sparse trajectory points, the trajectory data of the previous time slice can be used to assist cycle mining, thereby improving the accuracy of cycle mining.
[0057] In some embodiments, based on the lane-level trajectory data, the lane-level traffic light total cycle can be determined; based on the lane-level traffic light total cycle, the mined traffic light total cycle can be verified to obtain the confidence of the mined traffic light total cycle. Among them, the lane-level trajectory data can be the trajectory data of the left-turn lane or the straight lane. Based on the lane-level trajectory data, other methods different from the traffic light total cycle mining method provided in the embodiment of the present disclosure can be used to determine the lane-level traffic light total cycle. Among them, other traffic light total cycle mining methods mainly adopt the following schemes:
[0058] First, strategy mining based on large-scale trajectories: select traffic lights with trajectories in the same time period from a large number of driving navigation trajectories. Use the least squares principle to learn the deviations of pairwise switching points, and use the maximum likelihood method to estimate the total cycle of traffic lights.
[0059] Second, track map reproduction: by extracting the vehicle track information at the current moment and a certain period of time before and after, draw the stop light waiting time of each track at the corresponding moment. Comprehensively reproduce all tracks at the current moment, and determine the total cycle of the traffic light at that time based on the track map.
[0060] Third, manual driving verification: accurate traffic light cycle data is obtained by manually waiting for a complete round of traffic light cycles.
[0061] Fourth, distribution statistics: based on all trajectories within the target time period, find the first starting point. Calculate the deviation time of all trajectories and the starting point, and count the distribution information of the deviation time. According to the periodic regular characteristics of the distribution information, the total cycle of the traffic light is determined. The disclosed embodiment provides a method for mining the total cycle of traffic lights, which converts trajectory data into the form of images and can seamlessly connect to the image processing model, thereby significantly improving the efficiency of cycle mining; it can more flexibly adapt to changing traffic conditions, is more robust, and can achieve large-scale batch coverage; based on the image recognition model, there is no need for manual formulation and adjustment of strategies, the subsequent maintenance and update costs are low, and it is easy to achieve large-scale application.
[0062] Figure 3 A process 300 of another embodiment of a traffic light total cycle mining method according to the present disclosure is shown. The traffic light total cycle mining method introduces the actor-critic algorithm in reinforcement learning into the traffic light total cycle mining. The actor-critic algorithm can be an integrated algorithm that combines policy iteration and value iteration. The actor-critic algorithm can be composed of two parts: actor and critic. The critic can be a judgment network. When the current input is the environment state, the critic can evaluate the value of the current state; when the current input is the environment state and the action taken, the critic can evaluate the value of taking the action in the current state. The actor can be a policy network. The actor can take the current state as input and output the probability distribution of the action or the continuous action value, and then the critic network can evaluate the quality of the action, thereby adjusting the strategy. Here, the cycle boundary mining model can be the critic in the actor-critic algorithm. The cycle mining model can be the actor in the actor-critic algorithm. The traffic light total cycle mining method includes the following steps:
[0063] Step 301, converting target trajectory data into a target image.
[0064] In this embodiment, the execution subject of the traffic light total cycle mining method can convert the target trajectory data into a target image.
[0065] The execution subject of the traffic light total cycle mining method is usually a server. The server can be hardware or software. When the server is hardware, it can be implemented as a distributed server cluster consisting of multiple servers, or it can be implemented as a single server. When the server is software, it can be implemented as multiple software or software modules (for example, used to provide distributed services), or it can be implemented as a single software or software module. No specific limitation is made here.
[0066] An intersection may be an intersection of at least two roads, such as a crossroads, a T-junction, etc. At least one traffic light may be provided at the intersection. Vehicles passing through the intersection may travel according to the instructions of the corresponding traffic light. For example, at a crossroads, traffic lights may be provided on all four sides. A left-turn traffic light and a straight-ahead traffic light may be provided on each side. In some cases, a right-turn traffic light may also be provided at the same time. The total traffic light cycle of different traffic lights at the same intersection is the same. The total traffic light cycle may be the time interval between the last time a red light switched to green and the current time a red light switched to green.
[0067] The target trajectory data may include the trajectory data of vehicles passing through the traffic light within the target time period (such as within 1 minute). The target image may characterize the information of at least one type of trajectory point in the target trajectory data. By converting the trajectory data into the form of an image, the image processing model can be seamlessly connected, thereby significantly improving the efficiency of periodic mining. For the target trajectory data, the number of at least one type of trajectory points in each preset time period in the target trajectory data can be counted, and the number of at least one type of trajectory points can be converted into the value of the pixel point under at least one channel to generate a target image. Among them, the trajectory points may include but are not limited to: green light passing points, trajectory starting points, trajectory stopping points, and trajectory starting stopping points. For example, the number of green light passing points, trajectory starting points, trajectory stopping points, and trajectory starting stopping points per second in the target trajectory data is counted, and the number of green light passing points is normalized to the value of the pixel point under the first channel, the number of trajectory starting points is normalized to the value of the pixel point under the second channel, the number of trajectory stopping points is normalized to the value of the pixel point under the third channel, and the number of trajectory starting stopping points is normalized to the value of the pixel point under the fourth channel.
[0068] Step 302: convert the reference trajectory data into a reference image.
[0069] In this embodiment, the execution entity may convert the reference trajectory data into a reference image.
[0070] The reference trajectory data may include trajectory data of vehicles passing through the traffic light within a reference total cycle before the target time period. The reference total cycle may be a value set based on experience. In order to ensure that the reference trajectory data includes trajectory data within a complete cycle of the traffic light, it is necessary to ensure that the reference total cycle is greater than the traffic light total cycle. For example, by taking statistics on the total cycle of traffic lights in real life, it is known that the total cycle of traffic lights generally does not exceed 5 minutes. Therefore, 300 seconds can be set as the reference total cycle.
[0071] It should be noted that the reference image can represent information of at least one type of trajectory point in the reference trajectory data. The reference trajectory data can be converted into an image by referring to the specific implementation method of converting the target trajectory data into an image, which will not be described in detail here.
[0072] Step 303: input the target image and the reference image into a period boundary mining model to determine whether the target image and the reference image belong to the same period.
[0073] In this embodiment, the execution subject may input the target image and the reference image into the period boundary mining model to determine whether the target image and the reference image belong to the same period. If they do not belong to the same period, execute step 304; if they belong to the same period, execute step 305.
[0074] The cycle boundary mining model can be the critic in the actor-critic algorithm. The critic can be used to determine whether the target image and the reference image belong to the same cycle, so as to determine whether the boundary of the reference total cycle needs to be updated, improve the robustness of the model, and make it easier to maintain.
[0075] In some embodiments, the period boundary mining model can calculate the similarity between the target image and the reference image. Since the similarity between the trajectory data within the same total period is high, and the similarity between the trajectory data within different total periods is low, it can be determined whether the target image and the reference image belong to the same period based on the similarity. If the similarity is greater than the preset similarity threshold, it means that the target image and the reference image belong to the same period, and the boundary of the reference total period is updated at this time. If the similarity is not greater than the preset similarity threshold, it means that the target image and the reference image do not belong to the same period, and there is no need to update the boundary of the reference total period.
[0076] Step 304: input the reference image into the cycle mining model to obtain the total cycle of the traffic light.
[0077] In this embodiment, if the target image and the reference image do not belong to the same cycle, there is no need to use the target image to update the reference image. The execution entity may input the reference image into the cycle mining model to obtain the total cycle of the traffic light.
[0078] The cycle mining model can be an actor in the actor-critic algorithm. The actor can mine the total cycle of traffic lights based on the known trajectory data of the target. The cycle mining model can be trained with the sample reference image as input and the total cycle of the sample as the target output. The specific training process can be referred to Figure 1 The embodiments shown will not be described in detail here.
[0079] Step 305: splice the target image and the reference image to generate a new reference image.
[0080] In this embodiment, if the target image and the reference image belong to the same period, the reference image needs to be updated using the target image. The above-mentioned execution entity can splice the target image and the reference image to generate a new reference image, thereby completing the update of the boundary of the total reference period.
[0081] Step 306 , obtaining trajectory data in a time period next to the target time period, and converting it into a new target image.
[0082] In this embodiment, the execution subject may obtain the trajectory data in the next time period of the target time period, convert it into a new target image, and return to continue to execute step 303.
[0083] The disclosed embodiment provides a method for mining the total cycle of traffic lights, which introduces the actor-critic algorithm in reinforcement learning into the total cycle mining of traffic lights. The cycle boundary mining is analogous to the critic, which is used to determine whether the target trajectory data can be combined with the previous trajectory data. The total cycle mining is analogous to the actor, which is used to mine the total cycle according to the currently known trajectory data. Through this mode, the cycle boundary mining and cycle mining can be completely unified into one model.
[0084] Figure 4 Figure 2 shows a schematic diagram of the structure of the image recognition model. Figure 4 As shown, the image recognition model includes convolution kernels 401, 402, 403, a cycle boundary mining model 404 and a cycle mining model 405. The convolution kernel 401 is used to extract the features of the target image. At the same time, the convolution kernel 402 is used to extract the features of the reference image. The features of the target image and the features of the reference image are input into the cycle boundary mining model 404. The cycle boundary mining model 404 can be compared to a critic, which calculates the similarity between the features of the target image and the features of the reference image. If the similarity is greater than a preset similarity threshold, it means that the target image and the reference image belong to the same cycle. At this time, the target image and the reference image are spliced to update the reference image. The features of the updated reference image are extracted using the convolution kernel 403 and input into the cycle mining model 405. The cycle mining model 405 can be compared to an actor, which is used to mine the total cycle of the traffic light. If the similarity is not greater than the preset similarity threshold, it means that the target image and the reference image do not belong to the same cycle. At this time, the features of the reference image are extracted using the convolution kernel 403 and input into the cycle mining model 405. The cycle mining model 405 can be analogized to an actor, and is used to mine the total cycle of traffic lights.
[0085] Figure 5 A process 500 of an embodiment of a vehicle navigation method according to the present disclosure is shown. The vehicle navigation method comprises the following steps:
[0086] Step 501, obtaining the current position and current time of a vehicle that is about to pass the traffic light.
[0087] In this embodiment, the execution subject of the vehicle navigation method can obtain the current position and current time of the vehicle that is about to pass the traffic light.
[0088] The execution subject of the vehicle navigation method is usually a server. The server can be hardware or software. When the server is hardware, it can be implemented as a distributed server cluster consisting of multiple servers, or it can be implemented as a single server. When the server is software, it can be implemented as multiple software or software modules (for example, used to provide distributed services), or it can be implemented as a single software or software module. No specific limitation is made here.
[0089] Usually, the vehicle can travel along the route planned by the navigation. The route usually includes traffic lights corresponding to lanes (such as left-turn lanes and straight lanes). The current position of the vehicle is collected in real time, and the distance between the vehicle and the traffic light is calculated. When the distance between the vehicle and the traffic light is less than the preset distance threshold, it can be determined that the vehicle is about to pass the traffic light. At this time, the current time can be further obtained to assist in predicting the countdown of the traffic light.
[0090] Step 502, predicting the countdown of the traffic light based on the current position, the current time and the total cycle of the traffic light.
[0091] In this embodiment, the execution entity may predict the countdown of the traffic light based on the current location, the current time and the total cycle of the traffic light.
[0092] Usually, the traffic light timing scheme on the same day of different weeks remains basically unchanged. Therefore, the traffic light countdown information on the same day of this week can be predicted by the total traffic light cycle of a certain day last week. The traffic light countdown information can be calculated by combining the total traffic light cycle with the current position and current time of the vehicle.
[0093] It should be noted that the total cycle of the traffic light is based on Figure 2 or Figure 3 The described embodiments are obtained by mining and will not be repeated here.
[0094] Further references Figure 6 As an implementation of the methods shown in the above figures, the present disclosure provides an embodiment of a periodic mining model training device. Figure 1 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.
[0095] like Figure 6 As shown, the cycle mining model training device 600 of this embodiment may include: a conversion module 601, a prediction module 602, a calculation module 603 and an adjustment module 604. The conversion module 601 is configured to convert the sample reference trajectory data into a sample reference image, wherein the sample reference trajectory data includes the trajectory data of the vehicle passing through the sample traffic light within the sample total cycle of the sample traffic light, and the sample reference image represents the information of at least one type of trajectory points in the sample reference trajectory data; the prediction module 602 is configured to input the sample reference image into the initial model to obtain the predicted total cycle of the sample traffic light; the calculation module 603 is configured to calculate the loss based on the predicted total cycle of the traffic light and the sample total cycle; the adjustment module 604 is configured to adjust the parameters of the initial model based on the loss to obtain the cycle mining model.
[0096] In this embodiment, in the period mining model training device 600, the specific processing of the conversion module 601, the prediction module 602, the calculation module 603 and the adjustment module 604 and the technical effects thereof can be referred to respectively. Figure 1 The relevant descriptions of steps 101 - 104 in the corresponding embodiment are not repeated here.
[0097] In some optional implementations of this embodiment, the conversion module 601 is further configured to: count the number of at least one type of sample trajectory points in each preset time period in the sample reference trajectory data; convert the number of at least one type of sample trajectory points into the value of the sample pixel point in at least one channel to generate a sample reference image.
[0098] Further references Figure 7 As an implementation of the methods shown in the above figures, the present disclosure provides an embodiment of a traffic light total cycle mining device. Figure 2 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.
[0099] like Figure 7 As shown, the traffic light total cycle mining device 700 of this embodiment may include: a first conversion module 701, a second conversion module 702 and a recognition module 703. The first conversion module 701 is configured to convert the target trajectory data into a target image, wherein the target trajectory data includes the trajectory data of the vehicle passing the traffic light in the target time period, and the target image represents the information of at least one type of trajectory points in the target trajectory data; the second conversion module 702 is configured to convert the reference trajectory data into a reference image, wherein the reference trajectory data includes the trajectory data of the vehicle passing the traffic light in the reference total cycle before the target time period, and the reference image represents the information of at least one type of trajectory points in the reference trajectory data; the recognition module 703 is configured to input the target image and the reference image into the image recognition model to obtain the total cycle of the traffic light.
[0100] In this embodiment, in the traffic light total cycle mining device 700, the specific processing of the first conversion module 701, the second conversion module 702 and the identification module 703 and the technical effects thereof can be referred to respectively. Figure 2 The relevant descriptions of steps 201 - 203 in the corresponding embodiment are not repeated here.
[0101] In some optional implementations of this embodiment, the first conversion module 701 includes: a statistical submodule, configured to count the number of at least one type of trajectory points in each preset time period in the target trajectory data; a first conversion submodule, configured to convert the number of at least one type of trajectory points into the value of a pixel point in at least one channel to generate a target image.
[0102] In some optional implementations of this embodiment, the trajectory points include: green light pass points, trajectory start points, trajectory stop points and trajectory start stop points; and the first conversion submodule is further configured to: convert the number of green light pass points into pixel values under the first channel; convert the number of trajectory start points into pixel values under the second channel; convert the number of trajectory stop points into pixel values under the third channel; and convert the number of trajectory start stop points into pixel values under the fourth channel.
[0103] In some optional implementations of this embodiment, the image recognition model includes a periodic boundary mining model and a periodic mining model; and the recognition module 703 includes: a determination submodule, configured to input the target image and the reference image into the periodic boundary mining model to determine whether the target image and the reference image belong to the same period; a prediction submodule, configured to input the reference image into the periodic mining model in response to a determination that they do not belong to the same period, to obtain the total period of the traffic light, wherein the periodic mining model is trained with a sample reference image as input and a sample total period as target output.
[0104] In some optional implementations of this embodiment, the recognition module 703 also includes: a splicing submodule, configured to splice the target image and the reference image in response to determining that they belong to the same cycle to generate a new reference image; a second conversion submodule, configured to obtain trajectory data in the next time period of the target time period and convert it into a new target image; and a recognition submodule, configured to input the new target image and the new reference image into an image recognition model to obtain the total cycle of the traffic light.
[0105] In some optional implementations of this embodiment, the determination submodule is further configured to: calculate the similarity between the target image and the reference image; if the similarity is greater than a preset similarity threshold, determine that the target image and the reference image belong to the same period; if the similarity is not greater than the preset similarity threshold, determine that the target image and the reference image do not belong to the same period.
[0106] In some optional implementations of the present embodiment, the traffic light cycle mining device 700 also includes: a determination module, configured to determine the lane-level total traffic light cycle based on the lane-level trajectory data; a verification module, configured to verify the mined total traffic light cycle based on the lane-level total traffic light cycle to obtain the confidence level of the mined total traffic light cycle.
[0107] Further references Figure 8 As an implementation of the methods shown in the above figures, the present disclosure provides an embodiment of a vehicle navigation device, and the device embodiment is Figure 5Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.
[0108] like Figure 8 As shown, the vehicle navigation device 800 of this embodiment may include: an acquisition module 801 and a prediction module 802. The acquisition module 801 is configured to acquire the current position and current time of the vehicle that is about to pass the traffic light; the prediction module 802 is configured to predict the countdown of the traffic light based on the current position, current time and the total cycle of the traffic light, wherein the total cycle of the traffic light is calculated using Figure 7 The device described was excavated.
[0109] In the vehicle navigation device 800, the specific processing of the acquisition module 801 and the prediction module 802 and the technical effects thereof can be referred to in Figure 5 The relevant descriptions of steps 501-502 in the corresponding embodiment are not repeated here. In the technical solution of the present disclosure, the acquisition, storage and application of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good customs.
[0110] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.
[0111] Fig. 9 A schematic block diagram of an example electronic device 900 that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.
[0112] like Fig. 9 As shown, the device 900 includes a computing unit 901, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 902 or a computer program loaded from a storage unit 908 into a random access memory (RAM) 903. In the RAM 903, various programs and data required for the operation of the device 900 can also be stored. The computing unit 901, the ROM 902, and the RAM 903 are connected to each other via a bus 904. An input / output (I / O) interface 905 is also connected to the bus 904.
[0113] A number of components in the device 900 are connected to the I / O interface 905, including: an input unit 906, such as a keyboard, a mouse, etc.; an output unit 907, such as various types of displays, speakers, etc.; a storage unit 908, such as a disk, an optical disk, etc.; and a communication unit 909, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 909 allows the device 900 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.
[0114] The computing unit 901 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 901 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 901 performs the various methods and processes described above, such as the traffic light total cycle mining method. For example, in some embodiments, the traffic light total cycle mining method may be implemented as a computer software program, which is tangibly included in a machine-readable medium, such as a storage unit 908. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 900 via the ROM 902 and / or the communication unit 909. When the computer program is loaded into the RAM 903 and executed by the computing unit 901, one or more steps of the traffic light total cycle mining method described above may be performed. Alternatively, in other embodiments, the computing unit 901 may be configured to execute the traffic light total cycle mining method in any other appropriate manner (for example, by means of firmware).
[0115] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.
[0116] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.
[0117] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
[0118] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).
[0119] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.
[0120] A computer system may include a client and a server. The client and the server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises through computer programs running on respective computers and having a client-server relationship to each other. The server may be a cloud server, a server in a distributed system, or a server combined with a blockchain.
[0121] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions provided by this disclosure can be achieved, and this document does not limit this.
[0122] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.
Claims
1. A method for mining the total cycle of traffic lights, comprising: Converting target trajectory data into a target image, wherein the target trajectory data includes trajectory data of vehicles passing through a traffic light within a target time period, and the target image represents information of at least one type of trajectory points in the target trajectory data; Converting reference trajectory data into a reference image, wherein the reference trajectory data includes trajectory data of vehicles passing through the traffic light within a reference total cycle before the target time period, and the reference image represents information of at least one type of trajectory points in the reference trajectory data; The target image and the reference image are input into an image recognition model to obtain the total cycle of the traffic light.
2. The method according to claim 1, wherein: The step of converting the target trajectory data into a target image comprises: Counting the number of at least one type of trajectory points in each preset time period in the target trajectory data; The number of the at least one type of trajectory points is converted into the value of the pixel points in at least one channel to generate the target image.
3. The method according to claim 2, wherein: The track points include: a green light pass point, a track start point, a track stop point and a track start stop point; and The converting the number of at least one type of trajectory points into a value of a pixel point under at least one channel comprises: Convert the number of green light passing points into the value of pixel points in the first channel; Convert the number of the trajectory starting points into the value of the pixel point in the second channel; Convert the number of parking points of the track into the value of the pixel point under the third channel; The number of the starting and stopping points of the track is converted into the value of the pixel point under the fourth channel.
4. The method according to claim 1, wherein: The image recognition model includes a periodic boundary mining model and a periodic mining model; as well as The step of inputting the target image and the reference image into an image recognition model to obtain a total cycle of the traffic light comprises: The target inputs the target image and the reference image into the period boundary mining model to determine whether the target image and the reference image belong to the same period; In response to determining that they do not belong to the same cycle, the reference image is input into the cycle mining model to obtain the total cycle of the traffic light, wherein the cycle mining model is trained with the sample reference image as input and the sample total cycle as the target output.
5. The method according to claim 4, wherein: The step of inputting the target image and the reference image into an image recognition model to obtain the total cycle of the traffic light further includes: In response to determining that they belong to the same period, splicing the target image and the reference image to generate a new reference image; Acquire trajectory data in a time period next to the target time period and convert it into a new target image; The new target image and the new reference image are input into the image recognition model to obtain the total cycle of the traffic light.
6. The method according to claim 4, wherein: The step of inputting the target image and the reference image into the period boundary mining model to determine whether the target image and the reference image belong to the same period includes: Calculating the similarity between the target image and the reference image; If the similarity is greater than a preset similarity threshold, determining that the target image and the reference image belong to the same period; If the similarity is not greater than the preset similarity threshold, it is determined that the target image and the reference image do not belong to the same period.
7. The method according to any one of claims 1 to 6, wherein: The method further comprises: Determining a total traffic light cycle at the lane level based on the lane-level trajectory data; Based on the lane-level traffic light total cycle, the mined total cycle of the traffic light is verified to obtain the confidence of the mined total cycle of the traffic light.
8. A vehicle navigation method, comprising: Get the current position and time of the vehicle that is about to pass the traffic light; Based on the current position, the current time and the total cycle of the traffic light, the countdown of the traffic light is predicted, wherein the total cycle of the traffic light is mined using the method described in any one of claims 1-7.
9. A cycle mining model training method, comprising: Converting the sample reference trajectory data into a sample reference image, wherein the sample reference trajectory data includes trajectory data of vehicles passing through the sample traffic light within a total sample period of the sample traffic light, and the sample reference image represents information of at least one type of trajectory points in the sample reference trajectory data; Inputting the sample reference image into the initial model to obtain the predicted total period of the sample traffic light; Calculating the loss based on the predicted traffic light total cycle and the sample total cycle; Based on the loss, the parameters of the initial model are adjusted to obtain a cycle mining model.
10. The method according to claim 9, wherein: The converting the sample reference trajectory data into a sample reference image comprises: Counting the number of at least one type of sample trajectory points in each preset time period in the sample reference trajectory data; The number of the at least one type of sample trajectory points is converted into the value of the sample pixel points in at least one channel to generate the sample reference image.
11. A traffic light total cycle mining device, comprising: A first conversion module is configured to convert target trajectory data into a target image, wherein the target trajectory data includes trajectory data of vehicles passing through a traffic light within a target time period, and the target image represents information of at least one type of trajectory points in the target trajectory data; A second conversion module is configured to convert reference trajectory data into a reference image, wherein the reference trajectory data includes trajectory data of vehicles passing through the traffic light within a reference total cycle before the target time period, and the reference image represents information of at least one type of trajectory points in the reference trajectory data; The recognition module is configured to input the target image and the reference image into an image recognition model to obtain the total cycle of the traffic light.
12. The device according to claim 11, wherein The first conversion module comprises: A statistical submodule, configured to count the number of at least one type of trajectory points in each preset time period in the target trajectory data; The first conversion submodule is configured to convert the number of the at least one type of trajectory points into the value of the pixel points in at least one channel to generate the target image.
13. The device according to claim 12, wherein: The track points include a green light pass point, a track start point, a track stop point and a track start stop point; and The first conversion submodule is further configured to: Convert the number of green light passing points into the value of pixel points in the first channel; Convert the number of the trajectory starting points into the value of the pixel point in the second channel; Convert the number of parking points of the track into the value of the pixel point under the third channel; The number of the starting and stopping points of the track is converted into the value of the pixel point under the fourth channel.
14. The device according to claim 11, wherein: The image recognition model includes a periodic boundary mining model and a periodic mining model; as well as The identification module comprises: A determination submodule is configured to input the target image and the reference image into the period boundary mining model to determine whether the target image and the reference image belong to the same period; The prediction submodule is configured to, in response to determining that they do not belong to the same cycle, input the reference image into the cycle mining model to obtain the total cycle of the traffic light, wherein the cycle mining model is trained with the sample reference image as input and the sample total cycle as the target output.
15. The device according to claim 14, wherein: The identification module also includes: a splicing submodule, configured to splice the target image and the reference image to generate a new reference image in response to determining that they belong to the same period; A second conversion submodule is configured to obtain trajectory data in a time period next to the target time period and convert it into a new target image; The recognition submodule is configured to input the new target image and the new reference image into the image recognition model to obtain the total cycle of the traffic light.
16. The device according to claim 14, wherein: The determining submodule is further configured to: Calculating the similarity between the target image and the reference image; If the similarity is greater than a preset similarity threshold, determining that the target image and the reference image belong to the same period; If the similarity is not greater than the preset similarity threshold, it is determined that the target image and the reference image do not belong to the same period.
17. The device according to any one of claims 11 to 16, wherein: The device also includes: A determination module configured to determine a total traffic light cycle at the lane level based on the lane-level trajectory data; The verification module is configured to verify the mined total cycle of the traffic light based on the lane-level total cycle of the traffic light, and obtain the confidence of the mined total cycle of the traffic light.
18. A vehicle navigation device, comprising: An acquisition module is configured to acquire a current position and a current time of a vehicle that is about to pass a traffic light; A prediction module is configured to predict the countdown of the traffic light based on the current position, the current time and the total cycle of the traffic light, wherein the total cycle of the traffic light is mined using the device described in any one of claims 11-17.
19. A periodic mining model training device, comprising: a conversion module configured to convert the sample reference trajectory data into a sample reference image, wherein the sample reference trajectory data includes trajectory data of vehicles passing through the sample traffic light within a sample total cycle of the sample traffic light, and the sample reference image represents information of at least one type of trajectory points in the sample reference trajectory data; A prediction module, configured to input the sample reference image into an initial model to obtain a predicted total period of the sample traffic light; A calculation module, configured to calculate a loss based on the predicted traffic light total cycle and the sample total cycle; The adjustment module is configured to adjust the parameters of the initial model based on the loss to obtain a period mining model.
20. The device according to claim 19, wherein The conversion module is further configured to: Counting the number of at least one type of sample trajectory points in each preset time period in the sample reference trajectory data; The number of the at least one type of sample trajectory points is converted into the value of the sample pixel points in at least one channel to generate the sample reference image.
21. An electronic device, comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 10.
22. A non-transitory computer-readable storage medium storing computer instructions, the computer instructions being used to cause the computer to execute the method of any one of claims 1-10.
23. A computer program product comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 10.