Traffic route prediction method, system and medium for non-arterial roads in urban areas
By using deep learning models in the navigation system to identify and analyze traffic monitoring data, predict congestion on non-superior roads in the city, and adjust navigation routes, the problem that the navigation system is difficult to identify the occupied roads in the urban environment is improved, and navigation is improved Rationality and user experience of the route.
Patent Information
- Application Number
- CN202410209274.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-02-26
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2044-02-26
AI Technical Summary
The existing navigation system is difficult to identify the situation where pedestrians, non-motor vehicles and mobile stall vendors occupying roads in real time on non-main roads in the city, resulting in unreasonable navigation route planning and may guide drivers into congested roads.
Image data of non-main roads is obtained through traffic monitoring equipment, pre-trained deep learning models are used to identify pedestrians and non-motor vehicles, calculate real-time traffic, and predict congestion situations based on historical traffic data, upload congestion data to the cloud database, and adjust based on data Navigate the route to avoid congested sections.
Real-time congestion prediction on non-superior roads in cities has been achieved, the rationality of navigation routes has been improved, traffic congestion has been reduced, and users have more convenient travel.
Smart Images

Figure CN118262509B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of intelligent transportation technology, and in particular to a method, system and medium for predicting traffic routes on non-main roads in urban areas. Background Art
[0002] Currently, navigation systems mainly rely on big data analysis to identify and predict road congestion when providing route planning and navigation functions. They can help drivers choose the best route and avoid traffic congestion, which is very helpful for daily driving.
[0003] Since the navigation system is built based on big data analysis, the data mainly comes from the driving data of vehicles on the road and the road network data. It can analyze the road congestion in real time and provide route suggestions for drivers. However, relying solely on big data analysis is not enough to deal with all road conditions. For non-main roads in the city, pedestrians, non-motor vehicles, mobile vendors, etc. occupy the road, resulting in complex and changeable traffic conditions. Sudden traffic accidents often occur, which easily cause road congestion. In addition, such groups occupying roads cannot upload road condition data in real time, and the navigation system cannot truly reflect the real-time conditions of the above roads. This may cause the navigation system to ignore these congestion points when planning routes, resulting in unreasonable planned routes, thereby guiding drivers into congested roads.
[0004] It can be seen that the big data analysis of the navigation system has greatly facilitated route planning and navigation, but in complex urban environments, drivers still need to remain alert and use their own driving experience and judgment to deal with unforeseen accidents. Therefore, how to effectively plan routes to avoid traffic congestion on complex road conditions on non-main roads in cities has become an urgent problem to be solved. Summary of the invention
[0005] The present invention aims to solve at least one of the technical problems existing in the prior art. To this end, the present invention proposes a traffic route prediction method for non-arterial roads in urban areas, which can predict the congestion of non-arterial roads in urban areas, which is conducive to planning better navigation routes and reducing traffic congestion.
[0006] The present invention also provides a traffic route prediction system for non-main road in urban area, electronic equipment and computer-readable storage medium.
[0007] According to a first aspect of an embodiment of the present invention, a method for predicting a traffic route on a non-arterial road in an urban area comprises:
[0008] Acquire image data of non-arterial road sections through traffic monitoring equipment;
[0009] Processing the image data to identify pedestrian and non-motor vehicle images;
[0010] Analyze the pedestrian and non-motor vehicle images using a pre-trained deep learning model, calculate the real-time flow of pedestrians and non-motor vehicles in the road section, and predict the congestion of the road section in combination with historical traffic data;
[0011] When it is predicted that a traffic congestion event will occur on the road section within the predicted time period, uploading the congestion data to a cloud database, the congestion data including the road section information, the predicted time period and the real-time traffic volume;
[0012] A navigation route for the vehicle is determined based on the congestion data, the current location and the destination of the vehicle.
[0013] The method for predicting traffic routes on non-arterial roads in urban areas according to an embodiment of the present invention has at least the following beneficial effects:
[0014] The method of the embodiment first obtains image data of a non-main road section through a traffic monitoring device, and identifies the image data to obtain images of pedestrians and non-motor vehicles, and then uses a pre-trained deep learning model to analyze the images of pedestrians and non-motor vehicles. The real-time flow of pedestrians and non-motor vehicles in the section can be calculated through the deep learning module, and the congestion of the section is predicted based on the real-time flow and historical traffic data, that is, it is predicted whether the current real-time flow of pedestrians and non-motor vehicles may cause road traffic congestion, so as to predict the road congestion; when it is predicted that a traffic congestion event will occur in the above-mentioned section within the predicted time period, the congestion data is uploaded to the cloud database, and the congestion data includes section information, predicted time period and real-time flow, so that the route planning system can obtain the congestion data from the cloud database, and the navigation route of the vehicle can be determined based on the congestion data, the current position and destination of the vehicle. Since the route planning is combined with the congestion data, the navigation route can avoid the non-main roads in the urban area where there is a risk of traffic congestion, and a better navigation route can be planned, thereby reducing the situation where the navigation route encounters traffic congestion, making it more convenient for users to travel.
[0015] According to some embodiments of the present invention, predicting the congestion condition of the road section in combination with historical traffic data includes:
[0016] The deep learning model acquires the historical traffic data within a collection period, wherein the historical traffic data includes historical congestion flow and a historical time period corresponding to the historical congestion flow;
[0017] Analyze the historical congestion flow and the historical time period according to the deep learning model to obtain a minimum flow threshold;
[0018] The real-time traffic volume is compared with the minimum traffic volume threshold value. When the real-time traffic volume is greater than or equal to the minimum traffic volume threshold value, the time when the traffic congestion event occurs on the road section is predicted according to the real-time time period when the real-time traffic volume occurs.
[0019] According to some embodiments of the present invention, the acquiring of the historical traffic data by the deep learning model within a collection period includes:
[0020] Identify lane lines in an image and vehicles, pedestrians, and non-motor vehicles within the lane lines using the deep learning model;
[0021] When the speed of the vehicles in the lane is lower than the first threshold, and the density of pedestrians and non-motor vehicles is greater than or equal to the second threshold, it is determined to be a historical congestion event;
[0022] The historical congestion flow is calculated according to the density and the duration that the vehicle speed is lower than the first threshold, and the duration is recorded as the historical time period.
[0023] According to some embodiments of the present invention, the method further comprises:
[0024] Using the historical time period to train the deep learning model;
[0025] The predicting of the time when the traffic congestion event occurs on the road section according to the real-time time period of the real-time traffic flow includes:
[0026] When the real-time time period is greater than or equal to a preset time value, the real-time time period is divided into a plurality of time segments;
[0027] Extracting the maximum flow of pedestrians and non-motor vehicles corresponding to each of the time segments;
[0028] The maximum flow corresponding to each time segment is input into the trained deep learning model, the average time from the minimum flow threshold to the maximum flow is calculated, and the prediction result of the occurrence time of the traffic congestion event is obtained based on the average time.
[0029] According to some embodiments of the present invention, identifying the lane lines of the road section and the vehicles, pedestrians and non-motor vehicles located within the lane lines by using the deep learning model includes:
[0030] Analyzing objects within the lane line using the deep learning model to generate corresponding boundary features;
[0031] Associating the boundary features in continuous images to generate a motion trajectory of the object;
[0032] The vehicles, pedestrians and non-motor vehicles within the lane line are identified by matching the boundary features, the motion trajectory and a preset feature library.
[0033] According to some embodiments of the present invention, the method further comprises:
[0034] Collect the image data of the road section in real time, and analyze the image data using a pre-trained deep learning model to identify lane lines and vehicles and vehicle categories located within the lane lines;
[0035] The lane occupancy rate of pedestrians and non-motor vehicles in the road section is calculated according to the vehicles in the lane line, the vehicle category and the real-time traffic flow;
[0036] When the lane occupancy rate is greater than or equal to a third threshold, it is determined that a traffic congestion event currently occurs on the road section and the traffic congestion event is uploaded to the cloud database.
[0037] According to some embodiments of the present invention, determining the navigation route of the vehicle according to the congestion data, the current position and the destination of the vehicle includes:
[0038] Setting a navigation route according to the current location and the destination;
[0039] When the navigation route passes through the road section, the real-time traffic volume is greater than or equal to the historical congestion traffic volume, and the driving time of the vehicle on the road section intersects with the predicted time period, the navigation route is adjusted so that the adjusted navigation route does not pass through the road section.
[0040] According to some embodiments of the present invention, determining the navigation route of the vehicle according to the congestion data, the current position and the destination of the vehicle includes:
[0041] Setting a navigation route according to the current location and the destination;
[0042] When the destination is located within the road section and the driving time of the vehicle on the road section does not intersect with the predicted time period, the initially set navigation route is determined as the navigation route, and a reminder message is sent to the driver, wherein the reminder message includes a predicted result that a traffic congestion event will occur on the road section within the predicted time period.
[0043] According to some embodiments of the present invention, the method further comprises:
[0044] When the navigation function of the vehicle or mobile terminal is turned on, obtaining the driver's historical navigation route;
[0045] Determining the navigation route of the vehicle according to the congestion data, the current position and the destination of the vehicle includes:
[0046] setting an initial navigation route according to the current location and the destination;
[0047] When both the initial navigation route and the historical navigation route pass through the road section, the real-time traffic volume is greater than or equal to the historical congestion traffic volume, and the driving time of the vehicle on the road section intersects with the predicted time period, adjusting the navigation route so that the adjusted navigation route does not pass through the road section, and sending the initial navigation route and the adjusted navigation route to the vehicle or the mobile terminal to obtain an operation instruction of the driver to select the initial navigation route or the adjusted navigation route;
[0048] When the operation instruction is acquired, the route selected by the driver is determined as the current navigation route according to the operation instruction.
[0049] According to the second aspect of the embodiment of the present invention, the urban non-arterial road traffic route prediction system includes:
[0050] A collection module, used to obtain image data of non-main road sections through traffic monitoring equipment;
[0051] A recognition module is used to process the image data to identify pedestrian and non-motor vehicle images; and to analyze the pedestrian and non-motor vehicle images using a pre-trained deep learning model to calculate the real-time flow of pedestrians and non-motor vehicles in the road section;
[0052] A judgment module, used to predict the congestion of the road section by combining the real-time traffic with historical traffic data, and when it is predicted that a traffic congestion event will occur on the road section within the predicted time period, the congestion data is uploaded to a cloud database, wherein the congestion data includes the road section information, the predicted time period and the real-time traffic;
[0053] A path planning module is used to determine a navigation route for the vehicle based on the congestion data, the current location and destination of the vehicle.
[0054] The urban non-arterial road traffic route prediction system according to the embodiment of the present invention has at least the following beneficial effects:
[0055] The acquisition module obtains the image data of the road section through the traffic monitoring equipment, and uses the recognition module to identify the image data to obtain the images of pedestrians and non-motor vehicles, and then uses the pre-trained deep learning model to analyze the pedestrian and non-motor vehicle images. The deep learning module can calculate the real-time flow of pedestrians and non-motor vehicles in the section; then the judgment module is used to predict the congestion of the section according to the real-time flow and historical traffic data, that is, to predict whether the current real-time flow of pedestrians and non-motor vehicles may cause road traffic congestion, so as to predict the road congestion; when it is predicted that the above-mentioned section will have a traffic congestion event within the predicted time period, the congestion data will be uploaded to the cloud database, and the congestion data includes the section information, the predicted time period and the real-time flow; finally, the path planning module determines the navigation route of the vehicle according to the congestion data, the current position and destination of the vehicle. Since the congestion data is added for planning, the navigation route can avoid non-main roads in urban areas with traffic congestion risks, making the navigation route better, reducing the situation of traffic congestion encountered by the navigation route, and making it more convenient for users to travel.
[0056] An electronic device according to a third aspect of an embodiment of the present invention includes:
[0057] at least one processor; and
[0058] a memory communicatively connected to the at least one processor; wherein,
[0059] The memory stores instructions that can be executed by the at least one processor. The instructions are executed by the at least one processor so that the at least one processor can perform the method described in the above embodiment.
[0060] The computer-readable storage medium according to the fourth aspect of the embodiment of the present invention is characterized in that the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the method described in the above embodiment.
[0061] Since the electronic device and the computer-readable storage medium adopt all the technical solutions of the method of the above embodiment, they at least have all the beneficial effects brought by the technical solutions of the above embodiment, which will not be repeated here.
[0062] Other features and advantages of the present invention will be set forth in the description which follows, and in part will be apparent from the description, or may be learned by practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS
[0063] Figure 1 is a flow chart of the steps of a road traffic prediction method according to an embodiment of the present invention;
[0064] Figure 2is a schematic diagram of the principle of a road traffic prediction method according to an embodiment of the present invention;
[0065] Figure 3 is a principle block diagram of a road traffic route prediction method according to an embodiment of the present invention;
[0066] Figure 4 is a flow chart of the steps of predicting the congestion situation of a road section according to an embodiment of the present invention;
[0067] Figure 5 is a flowchart of steps for obtaining historical congestion flow and historical time periods according to an embodiment of the present invention;
[0068] Figure 6 is a flow chart of the steps of predicting the time of occurrence of a traffic congestion event according to an embodiment of the present invention;
[0069] Figure 7 is a flow chart of the steps of a road traffic prediction method according to another embodiment of the present invention;
[0070] Figure 8 is a flow chart of the steps of a route planning method according to an embodiment of the present invention;
[0071] Fig. 9 is a flow chart of steps for determining a navigation route of a vehicle according to an embodiment of the present invention;
[0072] Fig.10 is a flow chart of steps for determining a navigation route of a vehicle according to another embodiment of the present invention;
[0073] Fig.11 is a framework diagram of a road traffic prediction system according to an embodiment of the present invention;
[0074] Fig.12 It is a framework diagram of a route planning system according to an embodiment of the present invention. DETAILED DESCRIPTION
[0075] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.
[0076] In the description of the present invention, the meaning of "above", "below", "exceed", etc. is not inclusive of the number itself, and the meaning of "above", "below", "within", etc. is inclusive of the number itself. If there is a description of "first" or "second", it is only used to distinguish the technical features, and cannot be understood as indicating or implying the relative importance or implicitly indicating the number of the indicated technical features or implicitly indicating the order of the indicated technical features.
[0077] In the description of the present invention, it should be noted that the terms such as setting, installing, connecting, etc. should be understood in a broad sense, and technicians in the relevant technical field can reasonably determine the specific meanings of the above terms in the present invention in combination with the specific content of the technical solution.
[0078] The technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described below are only some embodiments of the present invention, not all embodiments.
[0079] When providing route planning and navigation functions, current navigation systems can identify road congestion in real time by collecting and analyzing massive amounts of data, thereby providing drivers with the best route suggestions. This not only saves the driver's time, but also makes driving smoother and reduces unnecessary troubles.
[0080] Since the navigation system is built based on big data analysis, the data mainly comes from the driving data of vehicles on the road and the road network data. However, relying solely on big data analysis is not enough to deal with all driving situations. For non-main roads in the city, pedestrians, non-motor vehicles, mobile vendors, etc. occupy the road, and sudden traffic accidents often occur, which can easily cause road congestion. In addition, such users who occupy the road cannot upload road condition data in real time, and the navigation system cannot truly reflect the real-time conditions of the above roads. Therefore, congestion analysis cannot be performed during route planning, resulting in unreasonable planned routes and easily directing vehicles to drive into congested roads.
[0081] It can be seen that the big data analysis of the navigation system has greatly facilitated route planning and navigation, but in complex urban environments, drivers still need to remain alert and use their own driving experience and judgment to deal with unforeseen accidents. Therefore, how to effectively plan routes to avoid traffic congestion on complex road conditions on non-main roads in cities has become an urgent problem to be solved.
[0082] An embodiment of the invention provides a non-main road traffic route prediction method, which includes two parts: a road traffic prediction method and a route planning method. The road traffic prediction method can upload congestion data to a cloud database, so that the route planning system can obtain congestion data from the cloud database. The route planning method can plan a better navigation route based on the congestion data, thereby reducing the situation where the navigation route encounters traffic congestion, and can effectively avoid traffic congestion in complex road conditions of non-main roads in cities, making travel more convenient for users.
[0083] refer to Figures 1 to 7 A road traffic prediction method according to an embodiment of the present invention is described, which is applicable to non-arterial roads in urban areas with traffic monitoring equipment. The road traffic prediction method is described below using a specific example.
[0084] Reference Figure 1 As shown, the road traffic prediction method of the embodiment of the present invention includes but is not limited to the following steps:
[0085] Step S100, acquiring image data of a non-main road section through a traffic monitoring device;
[0086] Step S200, processing the image data to identify pedestrian and non-motor vehicle images;
[0087] Step S300, using a pre-trained deep learning model to analyze pedestrian and non-motor vehicle images, calculate the real-time flow of pedestrians and non-motor vehicles in the road section, and predict the congestion of the road section in combination with historical traffic data;
[0088] Step S400, when a traffic congestion event is predicted to occur on a road section within a predicted time period, the congestion data is uploaded to a cloud database, and the congestion data includes road section information, predicted time period and real-time traffic.
[0089] It is understandable that the road traffic prediction method of the embodiment of the present invention is aimed at monitoring sections with complex road traffic, especially non-main roads and internal roads in urban areas. Such roads are characterized by intersections, pedestrians and non-motor vehicles, and even mobile vendors occupying the roads. Non-motor vehicles include bicycles, motorcycles, and power-assisted bicycles. When the flow of pedestrians and non-motor vehicles is large, the driving speed of the vehicle will decrease, which is easy to cause traffic congestion. Therefore, when monitoring traffic congestion on the above roads, it is necessary to first determine the flow of pedestrians and non-motor vehicles on the road.
[0090] Combination Figure 2 and Figure 3 It can be understood that in step S100, the image data of the non-main road section is first obtained through the traffic monitoring equipment of the road, for example, the pictures are taken by the cameras deployed on the road, the number of cameras is not limited to one, and multiple cameras on the road can also be used to continuously shoot the road section, so as to monitor the road conditions more accurately. Then, steps S200 and S300 are executed, and the collected pictures are analyzed and processed by the pre-trained deep learning model to identify the images of pedestrians and non-motor vehicles in the pictures, and the deep learning model can be used to calculate the flow of pedestrians and non-motor vehicles based on the above images.
[0091] Taking a specific example to illustrate, the input image is preprocessed, including resizing, normalization and other operations, so as to facilitate input into the deep learning model. The deep learning model used can be a convolutional neural network (CNN) or a pre-trained model, such as a visual geometry group network (VGG-16), a deep residual network (ResNet), etc. If a pre-trained model is used, the model needs to be fine-tuned to better adapt to specific tasks. For example, prepare a labeled data set containing images of pedestrians and non-motor vehicles, such as ImageNet. Use the prepared data set to train the model and adjust the parameters of the model to improve the accuracy of the model. The trained deep learning model can improve the accuracy of recognition, especially for complex tasks such as image processing. When using the trained model for calculation, the image to be detected is input into the trained model, and the recognition result can be output. For example, the number of pedestrians and non-motor vehicles identified in the image obtained within the sampling period, and the flow of pedestrians and non-motor vehicles can be calculated based on the time period and number.
[0092] Reference Figure 3 As shown, after obtaining the real-time traffic of pedestrians and non-motor vehicles, the congestion of the road section is predicted based on the real-time traffic and historical traffic data. This step is performed through a deep learning model, that is, the deep learning model can compare the real-time traffic with the historical traffic data. According to the comparison results, it can be predicted whether the current real-time traffic of pedestrians and non-motor vehicles may cause road traffic congestion, thereby realizing the prediction of road congestion. Among them, the historical traffic data can be understood as the traffic data of pedestrians and non-motor vehicles in historical traffic. For example, if the real-time traffic is greater than the preset traffic in historical traffic, it can be judged that the road section will have traffic congestion, that is, it is predicted that traffic congestion will occur in a certain time period in the future.
[0093] It is understandable that historical traffic data also includes time data corresponding to the traffic data. When it is predicted that traffic congestion will occur on a road section, the deep learning model can further predict the time period when the traffic congestion will occur based on the historical traffic data, which is the predicted time period. For example, when the current real-time traffic is greater than the preset traffic, the time when the real-time traffic is generated is paired with the historical time data to predict the predicted time period when the traffic congestion will occur.
[0094] In step S400 of the embodiment of the present invention, when it is predicted that a traffic congestion event will occur in a road section within the predicted time period, the congestion data is uploaded to the cloud database, and the congestion data includes the road section information, the predicted time period and the real-time flow rate, so as to store the above congestion data through the cloud database. Among them, the road section information includes the name, number, location, length, road grade and other information of the road section, and the road section where traffic congestion will occur in the predicted time period and the current real-time flow rate of the road section can be accurately located according to the road section information. It can be understood that when planning a navigation route through the navigation software, the road section information, the predicted time period and the real-time flow rate can be obtained from the cloud database, and a better navigation route can be planned according to the road section information, the predicted time period and the real-time flow rate, so that the navigation system can reflect the real-time status of the road, and the planned navigation route can bypass the road section predicted to have the risk of traffic congestion in the predicted time period, effectively reducing the situation where the navigation route encounters traffic congestion, and the planned route is more reasonable, making it more convenient for users to travel.
[0095] In addition, by uploading congestion data to the cloud database, it is also possible to restore, analyze and judge traffic congestion incidents, making it easier for traffic management departments to monitor and manage roads prone to traffic congestion.
[0096] It should be noted that, in some embodiments, when the real-time traffic flow of pedestrians and non-motor vehicles does not exceed the preset traffic flow, the probability of traffic congestion incidents occurring on the predicted road section is low. At this time, you can choose not to upload the collected traffic data, which can improve the operating efficiency of the road traffic monitoring system; in addition, the route planned by the navigation system can pass through the above-mentioned road section normally without the need for avoidance.
[0097] Reference Figure 4 As shown, in some embodiments, the step of predicting the congestion of a road section based on real-time traffic flow and historical traffic data specifically includes:
[0098] Step S310, analyzing historical congestion traffic and historical time periods according to the deep learning model to obtain a minimum traffic threshold;
[0099] Step S320, comparing the real-time traffic with the minimum traffic threshold, when the real-time traffic is greater than or equal to the minimum traffic threshold, predicting the time when the traffic congestion event will occur on the road section according to the real-time time period when the real-time traffic occurs.
[0100] It can be understood that historical traffic data includes historical congestion flow and historical time periods corresponding to historical congestion flow. Historical congestion flow can be understood as the flow of pedestrians and non-motor vehicles when historical traffic congestion events occur. The historical time period is the time period when historical traffic congestion events occur. The historical congestion flow and historical time period are acquired by the deep learning model within the collection period, wherein the collection period can be a preset time period or a time period learned through training. For example, a road section may easily generate a large amount of pedestrian and non-motor vehicle flow within a specific time. The specific time can be learned and trained by the deep learning model, so that the corresponding flow and time data can be collected regularly.
[0101] It can be understood that the greater the flow of pedestrians and non-motor vehicles, the more likely it is that traffic congestion events will occur. When the flow of pedestrians and non-motor vehicles is small and the flow of motor vehicles is not high, the road will remain unobstructed; as the flow of pedestrians and non-motor vehicles increases, the speed of motor vehicles will gradually decrease. When the flow of pedestrians and non-motor vehicles reaches a certain flow value, motor vehicles begin to slow down or stop driving, entering a traffic congestion state. Since the traffic congestion state will last for a certain period of time, the flow of pedestrians and non-motor vehicles will also change continuously. Therefore, in step S310 of the embodiment of the present invention, the historical congestion flow and historical time period are analyzed by a deep learning model, and the state in which motor vehicles slow down or stop driving can be regarded as the starting state of the traffic congestion event, and the corresponding time is the starting time point. According to the flow of the starting state and the starting time point, the minimum flow threshold for the occurrence of historical congestion events can be determined. If the flow of pedestrians and non-motor vehicles is lower than the minimum flow threshold, it means that the risk of pedestrians and non-motor vehicles causing congestion to road traffic is small, otherwise it means that the risk of pedestrians and non-motor vehicles causing congestion to road traffic is large.
[0102] In some embodiments, taking into account the random changes in the flow of pedestrians and non-motor vehicles, the average flow threshold within the real-time time period when the traffic congestion event occurs can be used as the minimum flow threshold. That is to say, when the real-time flow is greater than or equal to the average flow threshold within the real-time time period, the time when the traffic congestion event occurs on the road section is predicted based on the real-time time period when the real-time flow occurs.
[0103] After determining the minimum flow threshold, the real-time flow is compared with the minimum flow threshold to determine whether the real-time flow is greater than the minimum flow threshold. When the real-time flow is greater than or equal to the minimum flow threshold, it means that the flow of pedestrians and non-motor vehicles will bring a greater risk of road traffic congestion, and the risk of traffic congestion can be predicted more accurately; then the time of the traffic congestion event is predicted, and the prediction method is to predict the time when the traffic congestion event occurs on the road section based on the real-time time period when the real-time flow occurs, that is, the predicted time period for the occurrence of future traffic congestion events can be predicted.
[0104] It should be noted that in some embodiments, the deep learning model is trained using historical time periods, and the data set of historical time is input into the model for training. During the training process, the model automatically learns to extract useful features from the data by adjusting weights and parameters. The trained deep learning model can identify and analyze time periods and can cope with complex time processing tasks.
[0105] When the deep learning model is used to predict congestion, the time periods when historical congestion events occurred can be counted, and analysis can be performed based on the real-time time period when the real-time traffic occurs and the historical time period. If the real-time traffic exceeds the minimum traffic threshold and the real-time time period when the real-time traffic occurs is before the historical time period, or the real-time time period when the real-time traffic occurs partially overlaps with the historical time period, it can be predicted that the time period when future traffic congestion events will occur will be close to the historical time period. In this way, the predicted time period can be determined based on the historical time period.
[0106] Reference Figure 5 As shown, in some embodiments, the step of obtaining historical congestion traffic and historical time periods specifically includes:
[0107] Step S311, identifying lane lines in the image and vehicles, pedestrians and non-motor vehicles within the lane lines through a deep learning model;
[0108] Step S312: when the speed of the vehicles in the lane is lower than the first threshold, and the density of pedestrians and non-motor vehicles is greater than or equal to the second threshold, it is determined to be a historical congestion event;
[0109] Step S313: calculate the historical congestion flow according to the density and the duration that the vehicle speed is lower than the first threshold, and record the duration as the historical time period.
[0110] It is understandable that through steps S100 to S300 in the above embodiment, road traffic data can be obtained and uploaded for storage to form historical traffic data, which includes historical congestion events. When executing the prediction step, the image of the historical traffic data can be obtained through the deep learning model, and the image is recognized and processed to identify the lane lines, vehicles, pedestrians and non-motor vehicles in the image. For example, the feature information in the image is extracted through the image recognition algorithm to determine the lane lines in the road section. The lane lines include the boundary lines of the road and the middle lane separator lines. Then, the features of the objects located in the lane lines are identified to determine the vehicles, pedestrians and non-motor vehicles in the lane lines. The vehicles are motor vehicles, specifically cars, motorcycles, etc. By using the boundary lines and lane separator lines as the identification features of the lane lines, the vehicles, pedestrians and non-motor vehicles in the lane lines can be more accurately identified.
[0111] In the embodiment of the present invention, whether there is traffic congestion is determined based on the vehicle speed and the density of pedestrians and non-motor vehicles. Specifically, the vehicle speed is compared with a preset first threshold to determine whether the vehicle speed is in a slow driving state. When the vehicle speed is lower than the first threshold, it indicates that the vehicle is driving slowly or stopped. For example, the first threshold can be set to 5km / h, 10km / h, etc.; and the density of pedestrians and non-motor vehicles is compared with a preset second threshold to determine whether there are too many pedestrians and non-motor vehicles. The density of pedestrians and non-motor vehicles can be understood as the ratio of the number of pedestrians and non-motor vehicles to the area within the lane line, that is, the number of pedestrians and non-motor vehicles per square meter of the lane line. The greater the density of pedestrians and non-motor vehicles, the easier it is to cause traffic congestion. Therefore, the vehicle speed and the density of pedestrians and non-motor vehicles are used as conditions for determining whether there is traffic congestion. When the vehicle speed within the lane line is lower than the first threshold and the density of pedestrians and non-motor vehicles is greater than or equal to the second threshold, it is determined as a historical congestion event.
[0112] After determining the historical congestion events, the historical congestion flow is calculated based on the density that meets the traffic congestion conditions and the duration that the vehicle speed is lower than the first threshold. Considering that the density of pedestrians and non-motor vehicles will change continuously, the average density can be used to calculate the historical congestion flow. To illustrate with a specific example, the duration of the vehicle speed being lower than 5km / h is 30 minutes. During this period, the average density of pedestrians and motorcycles is 2 per square. According to the above duration and average density, the historical congestion flow can be calculated to be approximately 60 pedestrians and non-motor vehicles per minute. Of course, this is only an example. The specific values of the first threshold and the second threshold can be set according to the actual application of the road section, and are not limited here.
[0113] It can be understood that the duration when the vehicle speed is lower than the first threshold is the time when the historical congestion event occurred, that is, the historical time period corresponding to the historical congestion event. Therefore, the initial time point is recorded when the vehicle speed is lower than the first threshold, and the duration of the traffic congestion is recorded from the initial time point, so that the historical time period can be determined.
[0114] Reference Figure 6 As shown, in some embodiments, the step of predicting the time when a traffic congestion event will occur includes:
[0115] Step S321, when the real-time time period is greater than or equal to the preset time value, the real-time time period is divided into multiple time segments;
[0116] Step S322, extracting the maximum flow of pedestrians and non-motor vehicles corresponding to each time segment;
[0117] Step S323, input the maximum flow in each time segment into the trained deep learning model, calculate the average time from the minimum flow threshold to the maximum flow, and obtain the prediction result of the occurrence time of the traffic congestion event based on the average time.
[0118] Since real-time traffic is constantly changing, when the real-time traffic meets a threshold greater than or equal to the minimum traffic threshold and lasts for a long time, the real-time time period is divided into multiple time segments so that the size of each time segment is consistent, which is convenient for input into the deep learning model. Specifically, the real-time time period is compared with the preset time value to determine whether the real-time time period needs to be divided. When the real-time time period is greater than or equal to the preset time value, the above division operation is performed. After being divided into multiple time segments, the maximum traffic of pedestrians and non-motor vehicles corresponding to each time segment is extracted, and then the maximum traffic in each time segment is input into the deep learning model to calculate the average time from the minimum traffic threshold to the maximum traffic. Therefore, the time required for the real-time traffic to reach the maximum value can be estimated, which is conducive to improving the accuracy of the prediction.
[0119] Among them, the maximum flow can be understood as the flow value at which the flow reaches the peak in the time segment. The deep learning model can determine the maximum flow of the entire time based on the maximum flow of each time segment, and thus calculate the average time for the minimum flow threshold to reach the maximum flow.
[0120] In the embodiment of the present invention, the deep learning model is trained using historical congestion traffic and historical time periods, which can achieve higher accuracy than traditional machine learning models and effectively obtain time prediction results, thereby determining the prediction time period.
[0121] It can be understood that the process of traffic congestion is first slight congestion, then severe congestion, then congestion gradually weakens, and finally returns to smooth flow. At the same time, the traffic congestion flow also changes. When the maximum flow is reached, it can be understood as a severe congestion state. Therefore, the process of flow from the minimum value to the peak value, and then from the peak value to smooth flow, can be estimated by using the average time to obtain the approximate time of the traffic congestion event, that is, to predict the predicted time period. Specifically, in some embodiments, the deep learning model can establish a relationship formula between the average time and the predicted time period for calculation. The calculation formula is as follows: T 1 =kT 0 , where T 1 represents the forecast period, T 0represents the average time, k represents the weight parameter, and the value range of k is greater than or equal to 1. For example, the average time calculated by the deep learning model is 20 minutes, and the value of k is 2. According to the above formula, it can be estimated that the time for the traffic congestion event to occur is 40 minutes, which means that the traffic congestion event is predicted to occur within the next 40 minutes. In this way, when planning the route, the navigation route can be evasive in the next 40 minutes to bypass the road section with congestion risk.
[0122] It should be noted that when the real-time time period is less than the preset time value, it means that the duration of the real-time traffic exceeding the minimum traffic threshold is short, and there is no need to divide the real-time time period. At this time, the maximum traffic of pedestrians and non-motor vehicles corresponding to the real-time time period can be directly extracted, and then the above-mentioned step S323 is executed, and the details will not be repeated.
[0123] In the above step S311, lane lines, vehicles, pedestrians and non-motor vehicles in the image are identified by a deep learning model. In some embodiments, the specific steps of the identification process include:
[0124] Step S3111, using a deep learning model to analyze objects within the lane line and generate corresponding boundary features;
[0125] Step S3112, associating boundary features in continuous images to generate a motion trajectory of the object;
[0126] Step S3113, matching the boundary features, motion trajectories and a preset feature library to identify vehicles, pedestrians and non-motor vehicles within the lane.
[0127] It can be understood that boundary features include contours, shapes and other features. The deep learning model can recognize the edges and shapes of objects through training. For example, the boundary features of pedestrians and non-motor vehicles can be extracted through image vision algorithms. These features can help subsequent steps better identify the specific type and motion state of the object.
[0128] By extracting boundary features from continuous images, then associating all boundary features, generating the motion trajectory of the object, and then matching the boundary features and motion trajectory with the preset feature library, vehicles, pedestrians and non-motor vehicles can be identified. Since there are boundary features of objects in continuous images, by associating these features, the movement direction and speed of the object can be determined, thereby generating the motion trajectory of the object. The preset feature library includes feature data such as the outline, shape, and motion posture of vehicles, pedestrians, and non-motor vehicles. When the boundary features and motion trajectory of the object match the data in the feature library, the type of the object can be determined, and the recognition accuracy is high. For example, the outline edge and shape features of the pedestrian are first obtained, and the pedestrian features of the continuous images are associated, so that the motion trajectory of the pedestrian can be obtained. When it matches the outline and motion posture of the pedestrian in the feature library, the moving object can be identified as a pedestrian.
[0129] It should be noted that, considering that pedestrians and non-motor vehicles are the main factors causing congestion in non-main road sections in cities, only vehicles, pedestrians and non-motor vehicles are identified in the embodiment. Of course, this is only an example, and in some embodiments, it is also possible to add identification of mobile vendors. In addition, in the embodiment, only moving objects within the lane line are identified, and objects outside the lane line are not tracked to improve the recognition efficiency.
[0130] Taking a specific example to illustrate, the collected images are preprocessed, including denoising, enhancement, contrast adjustment and other operations to improve image quality; the deep learning model uses an image recognition algorithm to extract the features of the lane lines, such as color, texture, shape, etc., and determines the position of the lane lines in the image, and then uses the lane lines as boundaries to identify image objects within the lane lines, thereby determining the specific type of the object.
[0131] See also Figure 7 As shown, in some embodiments, the road traffic prediction method further includes:
[0132] Step S500, collecting image data of the road section in real time, and analyzing the image data using a pre-trained deep learning model to identify lane lines and vehicles and vehicle categories within the lane lines;
[0133] Step S600, calculating the lane occupancy rate of pedestrians and non-motor vehicles on the road section according to the vehicles in the lane line, the vehicle type and the real-time traffic flow;
[0134] Step S700: When the lane occupancy rate is greater than or equal to the third threshold, it is determined that a traffic congestion event currently occurs on the road section and the traffic congestion event is uploaded to the cloud database.
[0135] It is understood that in some embodiments, the real-time collected image data is processed by a deep learning model to identify lane lines and vehicles, and then the real-time traffic of vehicles, pedestrians and non-motor vehicles is counted and analyzed, and the lane occupancy rate of pedestrians and non-motor vehicles on the road section is calculated based on the above information, so that the traffic congestion situation can be quickly determined. It can be understood that the lane occupancy rate can be understood as the ratio of pedestrians and non-motor vehicles occupying the lane. The higher the lane occupancy rate, the more pedestrians and non-motor vehicles there are, the greater the traffic flow, and the easier it is to cause traffic congestion. Therefore, by presetting the third threshold, the lane occupancy rate is compared with the third threshold to determine whether a traffic congestion event occurs on the road section. When the lane occupancy rate is greater than or equal to the third threshold, it is determined that a traffic congestion event currently occurs on the road section, and the relevant congestion data is uploaded to the cloud database.
[0136] For example, the third threshold value can be set to an occupancy rate of 30%. When the lane occupancy rate of pedestrians and non-motor vehicles is greater than or equal to 30%, it is determined that a traffic congestion event has occurred and the data upload operation is performed. Of course, this is only an example, and the specific value of the third threshold value can be set according to the actual application of the road section, and is not limited here.
[0137] refer to Figures 8 to 9 A method for planning a traffic route for non-arterial roads in an urban area according to an embodiment of the present invention is described, which is applicable to a vehicle navigation system, a mobile terminal navigation system, etc. The route planning method is described below by taking a specific example.
[0138] Reference Figure 8 As shown, the route planning method of the embodiment of the present invention includes but is not limited to the following steps:
[0139] Step S800, acquiring image data of a non-main road section through a traffic monitoring device;
[0140] Step S900, processing the image data to identify pedestrian and non-motor vehicle images;
[0141] Step S1000, using a pre-trained deep learning model to analyze pedestrian and non-motor vehicle images, calculate the real-time flow of pedestrians and non-motor vehicles in the road section, and predict the congestion of the road section in combination with historical traffic data;
[0142] Step S1100, when a traffic congestion event is predicted to occur on a road section within a predicted time period, the congestion data is uploaded to a cloud database, and the congestion data includes road section information, predicted time period, and real-time traffic;
[0143] Step S1200, determining a navigation route of the vehicle based on congestion data, the current location and destination of the vehicle.
[0144] It can be understood that steps S800 to S1100 in the route planning method of the embodiment are intended to obtain congestion data of non-main roads. The specific execution process can refer to the operations of steps S100 to S400 in the above embodiment. That is to say, the congestion data is obtained by using the road traffic monitoring method and uploaded to the cloud database for storage. When planning the route, the navigation system can directly obtain the corresponding congestion data from the cloud database, and then plan the navigation route of the vehicle in combination with the current position and destination of the vehicle. Since the congestion data is added for planning, the congestion data can be used to know the risk of traffic congestion on the corresponding road section and the predicted time period when congestion may occur. In this way, when planning the route, the navigation route can avoid the road section with the risk of traffic congestion, make the navigation route better, reduce the situation of traffic congestion on the navigation route, and make travel more convenient for users.
[0145] In addition, by uploading congestion data to the cloud database, on the one hand, it is convenient for the navigation system to plan a better navigation route, and on the other hand, it can also realize the restoration, analysis and problem judgment of traffic congestion incidents, which is convenient for traffic management departments to monitor and manage roads prone to traffic congestion.
[0146] Reference Fig. 9 As shown, in some embodiments, the step of determining the navigation route of the vehicle according to the congestion data, the current position and the destination of the vehicle specifically includes:
[0147] Step S1210, setting a navigation route according to the current location and destination;
[0148] Step S1220, when the navigation route passes through a road section, the real-time traffic is greater than or equal to the historical congestion traffic and the vehicle's driving time on the road section intersects with the predicted time period, adjust the navigation route so that the adjusted navigation route does not pass through the road section.
[0149] Taking the in-vehicle navigation system as an example, the user enters the destination through the vehicle's interactive interface, and the navigation system calculates the navigation route based on the current location and the destination. This navigation route is the initial navigation route; then the congestion data is combined to determine whether the initial navigation route needs to be adjusted. If the navigation route passes through a section in the congestion data, it means that the navigation route is at risk of encountering traffic congestion, and the real-time traffic is compared with the historical congestion traffic to determine whether the real-time traffic is too large, and further determine whether the driving time of the vehicle in the navigation route intersects with the predicted time period of the traffic congestion event. If the two intersect, it means that the vehicle will pass through the above-mentioned section within the predicted time period.
[0150] Therefore, when the navigation route passes through the above-mentioned road section, the real-time traffic volume is greater than or equal to the historical congestion traffic volume, and the vehicle's driving time on the road section intersects with the predicted time period, the vehicle has a high risk of encountering traffic congestion. At this time, the navigation route is dynamically adjusted so that the adjusted navigation route does not pass through the road section with congestion risks, thereby reducing the risk of encountering traffic congestion, making the navigation route better, improving the efficiency and accuracy of navigation, and making travel more convenient for users.
[0151] In some embodiments, determining a navigation route for a vehicle based on congestion data, a current location of the vehicle, and a destination includes:
[0152] Step S1230, setting a navigation route according to the current location and destination;
[0153] Step S1240, when the destination is within the road section and the vehicle's travel time on the road section does not overlap with the predicted time period, the initially set navigation route is determined as the navigation route, and a reminder message is sent to the driver, including a prediction result that a traffic congestion event will occur on the road section within the predicted time period.
[0154] It is understandable that since the risk of traffic congestion events occurring within the predicted time period is higher, and the probability of traffic congestion occurring on sections during non-predicted time periods is greatly reduced, therefore, when the navigation route passes through a section and the vehicle's driving time on the section does not intersect with the predicted time period, it means that although the vehicle will pass through the above section, it will not pass through within the predicted time period, and the risk of encountering traffic congestion is low. The initial navigation route will not be adjusted, and a reminder message will be sent to the driver to remind the driver that a certain section on the navigation route is prone to traffic congestion during a certain time period. In this way, in order to inform the driver of the traffic conditions of the section in advance, it can help the driver better plan the driving route and time, improve navigation efficiency, avoid unnecessary delays and congestion, and provide a better navigation experience.
[0155] Combination Figure 3 It can be understood that in the embodiment, a reminder message can be sent to the driver through the central control interactive interface of the vehicle, so that the driver can quickly know the current traffic conditions. Of course, in the above step S1220, when the step of adjusting the navigation route is performed, the driver can also be informed through the interactive interface, so that the driver can understand the adjusted driving route and time.
[0156] Reference Fig.10 As shown, in some embodiments, determining a navigation route of a vehicle based on congestion data, a current location of the vehicle, and a destination includes:
[0157] Step S1250, setting an initial navigation route according to the current location and destination;
[0158] Step S1260, when both the initial navigation route and the historical navigation route pass through the road section, the real-time traffic volume is greater than or equal to the historical congestion traffic volume, and the driving time of the vehicle on the road section intersects with the predicted time period, the navigation route is adjusted so that the adjusted navigation route does not pass through the road section, and the initial navigation route and the adjusted navigation route are sent to the vehicle or mobile terminal to obtain the driver's operation instruction to select the initial navigation route or the adjusted navigation route;
[0159] Step S1270: When an operation instruction is obtained, the route selected by the driver is determined as the current navigation route according to the operation instruction.
[0160] It is understandable that drivers usually use the vehicle's onboard navigation system or a mobile terminal (such as a mobile phone) for navigation. When the navigation function of the vehicle or mobile terminal is turned on, the driver's historical navigation route is obtained. The historical navigation route includes information such as the starting location, destination, and road sections passed.
[0161] In step S1260 of the embodiment, when the navigation route passes through the above-mentioned road section, the real-time traffic volume is greater than or equal to the historical congestion traffic volume, and the vehicle's driving time on the road section intersects with the predicted time period, it is determined that the risk of the vehicle encountering traffic congestion is high, and the navigation route is dynamically adjusted at this time to obtain an adjusted navigation route; further determine whether the historical navigation route passes through the above-mentioned road section. If the historical navigation route passes through the above-mentioned road section, it means that the user may have passed by the above-mentioned road section and has a certain understanding of the road conditions of the road section. The initial navigation route and the adjusted navigation route are sent to the navigation system of the vehicle or the navigation system of the mobile phone. Taking the mobile phone as an example, the mobile phone will receive the initial navigation route and the adjusted navigation route. The driver manually selects the two routes, generates an operation instruction after selection, and provides feedback, so that the route selected by the driver is determined as the current navigation route.
[0162] It should be noted that the participants in manual operation can be the driver and the co-pilot, so that users can choose the route according to the actual situation. For the driver, the driving route can be adjusted according to the actual situation to avoid congestion and unexpected situations, or they can rely on their own driving experience and judgment to deal with complex urban environments. It can enhance the user's sense of participation, be more flexible to use, and enable user feedback to be resolved.
[0163] Considering that sudden road conditions, driving conditions of other vehicles and the driving habits of drivers are also important factors leading to traffic congestion or traffic accidents, how to effectively predict possible traffic congestion or traffic accidents in the future and make corresponding route adjustments is also a problem to be solved by the route planning method of the embodiment of the present invention. In some embodiments, the deep learning model is trained using historical navigation data, and the historical navigation data includes historical navigation routes, driver's selection habits, etc. The deep learning model can be used to analyze the average difference between the driving time predicted by the navigation system and the actual driving time of the driver in the historical navigation route, and the driver's driving habits can be further obtained. The driving habits include the driver's braking frequency in daily driving, the following distance from the vehicle in front, the preference for sudden braking, etc. The navigation route is combined with the above average difference and driving habits for analysis to evaluate whether the navigation route is suitable for the current driver. For example, if the driver's driving habit is that he drives slowly every day and tends to brake slowly, the deep learning model can be used to rate the driving habit as a cautious driving level, and the initial navigation route can be recommended to the driver based on this level; if the driver's driving habit is that he drives at a high average speed every day and brakes frequently, the deep learning model can be used to rate the driving habit as an aggressive driving level, and an adjusted navigation route can be recommended to the driver based on this level, so that the driver's navigation route does not pass through sections that are prone to traffic congestion, thereby helping to reduce the chance of traffic congestion or traffic accidents, and making route planning more flexible and reasonable.
[0164] The following is an explanation using specific examples. The embodiments specifically include data collection and data processing, wherein data collection includes collecting current vehicle information, surrounding vehicle information, and driver data, and uploading them to a cloud database. Each vehicle has its own corresponding vehicle identification for easy identification. The driver's data includes collecting the driver's eyeball image and pupil image within a certain period of time, the purpose of which is to extract the driver's iris features, which can reflect the driver's driving emotions and future driving tendencies to a certain extent; and collecting the driver's pulse information within a certain period of time, and the pulse information can be collected through the steering wheel. For example, the unit for collecting pulse information is set in the steering wheel, the purpose of which is to quickly collect relevant pulse information when the driver is holding the steering wheel;
[0165] The cloud database pre-stores the basic information of all drivers. The basic information focuses on the driver's emotional history, and each emotional history has a corresponding emotional level. According to research, the driver's historical or current emotional illness will affect the driver's actual driving habits, so the driver's emotional history needs to be taken into consideration. It should be noted that if the emotional history information is actively uploaded by the driver, there may be a situation where the driver's emotional history is empty. At this time, the driver's driving habits will be taken into consideration, and data such as driving habits can be collected through the vehicle and then uploaded to the cloud database for storage.
[0166] Specifically, when a driver's emotional history is stored in the database, such as depression, mania, etc., only the emotional history factor is considered when the emotional level is determined according to the actual type; when a driver in the database does not provide an emotional history or does not have an emotional history (the emotional level value is 0), the driver's driving habits in the historical time period are obtained. In addition, the method according to the above embodiment can obtain data such as the current location, destination, road section information, real-time traffic of the road section, and the predicted time period obtained by traffic prediction for the road section.
[0167] Then the above data is processed. For the driver's eyeball image and pupil image, the pupil image is projected onto the eyeball image, the projected image is synthesized, and the feature extraction of the projected image is performed to extract the driver's iris features in each frame, and the iris feature change curve is constructed based on the multi-frame iris features, and then the mean and variance of the iris feature change curve are calculated, and the ratio of the mean to the variance is calculated as the human iris feature vector. For the driver's pulse information and vehicle information, data cleaning is performed, and the driver's emotional level identification is predicted through a deep learning model combined with the driver's historical driving habits. In addition, for traffic congestion conditions and road surface related information, corresponding traffic signs are assigned according to the traffic congestion conditions; and the road surface related information needs to be calculated to integrate the road surface characteristics.
[0168] First, the human iris feature vector, pulse information and emotion level mark are used to predict the driver's driving tendency in the future and assign a coefficient corresponding to the driving tendency. Then, the vehicle information of multiple vehicles is used to predict the driving angle and speed of the current vehicle and other surrounding vehicles in the future; the initial navigation route is determined according to the current location destination, and the traffic congestion of the above-mentioned road section and the relevant information of the road surface are predicted according to the road section information, the real-time traffic flow of the road section and the traffic prediction of the road section. And the driving behavior of the current vehicle and other vehicles in the future is predicted, such as whether the vehicle will slow down, accelerate, brake suddenly, etc. This driving behavior can be represented by the acceleration (or deceleration) of the vehicle in the future and the increment (or decrement) of the angle. In the above prediction process, the coefficient corresponding to the driving tendency, the driving angle, speed, acceleration (or deceleration) of the current vehicle and other vehicles in the future are obtained. For each vehicle, the speed range of the vehicle at a future moment is determined based on the vehicle speed and its acceleration (or deceleration); the driving angle range of the vehicle at a future moment is determined based on the vehicle's driving angle and its angle increment (or decrement); and then, based on each vehicle, the driving trajectory of each vehicle is drawn through its driving angle range and speed range.
[0169] When the driving trajectories of the current vehicle and several vehicles around it overlap, the risk of traffic congestion on the above-mentioned road section in the navigation route is evaluated according to the overlap and the predicted time period obtained by the traffic forecast of the road section, and the risk level is predicted, so as to determine whether to adjust the navigation route according to the predicted level. The driver can also choose whether to adjust the navigation route, so as to determine the navigation route according to multiple factors to customize a more reasonable route, so that the driver can be more involved in the route customization process, which can help the driver better plan the driving route and time, improve navigation efficiency, avoid unnecessary delays and congestion, and provide a better navigation experience.
[0170] Reference Fig.11 As shown, the embodiment of the present invention also provides a traffic monitoring system for non-main road in urban area, which can perform the above Figures 1 to 7 The road traffic prediction method provided by the illustrated embodiment, specifically, the non-main road traffic monitoring system includes a collection module, an identification module and a judgment module.
[0171] Among them, the acquisition module is used to obtain image data of non-main road sections through traffic monitoring equipment; the recognition module is used to process the image data and identify pedestrian and non-motor vehicle images; and use the pre-trained deep learning model to analyze the pedestrian and non-motor vehicle images and calculate the real-time flow of pedestrians and non-motor vehicles in the section; the judgment module is used to combine the real-time flow with historical traffic data to predict the congestion of the section. When it is predicted that a traffic congestion incident will occur in the section within the predicted time period, the congestion data will be uploaded to the cloud database. The congestion data includes section information, predicted time period and real-time flow.
[0172] The acquisition module of the non-main road traffic monitoring system of the embodiment obtains image data of the road section through the traffic monitoring equipment, and uses the recognition module to identify the image data to obtain the images of pedestrians and non-motor vehicles, and then uses the pre-trained deep learning model to analyze the pedestrian and non-motor vehicle images. The deep learning module can calculate the real-time flow of pedestrians and non-motor vehicles in the road section; then the judgment module is used to predict the congestion situation of the road section according to the real-time flow and historical traffic data, that is, to predict whether the current real-time flow of pedestrians and non-motor vehicles may cause road traffic congestion, so as to realize the monitoring of road congestion; when it is predicted that the above-mentioned road section will have a traffic congestion event within the predicted time period, the congestion data will be uploaded to the cloud database, so that the route planning system can obtain the congestion data from the cloud database, and can plan a better navigation route according to the congestion data, thereby reducing the situation where the navigation route encounters traffic congestion, making user travel more convenient.
[0173] In addition, by uploading congestion data to the cloud database, on the one hand, it is helpful for the navigation system to plan a better navigation route, and on the other hand, it can also realize the restoration, analysis and problem judgment of traffic congestion incidents, which is convenient for traffic management departments to monitor and manage roads prone to traffic congestion.
[0174] Reference Fig.12 As shown, the embodiment of the present invention also provides a traffic route planning system for non-main road in urban area. The route planning system can be understood as a navigation system, which can execute the above Figures 8 to 10 The illustrated embodiment provides a method for planning a traffic route for non-arterial roads in an urban area. Specifically, the route planning system includes a collection module, a recognition module, a judgment module, and a path planning module.
[0175] Among them, the acquisition module is used to obtain image data of road sections through traffic monitoring equipment; the recognition module is used to process image data and identify images of pedestrians and non-motor vehicles; and use pre-trained deep learning models to analyze pedestrian and non-motor vehicle images and calculate the real-time flow of pedestrians and non-motor vehicles in the section; the judgment module is used to combine real-time flow with historical traffic data to predict the congestion of the section. When a traffic congestion event is predicted to occur in the predicted time period, the congestion data is uploaded to the cloud database. The congestion data includes section information, predicted time period and real-time flow; the path planning module is used to determine the vehicle's navigation route based on congestion data, the vehicle's current location and destination.
[0176] The acquisition module of the route planning system of the embodiment obtains the image data of the road section through the traffic monitoring equipment, and uses the recognition module to recognize the image data to obtain the images of pedestrians and non-motor vehicles, and then uses the pre-trained deep learning model to analyze the pedestrian and non-motor vehicle images. The deep learning module can calculate the real-time flow of pedestrians and non-motor vehicles in the road section; then the judgment module is used to predict the congestion of the road section according to the real-time flow and historical traffic data, that is, to predict whether the current real-time flow of pedestrians and non-motor vehicles may cause road traffic congestion, so as to realize the monitoring of road congestion; when the above prediction is When a traffic congestion event occurs on the road section within the predicted time period, the congestion data is uploaded to the cloud database. The congestion data includes the road section information, the predicted time period and the real-time traffic. Finally, the vehicle's navigation route is determined by the path planning module according to the congestion data, the vehicle's current position and the destination. Since the congestion data is added for planning, the risk of traffic congestion on the corresponding road section and the predicted time period when congestion may occur can be known through the congestion data. In this way, when planning the route, the navigation route can avoid the road section with the risk of traffic congestion, making the navigation route better, reducing the situation of traffic congestion on the navigation route, and making it more convenient for users to travel.
[0177] In addition, by uploading congestion data to the cloud database, on the one hand, it is convenient for the navigation system to plan a better navigation route, and on the other hand, it can also realize the restoration, analysis and problem judgment of traffic congestion incidents, which is convenient for traffic management departments to monitor and manage roads prone to traffic congestion.
[0178] An embodiment of the present invention further provides an electronic device, comprising at least one processor; and a memory connected to the at least one processor in communication; 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 road traffic prediction method or route planning method of the above embodiment. The processor and the memory may be connected via a bus or other means.
[0179] The memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transitory memory, such as at least one disk storage device, a flash memory device, or other non-transitory solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network.
[0180] The non-transient software program and instructions required to implement the method of the above embodiment are stored in the memory. When executed by the processor, the road traffic prediction method or route planning method in the above embodiment is executed, for example, the above described method is executed. Figure 2 Steps S100 to S400 of the method, Figure 4 Steps S310 to S320 of the method, Figure 5 Steps S311 to S313 of the method, Figure 6 Steps S321 to S323 of the method, Figure 7 Steps S500 to S700 of the method, Figure 8 Method steps S800 to S1200, etc.
[0181] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
[0182] Another embodiment of the present invention further provides a computer-readable storage medium, which stores computer-executable instructions. The computer-executable instructions are executed by a processor or a controller, for example, by a processor in the electronic device in the above embodiment, so that the processor can execute the road traffic prediction method or route planning method in the above embodiment, for example, execute the above-described Figure 2 Steps S100 to S400 of the method, Figure 4 Steps S310 to S320 of the method, Figure 5 Steps S311 to S313 of the method, Figure 6 Steps S321 to S323 of the method, Figure 7 Steps S500 to S700 of the method, Figure 8 Method steps S800 to S1200, etc.
[0183] It will be appreciated by those skilled in the art that all or some of the steps and systems in the disclosed method above may be implemented as software, firmware, hardware and appropriate combinations thereof. Some physical components or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor or a microprocessor, or may be implemented as hardware, or may be implemented as an integrated circuit, such as an application specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or a non-transitory medium) and a communication medium (or a temporary medium). As known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, disk storage or other magnetic storage devices, or any other medium that may be used to store desired information and may be accessed by a computer. Furthermore, it is well known to those skilled in the art that communication media typically embodies computer readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.
[0184] The above is a specific description of the preferred implementation of the present invention, but the present invention is not limited to the above-mentioned implementation mode. Technical personnel familiar with the field can also make various equivalent deformations or substitutions without violating the spirit of the present invention. These equivalent deformations or substitutions are all included in the scope defined by the claims of the present invention.
Claims
1. A method for predicting traffic routes on non-main roads in urban areas, characterized in that: include: Acquire image data of non-arterial road sections through traffic monitoring equipment; Processing the image data to identify pedestrian and non-motor vehicle images; Analyze the pedestrian and non-motor vehicle images using a pre-trained deep learning model, calculate the real-time flow of pedestrians and non-motor vehicles in the non-arterial road section, and predict the congestion of the non-arterial road section in combination with historical traffic data; When it is predicted that a traffic congestion event will occur on the non-arterial road section within the predicted time period, uploading the congestion data to a cloud database, the congestion data including the road section information, the predicted time period and the real-time traffic volume; determining a navigation route of the vehicle based on the congestion data, the current location and the destination of the vehicle; The predicting of the congestion situation of the non-arterial road section by combining historical traffic data includes: The deep learning model acquires the historical traffic data within a collection period, wherein the historical traffic data includes historical congestion flow and a historical time period corresponding to the historical congestion flow; Analyze the historical congestion flow and the historical time period according to the deep learning model to obtain a minimum flow threshold; The real-time traffic volume is compared with the minimum traffic volume threshold value, and when the real-time traffic volume is greater than or equal to the minimum traffic volume threshold value, the time when the traffic congestion event occurs on the non-main road section is predicted according to the real-time time period when the real-time traffic volume occurs; Using the historical time period to train the deep learning model; The predicting of the time when the traffic congestion event occurs on the non-main road section according to the real-time time period when the real-time traffic occurs includes: When the real-time time period is greater than or equal to a preset time value, the real-time time period is divided into a plurality of time segments; Extracting the maximum flow of pedestrians and non-motor vehicles corresponding to each of the time segments; The maximum flow corresponding to each time segment is input into the trained deep learning model, the average time from the minimum flow threshold to the maximum flow is calculated, and the prediction result of the occurrence time of the traffic congestion event is obtained based on the average time. The deep learning model establishes the relationship formula between the average time and the predicted time period for calculation. The calculation formula is as follows: T1=kT0, wherein T1 represents the predicted time period, T0 represents the average time, k represents the weight parameter, and the value range of k is greater than or equal to 1.
2. The method according to claim 1, characterized in that The acquiring of the historical traffic data by the deep learning model within a collection period includes: Identify lane lines in an image and vehicles, pedestrians, and non-motor vehicles within the lane lines using the deep learning model; When the speed of the vehicles in the lane is lower than the first threshold, and the density of pedestrians and non-motor vehicles is greater than or equal to the second threshold, it is determined to be a historical congestion event; The historical congestion flow is calculated according to the density and the duration that the vehicle speed is lower than the first threshold, and the duration is recorded as the historical time period.
3. The method according to claim 2, characterized in that The identifying the lane lines in the image and the vehicles, pedestrians and non-motor vehicles located within the lane lines by the deep learning model includes: Analyzing objects within the lane line using the deep learning model to generate corresponding boundary features; Associating the boundary features in continuous images to generate a motion trajectory of the object; The boundary features and the motion trajectory are matched with a preset feature library to identify vehicles, pedestrians and non-motor vehicles within the lane line.
4. The method according to claim 1, characterized in that The method further comprises: Collect the image data of the non-arterial road section in real time, and analyze the image data using a pre-trained deep learning model to identify lane lines and vehicles and vehicle categories located within the lane lines; The lane occupancy rate of pedestrians and non-motor vehicles in the non-main road section is calculated according to the vehicles in the lane line, the vehicle category and the real-time traffic flow; When the lane occupancy rate is greater than or equal to a third threshold, it is determined that a traffic congestion event currently occurs on the non-arterial road section and the traffic congestion event is uploaded to the cloud database.
5. The method according to claim 1, characterized in that Determining the navigation route of the vehicle according to the congestion data, the current position and the destination of the vehicle includes: Setting a navigation route according to the current location and the destination; When the navigation route passes through the non-arterial road section, the real-time traffic volume is greater than or equal to the historical congestion traffic volume and the driving time of the vehicle on the non-arterial road section intersects with the predicted time period, the navigation route is adjusted so that the adjusted navigation route does not pass through the non-arterial road section.
6. The method according to claim 5, characterized in that The determining the navigation route of the vehicle according to the congestion data, the current position and the destination of the vehicle further includes: When the destination is located on the non-arterial road section and the driving time of the vehicle on the non-arterial road section does not intersect with the predicted time period, the initially set navigation route is determined as the navigation route, and a reminder message is sent to the driver, wherein the reminder message includes a predicted result that a traffic congestion event will occur on the non-arterial road section within the predicted time period.
7. The method according to claim 1, characterized in that The method further comprises: When the navigation function of the vehicle or mobile terminal is turned on, obtaining the driver's historical navigation route; Determining the navigation route of the vehicle according to the congestion data, the current position and the destination of the vehicle includes: setting an initial navigation route according to the current location and the destination; When both the initial navigation route and the historical navigation route pass through the non-arterial road section, the real-time traffic volume is greater than or equal to the historical congestion traffic volume, and the driving time of the vehicle on the non-arterial road section intersects with the predicted time period, adjusting the navigation route so that the adjusted navigation route does not pass through the non-arterial road section, and sending the initial navigation route and the adjusted navigation route to the vehicle or the mobile terminal to obtain an operation instruction of the driver to select the initial navigation route or the adjusted navigation route; When the operation instruction is acquired, the route selected by the driver is determined as the current navigation route according to the operation instruction.
8. A traffic route prediction system for non-main road in urban area, implemented based on the traffic route prediction method for non-main road in urban area according to claim 1, characterized in that: include: A collection module, used to obtain image data of non-main road sections through traffic monitoring equipment; A recognition module, used to process the image data and recognize pedestrian and non-motor vehicle images; and using a pre-trained deep learning model to analyze the pedestrian and non-motor vehicle images to calculate the real-time flow of pedestrians and non-motor vehicles in the non-main road section; A judgment module, used to predict the congestion situation of the non-arterial road section by combining the real-time traffic with historical traffic data, and when it is predicted that the non-arterial road section will have a traffic congestion event within the predicted time period, uploading the congestion data to a cloud database, the congestion data including the road section information, the predicted time period and the real-time traffic; A path planning module is used to determine a navigation route for the vehicle based on the congestion data, the current location and destination of the vehicle.
9. An electronic device, characterized in that: include: at least one processor; as well as 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 urban non-arterial road traffic route prediction method described in any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the urban non-arterial road traffic route prediction method according to any one of claims 1 to 7.
Citation Information
Patent Citations
Intelligent traffic time period dividing method based on sliding average algorithm
CN109697849A
Peak start time prediction method, device and equipment and medium
CN113256973A
Road condition determination method and device and computer readable storage medium
CN115273452A
Recognition method and device of variable lane, equipment and storage medium
CN115731700A
Urban road intersection time period division method and storage medium
CN115798234A
