A panoramic segmentation-based global intelligent scheduling method and system

Abstract

The application discloses a kind of global intelligent scheduling method and system based on panorama segmentation, it is related to panorama segmentation technical field, method includes: according to the road monitoring data of target area is extracted to the preset time node, obtains traffic video data;Based on the collection position, according to the road planning map of target area, the traffic video data is carried out data segmentation, generates N traffic video data collection;Based on BP neural network, construct congestion duration prediction model;Generate N longitudinal duration prediction set and N transverse duration prediction set;N duration prediction bias is calculated;Based on the N duration prediction bias generates signal optimization instruction, and according to the signal optimization instruction, the traffic signal of target area is optimized control, the problem that the real-time of traffic scheduling work is poor due to not enough rigorous and lack of completeness in prior art is solved, effective solution about traffic congestion situation is realized.

Description

Technical Field

[0001] This invention relates to the field of panoramic segmentation technology, specifically to a global intelligent scheduling method and system based on panoramic segmentation. Background Technology

[0002] With the development of urban transportation and the acceleration of urbanization, traffic management has become increasingly complex. Traditional traffic management methods mainly rely on pre-set traffic lights to control traffic conditions, but these methods cannot be adjusted in a timely manner according to actual traffic conditions, ultimately leading to traffic congestion. Panoramic segmentation, an image processing technique, can divide a panoramic image into multiple smaller regions with similar attributes. Applying panoramic segmentation technology to traffic management, and making corresponding adjustments based on different road segments, is of great significance for improving vehicle traffic efficiency.

[0003] The existing traffic scheduling methods suffer from poor real-time performance due to insufficient rigor and completeness, which ultimately prevents effective solutions to traffic congestion. Summary of the Invention

[0004] This application provides a global intelligent scheduling method and system based on panoramic segmentation, which solves the problem of poor real-time performance in existing traffic scheduling work due to insufficient rigor and completeness, and achieves an effective solution to traffic congestion.

[0005] In view of the above problems, this application provides a global intelligent scheduling method based on panoramic segmentation.

[0006] Firstly, this application provides a global intelligent scheduling method based on panoramic segmentation. The method includes: extracting road monitoring data of a target area according to preset time nodes to obtain traffic video data, wherein the traffic video data carries a collection location identifier; segmenting the traffic video data according to the collection location and a road planning map of the target area to generate N traffic video data sets, where N is the number of traffic light-controlled intersections in the target area; constructing a congestion duration prediction model based on a BP neural network, wherein the congestion duration prediction model includes N congestion duration prediction channels; inputting the N traffic video data sets into the N congestion duration prediction channels respectively for prediction to generate N longitudinal duration prediction sets and N lateral duration prediction sets; calculating N duration prediction deviations based on the N longitudinal and N lateral duration prediction sets; generating signal optimization instructions based on the N duration prediction deviations, and optimizing the control of traffic lights in the target area according to the signal optimization instructions.

[0007] Secondly, this application provides a full-domain intelligent dispatching system based on panoramic segmentation. The system includes: a data extraction module: extracting road monitoring data of a target area according to preset time nodes to obtain traffic video data, wherein the traffic video data carries a collection location identifier; a data segmentation module: segmenting the traffic video data according to the collection location and the road planning map of the target area to generate N traffic video data sets, where N is the number of traffic light-controlled intersections in the target area; a model construction module: constructing a congestion duration prediction model based on a BP neural network, wherein the congestion duration prediction model includes N congestion duration prediction channels; a prediction set module: inputting the N traffic video data sets into the N congestion duration prediction channels respectively for prediction to generate N longitudinal duration prediction sets and N lateral duration prediction sets; a prediction deviation module: calculating N duration prediction deviations based on the N longitudinal duration prediction sets and the N lateral duration prediction sets; and an optimization instruction module: generating signal optimization instructions based on the N duration prediction deviations and optimizing the control of traffic lights in the target area according to the signal optimization instructions.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0009] This application provides a global intelligent scheduling method and system based on panoramic segmentation. It extracts road monitoring data of a target area according to preset time nodes to obtain traffic video data with acquisition location identifiers. Then, based on the acquisition location, the traffic video data is segmented according to the road planning map of the target area, generating N traffic video data sets. A congestion duration prediction model is constructed based on a BP neural network. The N traffic video data sets are input into N congestion duration prediction channels for prediction, generating N vertical duration prediction sets and N horizontal duration prediction sets. Based on these sets, N duration prediction deviations are calculated, and signal optimization instructions are generated. Finally, traffic lights in the target area are optimized and controlled according to the signal optimization instructions. This solves the problem of poor real-time performance in existing traffic scheduling methods due to insufficient rigor and completeness, and effectively addresses traffic congestion. Attached Figure Description

[0010] Figure 1 This application provides a schematic diagram of a global intelligent scheduling method based on panoramic segmentation;

[0011] Figure 2 This application provides a schematic diagram of a global intelligent scheduling system structure based on panoramic segmentation.

[0012] Figure labeling: Data extraction module 11, data segmentation module 12, model building module 13, prediction set module 14, prediction bias module 15, optimization instruction module 16. Detailed Implementation

[0013] This application provides a comprehensive intelligent traffic scheduling method and system based on panoramic segmentation. It extracts road monitoring data from a target area according to preset time nodes to obtain traffic video data with acquisition location identifiers. Then, based on the acquisition locations and the road planning map of the target area, the traffic video data is segmented to generate N traffic video data sets. A congestion duration prediction model is constructed based on a backpropagation neural network. The N traffic video data sets are input into N congestion duration prediction channels for prediction, generating N vertical and N horizontal duration prediction sets. Based on these sets, N duration prediction deviations are calculated, and signal optimization instructions are generated. Finally, traffic lights in the target area are optimized and controlled according to these instructions. This solves the problem of poor real-time performance in existing traffic scheduling methods due to insufficient rigor and completeness, achieving an effective solution to traffic congestion issues.

[0014] Example 1

[0015] like Figure 1 As shown, this application provides a global intelligent scheduling method and system based on panoramic segmentation. The method includes:

[0016] Road monitoring data of the target area is extracted according to preset time nodes to obtain traffic video data, and the traffic video data is marked with the collection location.

[0017] Because road conditions vary at different times, traffic lights need to be adjusted accordingly. Therefore, it's necessary to acquire road monitoring data at different times, set preset time points, and extract traffic video data from the target area according to these preset time points. The location information of the video data acquisition devices is then obtained and used to identify the traffic video data, resulting in traffic video data with acquisition location identifiers. This provides the data foundation for subsequent segmentation of the traffic video data based on the acquisition location and the road planning map of the target area, generating N traffic video data sets, where N represents the number of traffic light-controlled intersections in the target area.

[0018] Based on the acquisition location, the traffic video data is segmented according to the road planning map of the target area to generate N traffic video data sets, where N is the number of traffic light controlled intersections in the target area.

[0019] Based on the data collection location, a road planning map of the target area is obtained. This map includes road map data. The road structure of the target area is then determined from this map data. This structure is correlated with traffic video data, which is then segmented according to the structure to obtain N traffic video data sets. The number of these sets corresponds to the number of traffic light-controlled intersections in the target area. A traffic light-controlled intersection is one that has traffic lights and uses them for traffic control. This acquisition of the traffic video data sets provides the data foundation for subsequently inputting these N sets into N congestion duration prediction channels to generate N longitudinal and N lateral duration prediction sets.

[0020] A congestion duration prediction model is constructed based on a BP neural network, wherein the congestion duration prediction model includes N congestion duration prediction channels;

[0021] Acquire multiple historical traffic video datasets and multiple historical congestion durations;

[0022] The multiple sets of historical traffic video data and the multiple historical congestion durations are used as training sample data;

[0023] Congestion duration prediction sub-channel based on BP neural network;

[0024] The congestion duration prediction sub-channel is trained and validated using the training sample data to obtain a congestion duration prediction sub-channel that meets the expected indicators.

[0025] Congestion duration prediction channel is constructed based on two congestion duration prediction sub-channels;

[0026] The congestion duration prediction channel is replicated to obtain N congestion duration prediction channels, and the congestion duration prediction model is formed based on the N congestion duration prediction channels.

[0027] Backpropagation (BP) neural networks are multi-layer feedforward neural networks trained using an error backpropagation algorithm, making them well-suited for constructing predictive models. Building a congestion duration prediction model using BP neural networks allows for faster calculation of congestion durations, improving prediction efficiency. The congestion prediction model consists of multiple congestion duration prediction channels, the number of which corresponds to the number of traffic light-controlled intersections in the target area. Multiple historical traffic video datasets and historical congestion durations are acquired. The historical traffic video data refers to previously acquired traffic video data, and the historical congestion durations are the corresponding congestion durations from those historical traffic video data. These historical traffic video datasets and historical congestion durations are then used as training sample data. A congestion duration prediction sub-channel is constructed based on a backpropagation (BP) neural network. The framework is built using a BP neural network, and the congestion duration correspondence is filled to obtain the prediction sub-channel. Supervised training and validation of the congestion duration prediction sub-channel are performed using training sample data, which includes both training and validation data. The output results are obtained by inputting the training data and comparing them with the validation data. The model parameters are adjusted based on the comparison results to obtain a congestion duration prediction sub-channel that meets the expected indicators. A congestion duration prediction channel is constructed based on the two congestion duration prediction sub-channels. This channel is then replicated to obtain more congestion duration prediction channels, i.e., N congestion duration prediction channels. A congestion duration prediction model is then formed based on these N channels. The acquisition of the congestion duration prediction model provides the data foundation for subsequently inputting N sets of traffic video data into the N congestion duration prediction channels for prediction, generating N vertical duration prediction sets and N horizontal duration prediction sets.

[0028] The N traffic video data sets are respectively input into the N congestion duration prediction channels for prediction, generating N vertical duration prediction sets and N horizontal duration prediction sets;

[0029] The N traffic video data sets are divided by route orientation to generate N longitudinal traffic video data sets and N transverse traffic video data sets;

[0030] The N sets of longitudinal traffic video data and the N sets of transverse traffic video data are respectively input into the congestion duration prediction sub-channels of the N congestion duration prediction channels to perform congestion market prediction, and output N sets of longitudinal duration prediction and N sets of transverse duration prediction.

[0031] When predicting congestion duration from N traffic video datasets, the datasets must first be divided. Each dataset can be further divided according to the direction of vehicle travel. Traffic conditions differ for different directions, requiring separate analysis to ensure accurate congestion prediction. The datasets are divided into horizontal and vertical traffic video datasets. These N datasets are then further divided by route orientation. The N vertical and N horizontal traffic video datasets are input into the congestion duration prediction sub-channels of N congestion duration prediction channels for congestion duration prediction, resulting in N vertical and N horizontal congestion duration prediction sets. By dividing the traffic video datasets and performing congestion duration prediction on the divided datasets, more accurate congestion predictions for different directions can be obtained, improving the overall system accuracy.

[0032] N duration prediction deviations are calculated based on the N vertical duration prediction sets and the N horizontal duration prediction sets;

[0033] The process involves acquiring both a longitudinal duration prediction set and a horizontal market prediction set. The actual longitudinal congestion duration corresponding to the longitudinal duration prediction set is then obtained. The longitudinal duration prediction set is compared with the actual longitudinal congestion duration to determine the difference between them, yielding the longitudinal prediction deviation. Similarly, the actual horizontal congestion duration corresponding to the horizontal duration prediction set is acquired. The horizontal duration prediction set is compared with the actual horizontal congestion duration to determine the difference between them, yielding the horizontal prediction deviation. The longitudinal and horizontal prediction deviations are then fused to obtain the fused result, the duration prediction deviation. This acquisition of duration prediction deviations provides the data foundation for generating signal optimization instructions based on N duration prediction deviations and for optimizing traffic light control in the target area according to these instructions.

[0034] Based on the N duration prediction deviations, a signal optimization instruction is generated, and the traffic lights in the target area are optimized and controlled according to the signal optimization instruction.

[0035] A deviation threshold is set and used as the pass / fail criterion for determining the duration prediction deviation. When the absolute value of the duration prediction deviation exceeds the preset deviation threshold, it indicates that the current traffic light control is ineffective, the current traffic situation is not under optimal control, and congestion may result. To optimize traffic light control in the target area, a signal optimization command is generated, acquired, and its parameters are read. The traffic lights in the target area are then optimized according to the parameters in the signal optimization command. This ensures that the traffic lights in the target area maintain the signal control state that maximizes the current traffic flow, significantly improving overall traffic efficiency.

[0036] Furthermore, the method also includes:

[0037] Retrieve the historical road congestion record data set for the target area, including the congestion time points;

[0038] Congestion duration thresholds are determined based on a set of congestion time nodes, and frequency statistics are performed on the congestion time nodes within the congestion duration threshold according to a preset unit duration to generate multiple time period congestion coefficients.

[0039] Using the congestion duration threshold as a constraint for extracting road monitoring data, multiple preset time nodes are generated based on the congestion coefficients of the multiple time periods.

[0040] The historical road congestion record data set for the target area is retrieved. This data includes congestion time nodes, which represent the time periods of actual congestion. A congestion duration threshold is determined based on this time node set—the maximum congestion duration. Within this threshold, the frequency of congestion time nodes is statistically analyzed according to a preset unit time length. Multiple time-segment congestion coefficients are generated based on these thresholds. Using the congestion duration threshold as a constraint for road monitoring data extraction, multiple preset time nodes are generated based on these congestion coefficients. The relationship between the preset nodes and the congestion coefficient is that a higher congestion coefficient corresponds to a shorter time node. By rationally allocating resources based on the time nodes and congestion coefficients, the overall system utilization efficiency can be improved.

[0041] Furthermore, the method also includes:

[0042] The road planning map is used to extract intersections, resulting in multiple intersections;

[0043] By selecting intersections that are traffic-controlled from multiple intersections, N traffic-controlled intersections are obtained.

[0044] Based on the location of the traffic video data collection, the locations of the N traffic light-controlled intersections are matched to complete the data segmentation.

[0045] When segmenting traffic video data, it is necessary to determine the corresponding segmentation areas based on the road planning map, and then complete the segmentation of the traffic video data based on the area determination results. First, intersection identification is performed on the road planning map to obtain intersection identification results. Based on the intersection identification results, intersections are extracted to obtain multiple intersections. Intersections containing traffic lights are selected from the multiple intersections and extracted as traffic light-controlled intersections, resulting in N traffic light-controlled intersections. Based on the acquisition location of the traffic video data, the N traffic light-controlled intersections are matched for location, completing the data segmentation. The traffic video data is segmented according to the road planning map of the target area, resulting in N traffic video data sets. This provides the data foundation for subsequently inputting the N traffic video data sets into N congestion duration prediction channels for prediction, generating N vertical duration prediction sets and N horizontal duration prediction sets.

[0046] Furthermore, the method also includes:

[0047] Construct a time deviation-signal optimization comparison table;

[0048] The N duration prediction deviations are input into the duration deviation-signal optimization lookup table to optimize signal control matching, resulting in N optimized signal control parameters.

[0049] The traffic lights in the target area are optimized and controlled accordingly.

[0050] Based on big data, duration deviation data and its corresponding signal optimization data are acquired. An initial dataset is constructed from the acquired duration deviation data and its corresponding signal optimization data. This initial dataset is then filtered, and duration deviation data and their corresponding signal optimization data that do not meet the criteria are removed to obtain a control dataset. A comparison table is constructed based on the control dataset to obtain a duration deviation-signal optimization comparison table. N duration prediction deviations are input into the duration deviation-signal optimization comparison table and traversed to obtain the duration deviations corresponding to the predicted market deviations. The optimized signals are then obtained from the comparison table based on the traversed duration deviations, and the optimized signals are output to obtain N optimized signal control parameters. The traffic lights in the target area are then optimized and controlled according to the optimized signal control parameters. By establishing a comparison table and obtaining the corresponding optimized control parameters, the overall optimization efficiency can be greatly improved, thereby improving the overall system efficiency.

[0051] Example 2

[0052] Based on the same inventive concept as the panoramic segmentation-based global intelligent scheduling method in the foregoing embodiments, such as Figure 2 As shown, this application provides a global intelligent scheduling system based on panoramic segmentation, the system comprising:

[0053] Data extraction module 11: The data extraction module 11 is used to extract road monitoring data of the target area according to a preset time node to obtain traffic video data, and the traffic video data has a collection location identifier;

[0054] Data segmentation module 12: The data segmentation module 12 is used to segment the traffic video data based on the acquisition location and the road planning map of the target area to generate N traffic video data sets, where N is the number of traffic light controlled intersections in the target area;

[0055] Model building module 13: The model building module 13 is used to build a congestion duration prediction model based on a BP neural network, wherein the congestion duration prediction model includes N congestion duration prediction channels;

[0056] Prediction Set Module 14: The prediction set module 14 is used to input the N traffic video data sets into the N congestion duration prediction channels respectively for prediction, and generate N vertical duration prediction sets and N horizontal duration prediction sets.

[0057] Prediction deviation module 15: The prediction deviation module 15 is used to calculate N duration prediction deviations based on the N longitudinal duration prediction sets and the N horizontal duration prediction sets;

[0058] Optimization instruction module 16: The optimization instruction module 16 is used to generate signal optimization instructions based on the N duration prediction deviations, and to optimize the control of traffic lights in the target area according to the signal optimization instructions.

[0059] Furthermore, the system also includes:

[0060] Retrieve the historical road congestion record data set for the target area, including the congestion time points;

[0061] Congestion duration thresholds are determined based on a set of congestion time nodes, and frequency statistics are performed on the congestion time nodes within the congestion duration threshold according to a preset unit duration to generate multiple time period congestion coefficients.

[0062] Using the congestion duration threshold as a constraint for extracting road monitoring data, multiple preset time nodes are generated based on the congestion coefficients of the multiple time periods.

[0063] Furthermore, the data segmentation module 12 includes the following execution steps:

[0064] The road planning map is used to extract intersections, resulting in multiple intersections;

[0065] By selecting intersections that are traffic-controlled from multiple intersections, N traffic-controlled intersections are obtained.

[0066] Based on the location of the traffic video data collection, the locations of the N traffic light-controlled intersections are matched to complete the data segmentation.

[0067] Furthermore, the system also includes:

[0068] Acquire multiple historical traffic video datasets and multiple historical congestion durations;

[0069] The multiple sets of historical traffic video data and the multiple historical congestion durations are used as training sample data;

[0070] Congestion duration prediction sub-channel based on BP neural network;

[0071] The congestion duration prediction sub-channel is trained and validated using the training sample data to obtain a congestion duration prediction sub-channel that meets the expected indicators.

[0072] Congestion duration prediction channel is constructed based on two congestion duration prediction sub-channels;

[0073] The congestion duration prediction channel is replicated to obtain N congestion duration prediction channels, and the congestion duration prediction model is formed based on the N congestion duration prediction channels.

[0074] Furthermore, the prediction set module 14 includes the following execution steps:

[0075] The N traffic video data sets are divided by route orientation to generate N longitudinal traffic video data sets and N transverse traffic video data sets;

[0076] The N sets of longitudinal traffic video data and the N sets of transverse traffic video data are respectively input into the congestion duration prediction sub-channels of the N congestion duration prediction channels to perform congestion market prediction, and output N sets of longitudinal duration prediction and N sets of transverse duration prediction.

[0077] Furthermore, the optimized instruction module 16 includes the following execution steps:

[0078] Construct a time deviation-signal optimization comparison table;

[0079] The N duration prediction deviations are input into the duration deviation-signal optimization lookup table to optimize signal control matching, resulting in N optimized signal control parameters.

[0080] The traffic lights in the target area are optimized and controlled accordingly.

[0081] Through the foregoing detailed description of a global intelligent scheduling method based on panoramic segmentation, those skilled in the art can clearly understand that this embodiment is a global intelligent scheduling method based on panoramic segmentation. As for the apparatus disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and relevant parts can be referred to the method section description.

[0082] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A global intelligent scheduling method based on panoramic segmentation, characterized in that, The method includes: Road monitoring data of the target area is extracted according to preset time nodes to obtain traffic video data, and the traffic video data is marked with the collection location. Based on the acquisition location, the traffic video data is segmented according to the road planning map of the target area to generate N traffic video data sets, where N is the number of traffic light controlled intersections in the target area. A congestion duration prediction model is constructed based on a BP neural network, wherein the congestion duration prediction model includes N congestion duration prediction channels; The N traffic video data sets are respectively input into the N congestion duration prediction channels for prediction, generating N vertical duration prediction sets and N horizontal duration prediction sets; N duration prediction deviations are calculated based on the N vertical duration prediction sets and the N horizontal duration prediction sets; Based on the N duration prediction deviations, a signal optimization instruction is generated, and the traffic lights in the target area are optimized and controlled according to the signal optimization instruction. The method further includes: Acquire multiple historical traffic video datasets and multiple historical congestion durations; The multiple sets of historical traffic video data and the multiple historical congestion durations are used as training sample data; Congestion duration prediction sub-channel based on BP neural network; The congestion duration prediction sub-channel is trained and validated using the training sample data to obtain a congestion duration prediction sub-channel that meets the expected indicators. Congestion duration prediction channel is constructed based on two congestion duration prediction sub-channels; The congestion duration prediction channel is cloned to obtain N congestion duration prediction channels, and the congestion duration prediction model is composed of the N congestion duration prediction channels; it also includes: The N traffic video data sets are divided by route orientation to generate N longitudinal traffic video data sets and N transverse traffic video data sets; The N sets of vertical traffic video data and the N sets of horizontal traffic video data are respectively input into the congestion duration prediction sub-channels of the N congestion duration prediction channels to perform congestion market prediction, and output N sets of vertical duration prediction and N sets of horizontal duration prediction. This also includes constructing a duration deviation-signal optimization comparison table; The N duration prediction deviations are input into the duration deviation-signal optimization lookup table to optimize signal control matching, resulting in N optimized signal control parameters. The traffic lights in the target area are optimized and controlled accordingly.

2. The method as described in claim 1, characterized in that, The method further includes: Retrieve the historical road congestion record data set for the target area, including the congestion time points; Congestion duration thresholds are determined based on a set of congestion time nodes, and frequency statistics are performed on the congestion time nodes within the congestion duration threshold according to a preset unit duration to generate multiple time period congestion coefficients. Using the congestion duration threshold as a constraint for extracting road monitoring data, multiple preset time nodes are generated based on the congestion coefficients of the multiple time periods.

3. The method as described in claim 1, characterized in that, The step of segmenting the traffic video data based on the road planning map of the target area also includes: The road planning map is used to extract intersections, resulting in multiple intersections; By selecting intersections that are traffic-controlled from multiple intersections, N traffic-controlled intersections are obtained. Based on the location of the traffic video data collection, the locations of the N traffic light-controlled intersections are matched to complete the data segmentation.

4. A global intelligent scheduling system based on panoramic segmentation, used to execute the method of claim 1, characterized in that, The system includes: Data extraction module: Extracts road monitoring data of the target area according to preset time nodes to obtain traffic video data, and the traffic video data is marked with the collection location. Data segmentation module: Based on the acquisition location, the traffic video data is segmented according to the road planning map of the target area to generate N traffic video data sets, where N is the number of traffic light controlled intersections in the target area; Model building module: Congestion duration prediction model is built based on BP neural network, wherein the congestion duration prediction model includes N congestion duration prediction channels; Prediction Set Module: Input the N traffic video data sets into the N congestion duration prediction channels respectively for prediction, and generate N vertical duration prediction sets and N horizontal duration prediction sets; Prediction deviation module: Calculates N duration prediction deviations based on the N longitudinal duration prediction sets and the N horizontal duration prediction sets; Optimization instruction module: Generates signal optimization instructions based on the N duration prediction deviations, and optimizes the control of traffic lights in the target area according to the signal optimization instructions.

Images