Intelligent traffic signal lamp scheduling method and system based on pedestrian flow monitoring

Abstract

The invention provides an intelligent traffic signal lamp scheduling method and system based on pedestrian flow monitoring, and relates to the field of intelligent traffic, and the method comprises the steps: carrying out the analysis according to an image collection result, and obtaining a real-time pedestrian flow and a plurality of related pedestrian flows; inputting the real-time pedestrian flow and the plurality of associated pedestrian flows into a pedestrian flow prediction model, and outputting a predicted pedestrian flow at a future moment; and performing passing time simulation to obtain pedestrian simulation passing duration, and setting the pedestrian simulation passing duration as a green light control parameter at a future moment to manage and control traffic lights at the target intersection. The objective of the invention is to solve the technical problems that a conventional signal lamp scheduling method cannot be dynamically adjusted according to real-time pedestrian flow, traffic flow imbalance is easily caused, and road congestion is caused. Through pedestrian flow prediction based on a prediction model, in combination with simulation time, green light duration is reasonably set, dynamic regulation and control of traffic lights are realized, waiting time is reduced, and traffic fluency is improved.

Description

Technical Field

[0001] The present invention relates to the field of smart transportation, and in particular to a method and system for scheduling smart traffic lights based on pedestrian flow monitoring. Background Art

[0002] Traditional traffic light scheduling methods are usually designed based on fixed time patterns or traffic flow patterns, and cannot be dynamically adjusted based on the real-time pedestrian flow at a specific intersection. When the pedestrian flow at a certain time period or intersection suddenly increases, the signal light control cannot be adjusted in time, resulting in uneven traffic flow and, in turn, road congestion. In addition, because traditional traffic light scheduling cannot accurately control the length of the green light for pedestrians to cross the road, the signal light often switches too early, causing pedestrians to fail to cross safely, or the signal light does not switch for a long time during heavy traffic hours, causing pedestrians to stay for a long time, increasing safety hazards. Summary of the invention

[0003] The purpose of the present invention is to provide an intelligent traffic light scheduling method and system based on pedestrian flow monitoring, which is used to solve the technical problem that the traditional traffic light scheduling method cannot be dynamically adjusted according to the real-time pedestrian flow, which easily leads to uneven traffic flow and road congestion, including:

[0004] In the first aspect, the present invention provides an intelligent traffic signal scheduling method based on pedestrian flow monitoring, comprising: at a predetermined monitoring node, performing image acquisition on a target intersection and multiple related intersections, and obtaining real-time pedestrian flow and multiple related pedestrian flows based on image acquisition result analysis; constructing a pedestrian flow prediction model, and inputting the real-time pedestrian flow and multiple related pedestrian flows into the pedestrian flow prediction model, and outputting predicted pedestrian flow at a future moment; performing passing time simulation based on the predicted pedestrian flow, obtaining pedestrian simulation passing time, and setting the pedestrian simulation passing time as a green light control parameter at the future moment; at the future moment, controlling the traffic light at the target intersection based on the green light control parameter.

[0005] Preferably, the intelligent traffic light scheduling method based on pedestrian flow monitoring also includes: obtaining image acquisition results, wherein the image acquisition results include a target image and multiple associated images; inputting the target image and multiple associated images into a pedestrian flow recognition model respectively, and outputting real-time pedestrian flow and multiple associated pedestrian flows.

[0006] Preferably, the intelligent traffic signal scheduling method based on pedestrian flow monitoring also includes: constructing a pedestrian flow recognition model based on a convolutional neural network; collecting a sample pedestrian image set according to the historical pedestrian monitoring records of the area, and marking the pedestrian flow in different sample pedestrian images to obtain a sample pedestrian flow set; taking the sample pedestrian image as input and the sample pedestrian flow as output, and using the sample pedestrian image set and the sample pedestrian flow set to train the pedestrian flow recognition model until convergence.

[0007] Preferably, the intelligent traffic light scheduling method based on pedestrian flow monitoring also includes: collecting a sample pedestrian flow set and multiple sample associated pedestrian flow sets based on the historical pedestrian monitoring records of the target road; obtaining the pedestrian flow of the target road at historical moments under scenarios with different sample pedestrian flow sets and sample associated pedestrian flow sets to obtain a sample predicted pedestrian flow set; using the sample pedestrian flow set, multiple sample associated pedestrian flow sets and sample predicted pedestrian flow set to train the BP neural network until convergence to obtain the pedestrian flow prediction model.

[0008] Preferably, the intelligent traffic light scheduling method based on pedestrian flow monitoring also includes: collecting a sample pedestrian flow set and a sample passing time set based on the historical pedestrian monitoring records of the target road; calculating the pedestrian passing flow in multiple unit times based on the sample pedestrian flow set and the sample passing time set, and calculating the pedestrian passing flow in a standard unit time by the average; dividing the predicted pedestrian flow by the pedestrian passing flow in the standard unit time, and outputting the pedestrian simulation passing time.

[0009] Preferably, the intelligent traffic light scheduling method based on pedestrian flow monitoring also includes: taking the pedestrian flow in the standard unit time as a benchmark, performing deviation amplitude analysis on the pedestrian flow in the multiple unit times respectively, and determining multiple deviation ratios; calculating an error compensation coefficient based on the multiple deviation ratios; using the error compensation coefficient to compensate for the pedestrian simulation passing time, and obtaining a corrected pedestrian simulation passing time, which is set as the green light control parameter at the future moment.

[0010] Preferably, the intelligent traffic light scheduling method based on pedestrian flow monitoring also includes: arranging the multiple deviation ratios from large to small to generate a deviation ratio sequence; selecting the first fifth of the deviation ratios in the deviation ratio sequence, and setting the average value as the error compensation coefficient after calculating the average value.

[0011] In the second aspect, the present invention also provides an intelligent traffic signal light scheduling system based on pedestrian flow monitoring, which is used to execute an intelligent traffic signal light scheduling method based on pedestrian flow monitoring as described in the first aspect, including: an image acquisition result analysis module, which is used to perform image acquisition on a target intersection and multiple related intersections at a predetermined monitoring node, and obtain real-time pedestrian flow and multiple related pedestrian flows based on the image acquisition result analysis; a pedestrian flow prediction module, which is used to construct a pedestrian flow prediction model, and input the real-time pedestrian flow and multiple related pedestrian flows into the pedestrian flow prediction model, and output the predicted pedestrian flow at a future moment; a green light control parameter setting module, which is used to simulate the passing time according to the predicted pedestrian flow, obtain the pedestrian simulation passing time, and set the pedestrian simulation passing time as the green light control parameter at the future moment; a traffic light management and control module, which is used to control the traffic light at the target intersection according to the green light control parameter at the future moment.

[0012] The embodiments of the present invention include the following advantages:

[0013] By collecting images of a target intersection and multiple related intersections at a predetermined monitoring node, real-time pedestrian flow and multiple related pedestrian flow rates are obtained based on analysis of the image collection results; a pedestrian flow prediction model is constructed, and the real-time pedestrian flow and multiple related pedestrian flow rates are input into the pedestrian flow prediction model to output the predicted pedestrian flow rate at a future moment; a passing time simulation is performed based on the predicted pedestrian flow rate to obtain the simulated pedestrian passing time, and the simulated pedestrian passing time is set as the green light control parameter at the future moment; at the future moment, the traffic lights at the target intersection are controlled based on the green light control parameter. That is to say, by predicting the pedestrian flow based on the prediction model and combining the simulation time, the green light duration is reasonably set to achieve dynamic regulation of traffic lights, reduce waiting time, and improve traffic fluency. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 A flowchart of the steps of an intelligent traffic signal light scheduling method based on pedestrian flow monitoring of the present invention;

[0015] Figure 2 The present invention is a schematic structural diagram of an intelligent traffic signal light dispatching system based on pedestrian flow monitoring.

[0016] Description of reference numerals:

[0017] Image acquisition result analysis module 11, pedestrian flow prediction module 12, green light control parameter setting module 13, traffic light management and control module 14. DETAILED DESCRIPTION

[0018] The present invention provides an intelligent traffic light scheduling method and system based on pedestrian flow monitoring, which solves the technical problem that the traditional traffic light scheduling method cannot be dynamically adjusted according to the real-time pedestrian flow, which easily leads to uneven traffic flow and road congestion. Through pedestrian flow prediction based on the prediction model, combined with simulation time, the green light duration is reasonably set, dynamic regulation of traffic lights is achieved, waiting time is reduced, and traffic fluency is improved.

[0019] Below, the technical solutions in the present invention will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments of the present invention. It should be understood that the present invention is not limited to the example embodiments described herein. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. It should also be noted that, for the convenience of description, only the parts related to the present invention are shown in the accompanying drawings, rather than all of them.

[0020] For example, please refer to the attached Figure 1 The present invention provides an intelligent traffic light scheduling method based on pedestrian flow monitoring, which is applied to an intelligent traffic light scheduling system based on pedestrian flow monitoring, and specifically includes the following steps:

[0021] Step 101: Under a predetermined monitoring node, image acquisition is performed on a target intersection and multiple related intersections, and real-time pedestrian flow and multiple related pedestrian flows are acquired based on image acquisition result analysis; Step 102: A pedestrian flow prediction model is constructed, and the real-time pedestrian flow and multiple related pedestrian flows are input into the pedestrian flow prediction model, and the predicted pedestrian flow at a future moment is output; Step 103: A passing time simulation is performed based on the predicted pedestrian flow, and the pedestrian simulation passing time is acquired, and the pedestrian simulation passing time is set as the green light control parameter at the future moment; Step 104: At the future moment, the traffic light at the target intersection is controlled based on the green light control parameter.

[0022] Specifically, at a predetermined monitoring node, images are collected at a target intersection and multiple related intersections. For example, a camera or image sensor is used to collect images at these monitoring nodes. Through image analysis, the real-time pedestrian flow at the current intersection can be obtained, that is, the number of pedestrians passing or waiting to cross the road. Image analysis uses computer vision technology (such as target detection, image segmentation, etc.) to identify pedestrians from images taken by the camera, and then counts the pedestrian flow at each monitoring node. In addition to the target intersection, pedestrian flow data of other intersections associated with the target intersection will also be obtained. The data of these associated intersections can also help the system to have a more comprehensive understanding of the surrounding traffic conditions, thereby making more accurate scheduling decisions; real-time pedestrian flow and multiple associated pedestrian flows are obtained.

[0023] Then, in order to predict future traffic demand in advance, it is necessary to build a pedestrian flow prediction model based on historical data and real-time monitoring data. This model usually uses machine learning or deep learning algorithms, such as regression analysis, time series prediction models or other prediction techniques, to model and predict pedestrian flow. Then the real-time pedestrian flow and multiple associated pedestrian flows are input into the pedestrian flow prediction model, and the predicted pedestrian flow at the future moment is output. The prediction result will give the expected pedestrian flow at the target intersection at a certain time or time period in the future. Further, based on the predicted future pedestrian flow, the system will simulate the pedestrian passing time, that is, estimate the time required for pedestrians to pass through the target intersection. This simulation process takes into account factors such as the speed of pedestrians crossing the road, pedestrian density, and intersection width; based on the simulated passing time results, the system sets the green light duration at the future moment, which means that if the predicted future pedestrian flow is large, the system may appropriately extend the green light time; if the predicted flow is small, the green light time can be shortened. The simulation result of this passing time becomes the green light control parameter at the future moment.

[0024] Based on the green light duration calculated in the previous step, the system will automatically adjust the signal light switching strategy in the future. For example, when it predicts that a large number of pedestrians are about to pass, the system may extend the duration of the green light to ensure that pedestrians can pass the intersection safely. At this time, the switching of red and green lights will be intelligently adjusted according to the model's prediction and control parameters to avoid pedestrians waiting too long, while ensuring smooth traffic and avoiding traffic congestion caused by too long red lights.

[0025] Furthermore, the present invention further comprises the following steps:

[0026] Obtain image acquisition results, wherein the image acquisition results include a target image and multiple associated images; input the target image and the multiple associated images into a pedestrian flow recognition model respectively, and output real-time pedestrian flow and multiple associated pedestrian flow.

[0027] Specifically, an image acquisition result is obtained, wherein the image acquisition result includes a target image and multiple associated images, the target image refers to image data from a target intersection, that is, a real-time video or image of an intersection (target intersection) that the system mainly focuses on, and these images contain information about pedestrians passing or waiting at the target intersection; associated images refer to images from other intersections (or areas) associated with the target intersection. These associated intersections can be intersections in the traffic network that are adjacent to or have an impact on the target intersection. For example, if there are several intersections or streets near the target intersection, the associated images are real-time image data of these intersections, which help the system to more comprehensively understand the pedestrian flow of the entire traffic network. Then the target image and multiple associated images are respectively input into the pedestrian flow recognition model, and real-time pedestrian flow and multiple associated pedestrian flows are output.

[0028] Furthermore, the present invention further comprises the following steps:

[0029] A pedestrian flow recognition model is constructed based on a convolutional neural network; a sample pedestrian image set is collected according to the historical pedestrian monitoring records in the area, and the pedestrian flow in different sample pedestrian images is annotated to obtain a sample pedestrian flow set; with the sample pedestrian images as input and the sample pedestrian flow as output, the sample pedestrian image set and the sample pedestrian flow set are used to train the pedestrian flow recognition model until convergence.

[0030] Specifically, convolutional neural network (CNN) is a deep learning model that is particularly good at processing image data. Through the combination of convolutional layers, pooling layers, and fully connected layers, CNN can automatically extract features from images and perform classification or regression tasks. For the problem of pedestrian flow recognition, CNN can automatically learn and identify pedestrians from the input image and predict the number of pedestrians in the image. The goal of the pedestrian flow recognition model built by CNN is to identify pedestrians from the input image and estimate the number of pedestrians in the image. This model will be able to output pedestrian flow data in the image based on real-time images of different intersections.

[0031] Next, based on the historical pedestrian monitoring records in the area, a sample pedestrian image set is collected. The sample pedestrian image set is obtained by analyzing the historical monitoring data and selecting pedestrian images at different times, locations and scenes. These image sets include pedestrian images collected under different weather conditions, different intersections, different time periods and other conditions, which can reflect the behavior patterns of pedestrians in different situations. Then, the pedestrian flow in different sample pedestrian images is annotated to obtain a sample pedestrian flow set. The number of pedestrians contained in each sample image needs to be manually annotated. The annotation process refers to manually recording the number of pedestrians in each image or obtaining the accurate number of pedestrians through other automated methods. The annotation result is called a sample pedestrian flow set, that is, for each image, the corresponding number of pedestrians is annotated. When annotating, each pedestrian is usually manually circled in the image, or the image segmentation technology is used to automatically extract the boundaries of the pedestrians and calculate the number of pedestrians contained in the image. At this time, the sample image and the corresponding number of pedestrians (annotation value) will form a one-to-one correspondence.

[0032] Further, the sample pedestrian images are used as input and the sample pedestrian flow is used as output, and the sample pedestrian image set and the sample pedestrian flow set are used to train the pedestrian flow recognition model. That is, the purpose of using sample data for model training is to allow the convolutional neural network to learn how to predict the number of pedestrians in the image based on the input image. During the training process, CNN optimizes its parameters through the back propagation algorithm so that the prediction results of the model are as close to the actual labeled values ​​as possible; the training is carried out until the loss function of the model no longer decreases significantly, that is, the model reaches the convergence state of the training. At this time, the convolutional neural network has learned to estimate the number of pedestrians from the input image and has a certain generalization ability, and can accurately predict new images; and a pedestrian flow recognition model is obtained.

[0033] Furthermore, the present invention further comprises the following steps:

[0034] According to the historical pedestrian monitoring records of the target road, a sample pedestrian flow set and multiple sample associated pedestrian flow sets are collected; the pedestrian flow of the target road at historical moments in the scenarios of different sample pedestrian flow and sample associated pedestrian flow sets are obtained to obtain a sample predicted pedestrian flow set; the sample pedestrian flow set, multiple sample associated pedestrian flow sets and sample predicted pedestrian flow set are used to train the BP neural network until convergence to obtain the pedestrian flow prediction model.

[0035] Specifically, according to the historical pedestrian monitoring records of the target road, based on the historical pedestrian monitoring records of the target road (for example, pedestrian flow data of the past few months), the pedestrian flow data of the target road in different time periods are collected. The sample pedestrian flow set is usually composed of pedestrian flow records at historical moments, reflecting the number of pedestrians on the road at different times; in addition to the pedestrian flow data of the target road, the pedestrian flow data of other sections related to the target road must also be collected. The data of these related sections usually reflect the pedestrian flow conditions in the vicinity of the target road, and they may have an impact on the traffic flow of the target road. By collecting the pedestrian flow data of these related sections, more contextual information can be provided to the model to help predict the pedestrian flow of the target road.

[0036] Next, the pedestrian flow of the target road at historical moments in the scenarios of different sample pedestrian flow and sample associated pedestrian flow sets is obtained. By using historical data (pedestrian flow of the target road and pedestrian flow data of associated road sections), the pedestrian flow of the target road at future moments or historical moments is calculated or estimated as the predicted target value. At this time, the sample predicted pedestrian flow set is based on the analysis results of historical data and associated data, and represents the predicted pedestrian flow value of the target road under certain specific scenarios, thereby obtaining the sample predicted pedestrian flow set.

[0037] BP neural network is a commonly used feedforward neural network structure, which consists of an input layer, a hidden layer and an output layer. Its main feature is to optimize the network weights through the back propagation algorithm. In the pedestrian flow prediction task, the BP neural network can learn based on the input historical pedestrian flow data (including the flow of the target road and the flow of the associated road sections) to predict the pedestrian flow at future times. Then, the sample pedestrian flow set, multiple sample associated pedestrian flow sets and sample predicted pedestrian flow sets are used to train the BP neural network until convergence, wherein the BP neural network continuously updates the network weights through the back propagation algorithm to minimize the loss function until the model converges, and trains until the loss function of the model no longer decreases significantly, that is, the network has converged, at which time the model has a high accuracy and can be used for actual pedestrian flow prediction.

[0038] Furthermore, the present invention further comprises the following steps:

[0039] According to the historical pedestrian monitoring records of the target road, a sample pedestrian flow set and a sample passing time set are collected; the pedestrian passing flow in multiple unit times is calculated respectively according to the sample pedestrian flow set and the sample passing time set, and the pedestrian passing flow in the standard unit time is calculated by the average; the predicted pedestrian flow is divided by the pedestrian passing flow in the standard unit time, and the pedestrian simulation passing time is output.

[0040] Specifically, based on the historical pedestrian monitoring records of the target road, a sample pedestrian flow set and a sample passing time set are collected. The sample pedestrian flow set contains pedestrian flow information on the target road in different time periods. These data can reflect the number of pedestrians on the target road at different daily times (such as morning peak, evening peak, holidays, etc.). The sample pedestrian flow set usually includes the number of pedestrians or flow data at different time points, which are usually obtained through image acquisition, sensor monitoring, etc.; the sample passing time set contains the time information of pedestrians passing through the target road, for example, the time it takes for some pedestrians to enter a certain intersection and completely pass through the intersection. This data set reflects the time it takes for pedestrians to pass through the target road in different time periods. This time data is obtained based on historical records or actual observations.

[0041] Next, the pedestrian flow in multiple unit times is calculated based on the sample pedestrian flow set and the sample passing time set, and the pedestrian flow in the standard unit time is calculated by the average. The pedestrian flow in the unit time refers to the number of pedestrians passing through the target road in a given time period. For example, the number of pedestrians passing through the intersection in a certain period of time (such as 10 minutes) can be used to measure traffic flow; based on the sample pedestrian flow set and the sample passing time set, the number of pedestrians passing through the target road per unit time (such as per second, per minute, etc.) in a certain time period can be calculated. For example, assuming that the number of pedestrians passing through the intersection in 10 minutes is 100, the pedestrian flow per unit time in this time period is 100 people / 10 minutes. In order to obtain a more stable and reliable standard value, the pedestrian flow in the standard unit time is usually determined by calculating the average of the pedestrian flow in multiple unit times. For example, multiple unit time flow calculations are performed for the data of the past few months, and the average value is obtained as the standard unit time flow. Finally, the predicted pedestrian flow is divided by the pedestrian flow in the standard unit time, and the pedestrian simulation passing time is output. The pedestrian simulation passing time represents the time required for all pedestrians to pass the target intersection under the predicted pedestrian flow. By dividing the predicted pedestrian flow by the pedestrian flow in the standard unit time, the total time required for all pedestrians to pass the intersection in the target time period can be calculated. Finally, the calculated pedestrian simulation passing time can be used for traffic light scheduling decisions. According to the pedestrian flow and the required passing time, the system can dynamically adjust the duration of the green light to ensure the safe passage of pedestrians and optimize traffic flow.

[0042] Furthermore, the present invention further comprises the following steps:

[0043] Taking the pedestrian flow rate within the standard unit time as a benchmark, the pedestrian flow rate within the multiple unit times is respectively analyzed for deviation amplitudes to determine multiple deviation ratios; an error compensation coefficient is calculated based on the multiple deviation ratios; the pedestrian simulation passing time is compensated using the error compensation coefficient to obtain a corrected pedestrian simulation passing time, which is set as the green light control parameter at the future moment.

[0044] Specifically, taking the pedestrian flow rate in the standard unit time as a benchmark, the pedestrian flow rates in the multiple unit times are analyzed for deviation amplitudes, the deviation amplitude refers to the difference between the actual pedestrian flow rate in each unit time and the standard pedestrian flow rate per unit time, the deviation amplitude analysis is to understand the difference amplitude between the actual flow rate and the standard flow rate in different time periods; multiple deviation ratios are determined, the deviation ratio is obtained by calculating the ratio of the deviation amplitude in each unit time to the standard unit time flow rate, which represents the relative difference between the actual pedestrian flow rate and the standard pedestrian flow rate.

[0045] Next, an error compensation coefficient is calculated based on the multiple deviation ratios. The error compensation coefficient can be a weighted average of multiple deviation ratios or the result of other statistical methods. An overall compensation coefficient is calculated based on the deviation ratios in each time period. For example, if the values ​​of multiple deviation ratios are high, the error compensation coefficient may increase accordingly, indicating that the green light time needs to be extended to accommodate a larger pedestrian flow. Finally, the pedestrian simulation passing time is compensated using the error compensation coefficient to obtain a corrected pedestrian simulation passing time. By using the error compensation coefficient, the simulated passing time can be corrected. If the pedestrian flow is greater than the standard flow (the deviation ratio is large), the corrected passing time will increase, and vice versa. The corrected pedestrian simulation passing time is set as the green light control parameter at the future moment.

[0046] Furthermore, the present invention further comprises the following steps:

[0047] Arrange the multiple deviation ratios from large to small to generate a deviation ratio sequence; select the first fifth of the deviation ratios in the deviation ratio sequence, calculate the average and set them as the error compensation coefficient.

[0048] Specifically, the multiple deviation ratios are arranged from large to small to generate a deviation ratio sequence. This sorting can help us identify which time periods have larger pedestrian flow deviations and which time periods have smaller deviations. The sorted sequence provides deviation data arranged by deviation size, which is convenient for subsequent analysis. Then, in the sorted deviation ratio sequence, the first fifth of the deviation ratio refers to the data of the first 20% (one fifth) of the deviation ratio sequence. This part represents some time periods with the largest deviations, usually time periods with more significant flow deviations. By selecting these larger deviation ratios, it can help identify the most critical time periods that affect traffic light scheduling. Based on the first fifth of the data in the selected deviation ratio sequence, their mean is calculated as the error compensation coefficient. The mean calculation is the average value obtained by adding up these deviation ratios and dividing them by their number.

[0049] In summary, the intelligent traffic signal light scheduling method based on pedestrian flow monitoring provided by the present invention has the following technical effects:

[0050] By collecting images of a target intersection and multiple related intersections at a predetermined monitoring node, real-time pedestrian flow and multiple related pedestrian flow rates are obtained based on analysis of the image collection results; a pedestrian flow prediction model is constructed, and the real-time pedestrian flow and multiple related pedestrian flow rates are input into the pedestrian flow prediction model to output the predicted pedestrian flow rate at a future moment; a passing time simulation is performed based on the predicted pedestrian flow rate to obtain the simulated pedestrian passing time, and the simulated pedestrian passing time is set as the green light control parameter at the future moment; at the future moment, the traffic lights at the target intersection are controlled based on the green light control parameter. That is to say, by predicting the pedestrian flow based on the prediction model and combining the simulation time, the green light duration is reasonably set to achieve dynamic regulation of traffic lights, reduce waiting time, and improve traffic fluency.

[0051] Embodiment 2, based on the same inventive concept as the intelligent traffic light dispatching method based on pedestrian flow monitoring in the aforementioned embodiment, the present invention also provides an intelligent traffic light dispatching system based on pedestrian flow monitoring, please refer to the attached Figure 2 ,include:

[0052] The image acquisition result analysis module 11 is used to acquire images of the target intersection and multiple related intersections at a predetermined monitoring node, and obtain real-time pedestrian flow and multiple related pedestrian flow according to the image acquisition result analysis; the pedestrian flow prediction module 12 is used to construct a pedestrian flow prediction model, and input the real-time pedestrian flow and multiple related pedestrian flow into the pedestrian flow prediction model, and output the predicted pedestrian flow at a future time; the green light control parameter setting module 13 is used to simulate the passing time according to the predicted pedestrian flow, obtain the pedestrian simulation passing time, and set the pedestrian simulation passing time as the green light control parameter at the future time; the traffic light control module 14 is used to control the traffic light at the target intersection according to the green light control parameter at the future time.

[0053] Furthermore, the intelligent traffic light scheduling system based on pedestrian flow monitoring is also used to: obtain image acquisition results, wherein the image acquisition results include a target image and multiple associated images; input the target image and multiple associated images into a pedestrian flow recognition model respectively, and output real-time pedestrian flow and multiple associated pedestrian flows.

[0054] Furthermore, the intelligent traffic signal light dispatching system based on pedestrian flow monitoring is also used to: construct a pedestrian flow recognition model based on a convolutional neural network; collect a sample pedestrian image set according to the historical pedestrian monitoring records of the area, and mark the pedestrian flow in different sample pedestrian images to obtain a sample pedestrian flow set; use sample pedestrian images as input and sample pedestrian flow as output, and use the sample pedestrian image set and the sample pedestrian flow set to train the pedestrian flow recognition model until convergence.

[0055] Furthermore, the intelligent traffic light dispatching system based on pedestrian flow monitoring is also used to: collect a sample pedestrian flow set and multiple sample associated pedestrian flow sets based on historical pedestrian monitoring records of the target road; obtain the pedestrian flow of the target road at historical moments under scenarios with different sample pedestrian flow and sample associated pedestrian flow sets to obtain a sample predicted pedestrian flow set; use the sample pedestrian flow set, multiple sample associated pedestrian flow sets and sample predicted pedestrian flow set to train the BP neural network until convergence to obtain the pedestrian flow prediction model.

[0056] Furthermore, the intelligent traffic light scheduling system based on pedestrian flow monitoring is also used to: collect sample pedestrian flow sets and sample passing time sets according to historical pedestrian monitoring records of the target road; calculate the pedestrian passing flow in multiple unit times according to the sample pedestrian flow sets and sample passing time sets, and calculate the pedestrian passing flow in a standard unit time by the average; divide the predicted pedestrian flow by the pedestrian passing flow in the standard unit time, and output the pedestrian simulation passing time.

[0057] Furthermore, the intelligent traffic light dispatching system based on pedestrian flow monitoring is also used to: take the pedestrian flow in the standard unit time as a benchmark, perform deviation amplitude analysis on the pedestrian flow in the multiple unit times respectively, and determine multiple deviation ratios; calculate an error compensation coefficient based on the multiple deviation ratios; use the error compensation coefficient to compensate for the pedestrian simulation passing time, obtain the corrected pedestrian simulation passing time, and set it as the green light control parameter at the future moment.

[0058] Furthermore, the intelligent traffic light dispatching system based on pedestrian flow monitoring is also used to: arrange the multiple deviation ratios from large to small to generate a deviation ratio sequence; select the first fifth of the deviation ratios in the deviation ratio sequence, and set them as the error compensation coefficient after average calculation.

[0059] Each embodiment in this specification is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The intelligent traffic light scheduling method based on pedestrian flow monitoring and the specific examples in the aforementioned embodiment 1 are also applicable to the intelligent traffic light scheduling system based on pedestrian flow monitoring in this embodiment. Through the aforementioned detailed description of the intelligent traffic light scheduling method based on pedestrian flow monitoring, those skilled in the art can clearly know the intelligent traffic light scheduling system based on pedestrian flow monitoring in this embodiment, so for the sake of brevity of the specification, it will not be described in detail here. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description.

[0060] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

[0061] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention belong to the scope of the present invention and its equivalent technologies, the present invention is also intended to include these changes and variations.

Claims

1. An intelligent traffic light scheduling method based on pedestrian flow monitoring, characterized in that: include: At a predetermined monitoring node, image acquisition is performed on the target intersection and multiple associated intersections, and real-time pedestrian flow and multiple associated pedestrian flow are acquired based on image acquisition results analysis; Constructing a pedestrian flow prediction model, and inputting the real-time pedestrian flow and a plurality of associated pedestrian flows into the pedestrian flow prediction model, and outputting the predicted pedestrian flow at a future moment; Performing a passing time simulation according to the predicted pedestrian flow, obtaining the pedestrian simulated passing time, and setting the pedestrian simulated passing time as a green light control parameter at the future moment; At the future moment, the traffic light at the target intersection is controlled according to the green light control parameters.

2. The intelligent traffic signal light scheduling method based on pedestrian flow monitoring according to claim 1 is characterized in that: According to the image acquisition results, the real-time pedestrian flow and multiple related pedestrian flows are obtained, including: Acquire an image acquisition result, wherein the image acquisition result includes a target image and a plurality of associated images; The target image and multiple associated images are respectively input into a pedestrian flow recognition model, and real-time pedestrian flow and multiple associated pedestrian flow are output.

3. The intelligent traffic signal light scheduling method based on pedestrian flow monitoring according to claim 2 is characterized in that: Build a pedestrian flow recognition model, including: Build a pedestrian flow recognition model based on convolutional neural network; According to the historical pedestrian monitoring records in the area, a sample pedestrian image set is collected, and the pedestrian flow in different sample pedestrian images is annotated to obtain a sample pedestrian flow set; The sample pedestrian images are used as input and the sample pedestrian flow is used as output, and the pedestrian flow recognition model is trained using the sample pedestrian image set and the sample pedestrian flow set until convergence.

4. The intelligent traffic signal light dispatching method based on pedestrian flow monitoring according to claim 1 is characterized in that: Build a pedestrian flow prediction model, including: According to the historical pedestrian monitoring records of the target road, a sample pedestrian flow set and multiple sample associated pedestrian flow sets are collected; Obtain the pedestrian flow of the target road at historical moments in scenarios with different sample pedestrian flow and sample-associated pedestrian flow sets, and obtain a sample predicted pedestrian flow set; The sample pedestrian flow set, multiple sample associated pedestrian flow sets and sample predicted pedestrian flow set are used to train the BP neural network until convergence, thereby obtaining the pedestrian flow prediction model.

5. The intelligent traffic signal light scheduling method based on pedestrian flow monitoring according to claim 1 is characterized in that: The passing time is simulated according to the predicted pedestrian flow to obtain the pedestrian simulation passing time, including: According to the historical pedestrian monitoring records of the target road, a sample pedestrian flow set and a sample passing time set are collected; According to the sample pedestrian flow set and the sample passing time set, pedestrian passing flow in multiple unit times is calculated respectively, and the pedestrian passing flow in the standard unit time is calculated by average; The predicted pedestrian flow is divided by the pedestrian flow in the standard unit time, and the pedestrian simulation passage time is output.

6. The intelligent traffic signal light dispatching method based on pedestrian flow monitoring according to claim 5 is characterized in that: The pedestrian simulation passing time is set as the green light control parameter at the future moment, and the method also includes: Taking the pedestrian flow rate in the standard unit time as a benchmark, respectively analyzing the deviation amplitude of the pedestrian flow rates in the multiple unit times to determine multiple deviation ratios; Calculate an error compensation coefficient according to the multiple deviation ratios; The error compensation coefficient is used to compensate the pedestrian simulation passing time to obtain a corrected pedestrian simulation passing time, which is set as the green light control parameter at the future moment.

7. The intelligent traffic signal light dispatching method based on pedestrian flow monitoring according to claim 6 is characterized in that: The error compensation coefficient is calculated according to the multiple deviation ratios, including: Arrange the multiple deviation ratios from large to small to generate a deviation ratio sequence; The first fifth of the deviation ratios in the deviation ratio sequence are selected, and the average is calculated and set as the error compensation coefficient.

8. Intelligent traffic light dispatching system based on pedestrian flow monitoring, characterized in that: The steps for implementing the intelligent traffic light scheduling method based on pedestrian flow monitoring according to any one of claims 1 to 7 include: An image acquisition result analysis module is used to acquire images of a target intersection and multiple associated intersections at a predetermined monitoring node, and obtain real-time pedestrian flow and multiple associated pedestrian flow according to the image acquisition result analysis; A pedestrian flow prediction module is used to construct a pedestrian flow prediction model, input the real-time pedestrian flow and multiple associated pedestrian flows into the pedestrian flow prediction model, and output the predicted pedestrian flow at a future moment; A green light control parameter setting module, used to simulate the passing time according to the predicted pedestrian flow, obtain the pedestrian simulation passing time, and set the pedestrian simulation passing time as the green light control parameter at the future moment; The traffic light control module is used to control the traffic light at the target intersection according to the green light control parameters at the future moment.

Images