Intelligent shared bicycle scheduling management method, system and equipment based on computer vision and medium

Abstract

The invention relates to a shared bicycle intelligent scheduling management method, system and device based on computer vision and a medium. The method comprises the following steps: acquiring shared bicycle monitoring image data of a shared bicycle scheduling monitoring point; inputting the shared bicycle monitoring image data into a shared bicycle identification deep learning model to generate shared bicycle identification information; inputting the shared bicycle identification information into a shared bicycle parking feature identification deep learning model to generate shared bicycle parking feature identification information; inputting the shared bicycle identification information and the shared bicycle parking feature identification information into a shared bicycle parking feature prediction deep learning model to generate shared bicycle parking feature prediction information; and generating shared bicycle intelligent scheduling management information based on the shared bicycle identification information, the shared bicycle parking feature identification information and the shared bicycle parking feature prediction information. By adopting the method, the real-time performance, scientificity and intelligence of a shared bicycle scheduling strategy can be improved.

Description

Technical Field

[0001] This application belongs to the field of scheduling management, and in particular relates to a computer vision-based intelligent scheduling management method, system, equipment and medium for shared bicycles. Background Technology

[0002] With the rapid development of the sharing economy, shared bicycles have become a widely adopted tool for short-distance urban travel globally. However, problems such as disorderly parking and imbalanced resource allocation have gradually emerged, making it difficult for traditional management models relying on manual inspections and static rules to meet the needs of dynamic and refined management. Against this backdrop, computer vision technology, with its non-contact perception capabilities and automated processing advantages, can quickly identify specific targets and extract key features through real-time analysis of video streams or image data. This has become a key means of solving the shared bicycle management challenges, driving the transformation of shared bicycle management from manual inspection to automated perception.

[0003] However, existing computer vision-based shared bicycle management technologies have weak perception capabilities for multi-dimensional features such as bicycle parking posture and regional distribution density. They often only count the number of bicycles and cannot accurately assess the risk of congestion. In addition, existing computer vision-based shared bicycle management systems mostly use static rules for scheduling decisions, lacking the ability to predict and respond to dynamic environments, resulting in low resource allocation efficiency. Summary of the Invention

[0004] Therefore, it is necessary to provide a computer vision-based intelligent scheduling and management method, system, equipment, and medium for shared bicycles that can improve the real-time performance, effectiveness, and intelligence level of shared bicycle scheduling and management, in response to the above-mentioned technical problems.

[0005] Firstly, this application provides a computer vision-based intelligent scheduling and management method for shared bicycles, including:

[0006] Acquire shared bicycle monitoring image data from shared bicycle dispatch monitoring points;

[0007] The shared bicycle monitoring image data is input into the shared bicycle recognition deep learning model to generate shared bicycle recognition information, which includes shared bicycle recognition mask image data.

[0008] The shared bicycle identification mask image data is input into the shared bicycle parking feature recognition deep learning model to generate shared bicycle parking feature recognition information;

[0009] The shared bicycle identification information and shared bicycle parking feature identification information are input into the shared bicycle parking feature prediction deep learning model to generate shared bicycle parking feature prediction information;

[0010] Intelligent scheduling and management information for shared bicycles is generated based on shared bicycle identification information, shared bicycle parking feature identification information, and shared bicycle parking feature prediction information.

[0011] In one embodiment, the shared bicycle identification information also includes shared bicycle quantity information and shared bicycle pose information. Shared bicycle monitoring image data is input into a shared bicycle identification deep learning model to generate shared bicycle identification information, including:

[0012] The shared bicycle monitoring image data is input into the shared bicycle recognition deep learning model to generate shared bicycle recognition pre-selection box information and shared bicycle posture information;

[0013] Based on the total number of shared bicycle pre-selection boxes corresponding to the shared bicycle identification pre-selection box information, the number of shared bicycles is obtained by setting the number of shared bicycles.

[0014] Based on the relative position information of the shared bicycle preselection box corresponding to the shared bicycle identification preselection box information, combined with the preset camera intrinsic parameter matrix and the preset homography matrix, the ground positioning information of the shared bicycle is generated.

[0015] Integrate shared bicycle attitude information and shared bicycle ground positioning information to generate shared bicycle ground positioning information;

[0016] Based on the pre-selection box information for shared bicycle identification, image segmentation is performed on the shared bicycle monitoring image data to generate shared bicycle identification mask image data.

[0017] In one embodiment, the shared bicycle recognition deep learning model is a YOLO model. The shared bicycle posture information includes the overall angle information, the front angle information, and the overturned state information. The recognition model loss function of the shared bicycle recognition deep learning model includes an overall angle loss function, a front angle loss function, and an overturned state loss function. The expressions for the overall angle loss function, the front angle loss function, and the overturned state loss function are as follows:

[0018]

[0019]

[0020]

[0021] In the formula, , and These are the overall angle loss function, the front angle loss function, and the collapse state loss function, respectively. This represents the total number of all detected shared bicycle targets in the training sample image data for shared bicycle monitoring. For the first The overall angle information of the shared bicycle is output by the deep learning model for shared bicycle identification. For the first A true overall perspective label for the shared bicycle target. For the first The shared bicycle target is identified by a deep learning model, which outputs the angle information of the bicycle's front end. For the first The actual front-end angle label of the target shared bicycle. For the first The actual collapsed state label of the shared bicycle target. The output of the deep learning model for shared bicycle recognition The probability that a shared bicycle is in a fallen state.

[0022] In one embodiment, the intelligent dispatch management information for shared bicycles includes dispatch management prompts for the impact of road congestion on shared bicycles, dispatch management prompts for a decrease in the number of shared bicycles, and dispatch management prompts for an increase in the number of shared bicycles. The shared bicycle parking feature identification information includes a first real-time index of the impact of road congestion on shared bicycles.

[0023] Intelligent scheduling and management information for shared bicycles is generated based on shared bicycle identification information, shared bicycle parking feature identification information, and shared bicycle parking feature prediction information, including:

[0024] The second real-time shared bicycle road congestion impact index is calculated based on the shared bicycle pose information.

[0025] The difference between the first real-time shared bicycle road congestion impact index and the second real-time shared bicycle road congestion impact index is calculated to obtain the road congestion impact error index;

[0026] If the impact error index of silted road exceeds the preset impact error threshold of silted road, an abnormal prompt message for the impact error of silted road will be generated.

[0027] If the impact error index of silted road is greater than the preset impact error threshold of silted road, a comprehensive real-time impact index of silted road for shared bicycles is generated based on the weighted sum of the first real-time impact index of silted road for shared bicycles and the second real-time impact index of silted road for shared bicycles.

[0028] If the comprehensive real-time shared bicycle road congestion impact index exceeds the preset comprehensive real-time road congestion impact threshold, a shared bicycle road congestion impact scheduling and management prompt message will be generated.

[0029] If the number of shared bicycles exceeds the preset upper limit, a dispatch management prompt message indicating a reduction in the number of shared bicycles will be generated.

[0030] If the number of shared bikes is less than the preset minimum number of shared bikes, a dispatch management prompt message indicating an increase in the number of shared bikes will be generated.

[0031] In one embodiment, the deep learning model for shared bicycle parking feature recognition is a ResNet model, and the expression for the feature recognition loss function of the deep learning model for shared bicycle parking feature recognition is as follows:

[0032]

[0033] In the formula, For feature recognition loss function, The number of batch samples used in a single training iteration for a deep learning model to identify shared bicycle parking features. For the first The actual road congestion level of shared bicycles in a sample The output of the deep learning model for shared bicycle parking feature recognition The first real-time impact indicator of shared bicycle congestion on roads for each sample.

[0034] In one embodiment, the shared bicycle parking feature prediction information includes shared bicycle road congestion impact prediction information and shared bicycle quantity prediction information, and the shared bicycle parking feature prediction deep learning model includes an LSTM module and a fully connected module.

[0035] The shared bicycle identification information and shared bicycle parking feature identification information are input into the shared bicycle parking feature prediction deep learning model to generate shared bicycle parking feature prediction information, including:

[0036] A time-series sequence of shared bicycle parking features is constructed based on shared bicycle identification information and shared bicycle parking feature identification information;

[0037] The time series sequence of shared bicycle parking features is input into the LSTM module of the shared bicycle parking feature prediction deep learning model to generate the hidden state of the shared bicycle parking feature prediction.

[0038] The hidden state of the predicted parking features of shared bicycles is input into the fully connected module of the deep learning model for predicting the parking features of shared bicycles, thereby generating prediction information on the impact of shared bicycles on road congestion and prediction information on the number of shared bicycles.

[0039] In one embodiment, the intelligent dispatch management information for shared bicycles also includes dispatch management prompts for the expected impact of road congestion on shared bicycles, dispatch management prompts for the expected decrease in the number of shared bicycles, and dispatch management prompts for the expected increase in the number of shared bicycles.

[0040] Intelligent scheduling and management information for shared bicycles is generated based on shared bicycle identification information, shared bicycle parking feature identification information, and shared bicycle parking feature prediction information. This also includes:

[0041] If the predicted impact of shared bicycles on road congestion exceeds the preset threshold for expected road congestion impact, a scheduling and management prompt message for the expected impact of shared bicycles on road congestion will be generated.

[0042] If the predicted number of shared bicycles is greater than the preset upper limit of the expected number of shared bicycles, a scheduling management prompt message indicating a reduction in the expected number of shared bicycles will be generated.

[0043] If the predicted number of shared bicycles is less than the preset lower limit of the expected number of shared bicycles, a scheduling management prompt message indicating an increase in the expected number of shared bicycles will be generated.

[0044] Secondly, this application also provides a computer vision-based intelligent dispatch and management system for shared bicycles, including:

[0045] The monitoring image acquisition module is used to acquire shared bicycle monitoring image data from shared bicycle dispatch monitoring points;

[0046] The shared bicycle identification module is used to input shared bicycle monitoring image data into the shared bicycle identification deep learning model to generate shared bicycle identification information, which includes shared bicycle identification mask image data.

[0047] The parking feature recognition module is used to input the shared bicycle identification mask image data into the shared bicycle parking feature recognition deep learning model to generate shared bicycle parking feature recognition information;

[0048] The parking feature prediction module is used to input shared bicycle identification information and shared bicycle parking feature identification information into the shared bicycle parking feature prediction deep learning model to generate shared bicycle parking feature prediction information;

[0049] The management information generation module is used to generate intelligent scheduling and management information for shared bicycles based on shared bicycle identification information, shared bicycle parking feature identification information, and shared bicycle parking feature prediction information.

[0050] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method as described in any of the first aspects of this application.

[0051] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in any of the first aspects of this application.

[0052] The aforementioned computer vision-based intelligent scheduling and management methods, systems, equipment, and media for shared bicycles, through a deep learning model for shared bicycle recognition, can accurately identify shared bicycles from shared bicycle monitoring image data, improving the accuracy of shared bicycle recognition. By inputting shared bicycle identification mask image data into a deep learning model for shared bicycle parking feature recognition, it can accurately extract shared bicycle parking features and generate parking feature recognition information that truly reflects the actual parking status of the vehicles. Since the mask image data has already accurately defined the shared bicycle targets, the deep learning model for shared bicycle parking feature recognition can focus on the feature dimensions related to vehicle parking, effectively improving the targeting and accuracy of parking feature extraction. Through a deep learning model for shared bicycle parking feature prediction, it can accurately predict the future trend of shared bicycle parking features at monitoring points based on the current shared bicycle identification information and parking feature recognition information. This allows for early prediction of parking feature changes, enabling early warning of potential shared bicycle congestion areas and shared bicycle shortage areas, reserving sufficient scheduling preparation time for operations, and improving the real-time, scientific, and intelligent nature of shared bicycle scheduling strategies.

[0053] Intelligent scheduling and decision-making technology driven by multi-dimensional information fusion can comprehensively consider real-time shared bicycle identification information, current parking feature identification information, and future parking feature prediction information to generate scientific and accurate intelligent scheduling and management information for shared bicycles. This technology integrates real-time status data and future trend data, ensuring that the generated scheduling and management information fully matches the actual needs and development trends of monitoring points, effectively improving the rationality and effectiveness of scheduling decisions, thereby solving the technical problems of blindness and strong subjectivity inherent in traditional experience-based scheduling. By outputting accurate scheduling and management information, it can guide operators to carry out targeted scheduling work, effectively improving vehicle resource allocation efficiency, avoiding resource waste, further reducing operating and management costs, while ensuring the order of shared bicycle parking in the city, improving the user travel experience, and achieving a refined and intelligent upgrade of shared bicycle operation and management. Attached Figure Description

[0054] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1 A flowchart illustrating a computer vision-based intelligent scheduling and management method for shared bicycles, provided as an embodiment of this application. Figure 1 ;

[0056] Figure 2 A flowchart illustrating a computer vision-based intelligent scheduling and management method for shared bicycles, provided as an embodiment of this application. Figure 2 ;

[0057] Figure 3 This is a schematic diagram of the structure of a computer vision-based intelligent dispatch and management system for shared bicycles, provided as an embodiment of this application. Detailed Implementation

[0058] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0059] In one exemplary embodiment of this application, such as Figure 1 As shown, a computer vision-based intelligent scheduling and management method for shared bicycles is provided. This embodiment illustrates the application of this method to a bicycle scheduling terminal. It is understood that this method can also be applied to a bicycle scheduling server, and further to a bicycle scheduling system including both a bicycle scheduling terminal and a bicycle scheduling server, and is implemented through the interaction between the bicycle scheduling terminal and the bicycle scheduling server. In this embodiment, the method may include the following steps:

[0060] Step S101: Obtain shared bicycle monitoring image data from the shared bicycle dispatch monitoring point.

[0061] Specifically, the bike dispatch terminal can acquire shared bike monitoring image data within the monitoring area of ​​each shared bike dispatch monitoring point through visual acquisition devices deployed at each shared bike dispatch monitoring point.

[0062] Optionally, the monitoring area corresponding to the shared bicycle dispatch monitoring point may include, but is not limited to, areas where shared bicycles are frequently used and parked, such as subway station entrances, business districts, communities, and major traffic arteries.

[0063] Furthermore, the bicycle dispatch terminal can record the timestamp of the collection of each frame of shared bicycle monitoring image data.

[0064] Optionally, the bicycle dispatch terminal can perform image filtering and image enhancement on the acquired raw shared bicycle monitoring image data.

[0065] Step S102: Input the shared bicycle monitoring image data into the shared bicycle recognition deep learning model to generate shared bicycle recognition information, which includes shared bicycle recognition mask image data.

[0066] Optionally, the bike dispatch terminal can input the shared bike monitoring image data into a preset shared bike recognition deep learning model to generate shared bike recognition pre-selection box information. The bike dispatch terminal can then perform image segmentation on the shared bike monitoring image data based on the shared bike recognition pre-selection box information to obtain shared bike recognition mask image data.

[0067] Optionally, the deep learning model for shared bicycle recognition can be a YOLO model. The bicycle dispatching terminal can generate pre-selected bounding boxes and bicycle attitude information using this model. Based on the total number of pre-selected bounding boxes, the terminal can determine the number of shared bicycles. Based on the relative position information of the pre-selected bounding boxes, combined with a preset camera intrinsic matrix and homography matrix, the terminal can calculate the ground positioning information of the shared bicycles. The terminal can integrate the bicycle attitude information and ground positioning information to generate bicycle pose information. Based on the pre-selected bounding box information, the terminal can perform image segmentation on the monitored image to obtain shared bicycle identification mask image data. Finally, the terminal can integrate the shared bicycle quantity information, bicycle pose information, and shared bicycle identification mask image data to generate shared bicycle recognition information.

[0068] Step S103: Input the shared bicycle identification mask image data into the shared bicycle parking feature recognition deep learning model to generate shared bicycle parking feature recognition information.

[0069] Optionally, the bike dispatch terminal can input the shared bike identification mask image data into a preset deep learning model for shared bike parking feature recognition. The bike dispatch terminal can extract features such as the spatial distribution, parking regularity, and area occupancy of shared bikes from the shared bike identification mask image data through the deep learning model for shared bike parking feature recognition, and generate shared bike parking feature recognition information.

[0070] Optionally, the shared bicycle parking feature identification information may include a first real-time shared bicycle road congestion impact index, which can quantitatively reflect the degree of impact of shared bicycles on road traffic in the current monitoring area.

[0071] Optionally, the deep learning model for shared bicycle parking feature recognition can be a ResNet model.

[0072] Step S104: Input the shared bicycle identification information and the shared bicycle parking feature identification information into the shared bicycle parking feature prediction deep learning model to generate shared bicycle parking feature prediction information.

[0073] Optionally, the bike dispatch terminal can perform time-series processing on the shared bike identification information and shared bike parking feature identification information to construct a time-series sequence of shared bike parking features. The bike dispatch terminal can then input the time-series sequence of shared bike parking features into a preset deep learning model for shared bike parking feature prediction to generate predicted information for shared bike parking features.

[0074] Optionally, the shared bicycle parking feature prediction information may include, but is not limited to, the shared bicycle quantity prediction information and the shared bicycle road congestion impact prediction information.

[0075] Step S105: Generate intelligent scheduling and management information for shared bicycles based on shared bicycle identification information, shared bicycle parking feature identification information, and shared bicycle parking feature prediction information.

[0076] Optionally, the bike dispatch terminal can calculate a second real-time shared bike road congestion impact index based on the shared bike pose information in the shared bike identification information. The bike dispatch terminal can calculate the difference between the first and second real-time shared bike road congestion impact indices to obtain a road congestion impact error index. If the road congestion impact error index is greater than a preset road congestion impact error threshold, the bike dispatch terminal can generate an abnormal road congestion impact error message. If the road congestion impact error index is less than the preset road congestion impact error threshold, the bike dispatch terminal can generate a comprehensive real-time shared bike road congestion impact index based on the weighted sum of the first and second real-time shared bike road congestion impact indices.

[0077] Furthermore, the bike dispatch terminal can make multi-dimensional judgments by combining real-time data and predicted data: if the comprehensive real-time shared bike road congestion impact index is greater than the preset comprehensive real-time road congestion impact threshold, the bike dispatch terminal can generate a shared bike road congestion impact dispatch management prompt; if the shared bike quantity information is greater than the preset shared bike quantity upper limit, the bike dispatch terminal can generate a shared bike quantity reduction dispatch management prompt; if the shared bike quantity information is less than the preset shared bike quantity lower limit, the bike dispatch terminal can generate a shared bike quantity increase dispatch management prompt; if the shared bike road congestion impact prediction information is greater than the preset expected road congestion impact threshold, the bike dispatch terminal can generate a shared bike road congestion impact expected dispatch management prompt; if the shared bike quantity prediction information is greater than the preset expected shared bike quantity upper limit, the bike dispatch terminal can generate a shared bike quantity reduction expected dispatch management prompt; if the shared bike quantity prediction information is less than the preset expected shared bike quantity lower limit, the bike dispatch terminal can generate a shared bike quantity increase expected dispatch management prompt.

[0078] In the aforementioned computer vision-based intelligent scheduling and management method for shared bicycles, a deep learning model for shared bicycle recognition can accurately identify shared bicycles from shared bicycle monitoring image data, improving the accuracy of shared bicycle recognition. By inputting shared bicycle identification mask image data into a deep learning model for shared bicycle parking feature recognition, shared bicycle parking features can be accurately extracted, generating parking feature recognition information that truly reflects the actual parking status of the vehicles. Since the mask image data has already accurately defined the shared bicycle targets, the deep learning model for shared bicycle parking feature recognition can focus on the feature dimensions related to vehicle parking, effectively improving the targeting and accuracy of parking feature extraction. Through a deep learning model for shared bicycle parking feature prediction, based on the current shared bicycle identification information and parking feature recognition information, the future trend of shared bicycle parking features at monitoring points can be accurately predicted. This allows for early prediction of parking feature changes, enabling early warning of potential shared bicycle congestion areas and shared bicycle shortage areas, reserving sufficient scheduling preparation time for operations, and improving the real-time, scientific, and intelligent nature of shared bicycle scheduling strategies.

[0079] In an optional embodiment of this application, the shared bicycle identification information may further include shared bicycle quantity information and shared bicycle position information. Please refer to [reference needed]. Figure 1 and Figure 2 Step S102 involves inputting the shared bicycle monitoring image data into the shared bicycle recognition deep learning model to generate shared bicycle recognition information, which may include:

[0080] Step S202: Input the shared bicycle monitoring image data into the shared bicycle recognition deep learning model to generate shared bicycle recognition pre-selection box information and shared bicycle posture information.

[0081] Step S203: Based on the total number of shared bicycle pre-selection boxes corresponding to the shared bicycle identification pre-selection box information, set the shared bicycle quantity information.

[0082] Step S204: Based on the relative position information of the shared bicycle pre-selection box corresponding to the shared bicycle identification pre-selection box information, and combined with the preset camera intrinsic parameter matrix and the preset homography matrix, generate the ground positioning information of the shared bicycle.

[0083] Indicatively, the camera intrinsic parameter matrix can include core parameters such as the focal length and principal point coordinates of the visual acquisition device. The camera intrinsic parameter matrix can be used to convert image pixel coordinates to coordinates in the camera coordinate system. The homography matrix can be obtained from the initial camera calibration and ground feature point matching. The homography matrix can be used to describe the projection transformation relationship between the image plane and the ground world plane.

[0084] Optionally, the bicycle dispatch terminal can convert the pixel coordinates of the relative position of the pre-selected box into three-dimensional coordinates in the camera coordinate system through the camera intrinsic parameter matrix. The bicycle dispatch terminal can then project these three-dimensional coordinates into the ground world coordinate system through the homography matrix, completing the mapping from the two-dimensional position in the image to the three-dimensional position on the real ground, and generating ground positioning information for the shared bicycle.

[0085] Step S205: Integrate the shared bicycle attitude information and the shared bicycle ground positioning information to generate shared bicycle ground positioning information.

[0086] Step S206: Based on the shared bicycle identification pre-selection box information, perform image segmentation on the shared bicycle monitoring image data to generate shared bicycle identification mask image data.

[0087] In an optional embodiment of this application, the deep learning model for shared bicycle recognition can be a YOLO model. The shared bicycle posture information can include the overall angle information of the shared bicycle, the front angle information of the shared bicycle, and the overturned state information of the shared bicycle. The recognition model loss function of the deep learning model for shared bicycle recognition can include the overall angle loss function, the front angle loss function, and the overturned state loss function. The expressions for the overall angle loss function, the front angle loss function, and the overturned state loss function can be:

[0088]

[0089]

[0090]

[0091] In the formula, , and These are the overall angle loss function, the front angle loss function, and the collapse state loss function, respectively. This represents the total number of all detected shared bicycle targets in the training sample image data for shared bicycle monitoring. For the first The overall angle information of the shared bicycle is output by the deep learning model for shared bicycle identification. For the first A true overall perspective label for the shared bicycle target. For the first The shared bicycle target is identified by a deep learning model, which outputs the angle information of the bicycle's front end. For the first The actual front-end angle label of the target shared bicycle. For the first The actual collapsed state label of the shared bicycle target. The output of the deep learning model for shared bicycle recognition The probability that a shared bicycle is in a fallen state.

[0092] Indicative The actual collapsed state of a shared bicycle can be represented as "falling down". This can indicate that the actual collapsed state of a shared bicycle is that it is parked normally.

[0093] Indicative It can provide the overall angle information of shared bicycles Convert to coordinates on the unit circle. You can label the true overall angle. Convert to coordinates on the unit circle.

[0094] In an optional embodiment of this application, the intelligent scheduling and management information for shared bicycles may include scheduling and management prompts for the impact of road congestion on shared bicycles, scheduling and management prompts for the decrease in the number of shared bicycles, and scheduling and management prompts for the increase in the number of shared bicycles. The shared bicycle parking feature identification information may include a first real-time shared bicycle road congestion impact index.

[0095] Intelligent scheduling and management information for shared bicycles is generated based on shared bicycle identification information, shared bicycle parking feature identification information, and shared bicycle parking feature prediction information. This information may include:

[0096] Specifically, the bicycle dispatch terminal can calculate the second real-time shared bicycle road congestion impact index based on the shared bicycle's position and pose information.

[0097] Optionally, the bicycle dispatch terminal can calculate the shared bicycle density, the regularity of shared bicycle parking, and the road encroachment rate of shared bicycles based on the pose information of shared bicycles. The bicycle dispatch terminal can also calculate a second real-time road congestion impact index of shared bicycles based on the weighted sum of the shared bicycle density, the regularity of shared bicycle parking, and the road encroachment rate of shared bicycles.

[0098] Indicatively, the density of shared bicycles can be used to characterize the density of shared bicycles within the monitoring area of ​​a shared bicycle dispatch monitoring point; the regularity of shared bicycle parking can be calculated, but is not limited to, based on the variance of the overall angle information and the front angle information of shared bicycles, and the regularity of shared bicycle parking can reflect the degree of orderliness of shared bicycle parking; the road encroachment rate of shared bicycles can reflect the degree to which shared bicycles occupy key road passage space.

[0099] Specifically, the bicycle dispatch terminal can calculate the difference between the first real-time shared bicycle road congestion impact index and the second real-time shared bicycle road congestion impact index to obtain the road congestion impact error index.

[0100] Specifically, if the impact error index of silted road exceeds the preset impact error threshold of silted road, the single-vehicle dispatch terminal can generate an abnormal prompt message for the impact error of silted road.

[0101] Specifically, if the impact error index of silted-up roads is greater than the preset impact error threshold of silted-up roads, the bicycle dispatch terminal can generate a comprehensive real-time impact index of silted-up roads for shared bicycles based on the weighted sum of the first real-time impact index of silted-up roads for shared bicycles and the second real-time impact index of silted-up roads for shared bicycles.

[0102] Specifically, if the comprehensive real-time impact index of shared bicycles on congested roads exceeds the preset comprehensive real-time impact threshold of congested roads, the bicycle dispatch terminal can generate a dispatch management prompt message regarding the impact of shared bicycles on congested roads.

[0103] Specifically, if the number of shared bicycles exceeds the preset upper limit, the bicycle dispatch terminal can generate a dispatch management prompt message indicating a reduction in the number of shared bicycles.

[0104] Specifically, if the number of shared bicycles is less than the preset minimum number of shared bicycles, the bicycle dispatch terminal can generate a dispatch management prompt message indicating an increase in the number of shared bicycles.

[0105] In an optional embodiment of this application, the deep learning model for shared bicycle parking feature recognition can be a ResNet model, and the expression for the feature recognition loss function of the deep learning model for shared bicycle parking feature recognition can be:

[0106]

[0107] In the formula, For feature recognition loss function, The number of batch samples used in a single training iteration for a deep learning model to identify shared bicycle parking features. For the first The actual road congestion level of shared bicycles in a sample The output of the deep learning model for shared bicycle parking feature recognition The first real-time impact indicator of shared bicycle congestion on roads for each sample.

[0108] In an optional embodiment of this application, the shared bicycle parking feature prediction information may include shared bicycle road congestion impact prediction information and shared bicycle quantity prediction information, and the shared bicycle parking feature prediction deep learning model may include an LSTM module and a fully connected module.

[0109] Please refer to Figure 1 and Figure 2Step S104 involves inputting the shared bicycle identification information and shared bicycle parking feature identification information into the shared bicycle parking feature prediction deep learning model to generate shared bicycle parking feature prediction information, which may include:

[0110] Step S208: Construct a time series sequence of shared bicycle parking features based on shared bicycle identification information and shared bicycle parking feature identification information.

[0111] Step S209: Input the time series sequence of shared bicycle parking features into the LSTM module of the shared bicycle parking feature prediction deep learning model to generate the hidden state of the predicted parking features of shared bicycles.

[0112] Step S210: Input the hidden state of the predicted parking features of shared bicycles into the fully connected module in the deep learning model for predicting the parking features of shared bicycles to generate prediction information on the impact of road congestion caused by shared bicycles and prediction information on the number of shared bicycles.

[0113] In an optional embodiment of this application, the intelligent scheduling and management information for shared bicycles may further include scheduling and management prompts for the expected impact of road congestion on shared bicycles, scheduling and management prompts for the expected decrease in the number of shared bicycles, and scheduling and management prompts for the expected increase in the number of shared bicycles.

[0114] Intelligent scheduling and management information for shared bicycles is generated based on shared bicycle identification information, shared bicycle parking feature identification information, and shared bicycle parking feature prediction information. This information may also include:

[0115] Specifically, if the predicted impact of shared bicycles accumulating on roads is greater than the preset threshold for expected road accumulation impact, the bicycle dispatch terminal can generate a dispatch management prompt message regarding the expected impact of shared bicycles accumulating on roads.

[0116] Specifically, if the predicted number of shared bicycles is greater than the preset upper limit for the expected number of shared bicycles, the bicycle dispatch terminal can generate a dispatch management prompt message indicating a reduction in the expected number of shared bicycles.

[0117] Specifically, if the predicted number of shared bicycles is less than the preset lower limit of the expected number of shared bicycles, the bicycle dispatch terminal can generate a dispatch management prompt message indicating an increase in the expected number of shared bicycles.

[0118] In one exemplary embodiment of this application, such as Figure 2 As shown, a computer vision-based intelligent scheduling and management method for shared bicycles is provided, which may include:

[0119] Step S201: Obtain shared bicycle monitoring image data from the shared bicycle dispatch monitoring point.

[0120] Step S202: Input the shared bicycle monitoring image data into the shared bicycle recognition deep learning model to generate shared bicycle recognition pre-selection box information and shared bicycle posture information.

[0121] Step S203: Based on the total number of shared bicycle pre-selection boxes corresponding to the shared bicycle identification pre-selection box information, set the shared bicycle quantity information.

[0122] Step S204: Based on the relative position information of the shared bicycle pre-selection box corresponding to the shared bicycle identification pre-selection box information, and combined with the preset camera intrinsic parameter matrix and the preset homography matrix, generate the ground positioning information of the shared bicycle.

[0123] Step S205: Integrate the shared bicycle attitude information and the shared bicycle ground positioning information to generate shared bicycle ground positioning information.

[0124] Step S206: Based on the shared bicycle identification pre-selection box information, perform image segmentation on the shared bicycle monitoring image data to generate shared bicycle identification mask image data.

[0125] Step S207: Input the shared bicycle identification mask image data into the shared bicycle parking feature recognition deep learning model to generate shared bicycle parking feature recognition information.

[0126] Step S208: Construct a time series sequence of shared bicycle parking features based on shared bicycle identification information and shared bicycle parking feature identification information.

[0127] Step S209: Input the time series sequence of shared bicycle parking features into the LSTM module of the shared bicycle parking feature prediction deep learning model to generate the hidden state of the predicted parking features of shared bicycles.

[0128] Step S210: Input the hidden state of the predicted parking features of shared bicycles into the fully connected module in the deep learning model for predicting the parking features of shared bicycles to generate prediction information on the impact of road congestion caused by shared bicycles and prediction information on the number of shared bicycles.

[0129] Step S211: Generate intelligent scheduling and management information for shared bicycles based on shared bicycle identification information, shared bicycle parking feature identification information, and shared bicycle parking feature prediction information.

[0130] The aforementioned computer vision-based intelligent scheduling and management method for shared bicycles accurately acquires shared bicycle monitoring image data and generates recognition results containing pre-selected boxes, pose information, etc., through computer vision technology and deep learning models. This allows for the derivation of recognition information such as the number of shared bicycles, ground positioning, and mask images. Furthermore, parking features of the mask images are extracted based on deep learning models, and time-series data and prediction models are combined to predict the impact of congestion and the changing trends in the number of shared bicycles. This enables precise real-time perception of the parking status of shared bicycles and accurate prediction of future trends, significantly improving the targeting and foresight of shared bicycle scheduling and management, optimizing the efficiency of shared bicycle resource allocation, reducing operating and management costs, ensuring urban road traffic order, and enhancing the user travel experience.

[0131] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0132] Based on the same inventive concept, this application also provides a computer vision-based intelligent dispatch and management system for implementing the aforementioned computer vision-based intelligent dispatch and management method for shared bicycles. The solution provided by this system is similar to the implementation scheme described in the above method. Therefore, the specific limitations of one or more computer vision-based intelligent dispatch and management system embodiments provided below can be found in the limitations of the computer vision-based intelligent dispatch and management method for shared bicycles described above, and will not be repeated here.

[0133] In one exemplary embodiment, such as Figure 3 As shown, a computer vision-based intelligent dispatch and management system 300 for shared bicycles is provided, including:

[0134] The monitoring image acquisition module 301 can be used to acquire shared bicycle monitoring image data from shared bicycle dispatch monitoring points.

[0135] The shared bicycle identification module 302 can be used to input shared bicycle monitoring image data into a shared bicycle identification deep learning model to generate shared bicycle identification information, which includes shared bicycle identification mask image data.

[0136] The parking feature recognition module 303 can be used to input shared bicycle identification mask image data into the shared bicycle parking feature recognition deep learning model to generate shared bicycle parking feature recognition information.

[0137] The parking feature prediction module 304 can be used to input shared bicycle identification information and shared bicycle parking feature identification information into the shared bicycle parking feature prediction deep learning model to generate shared bicycle parking feature prediction information.

[0138] The management information generation module 305 can be used to generate intelligent scheduling and management information for shared bicycles based on shared bicycle identification information, shared bicycle parking feature identification information, and shared bicycle parking feature prediction information.

[0139] In an optional embodiment of this application, the shared bicycle identification module 302 can also be used for:

[0140] The shared bicycle monitoring image data is input into the shared bicycle recognition deep learning model to generate shared bicycle recognition pre-selection box information and shared bicycle posture information.

[0141] Based on the total number of shared bicycle pre-selection boxes corresponding to the shared bicycle identification pre-selection box information, the number of shared bicycles is obtained.

[0142] Based on the relative position information of the shared bicycle pre-selection box corresponding to the shared bicycle identification pre-selection box information, combined with the preset camera intrinsic parameter matrix and the preset homography matrix, the ground positioning information of the shared bicycle is generated.

[0143] Integrate shared bicycle attitude information and shared bicycle ground positioning information to generate shared bicycle ground positioning information.

[0144] Based on the pre-selection box information for shared bicycle identification, image segmentation is performed on the shared bicycle monitoring image data to generate shared bicycle identification mask image data.

[0145] In an optional embodiment of this application, the management information generation module 305 may also be used for:

[0146] The second real-time road congestion impact index of shared bicycles is calculated based on the pose information of shared bicycles.

[0147] The difference between the first real-time shared bicycle road congestion impact index and the second real-time shared bicycle road congestion impact index is calculated to obtain the road congestion impact error index.

[0148] If the impact error index of silted road exceeds the preset impact error threshold of silted road, an abnormal prompt message for the impact error of silted road will be generated.

[0149] If the impact error index of silted-up roads is greater than the preset impact error threshold of silted-up roads, a comprehensive real-time impact index of silted-up roads for shared bicycles is generated based on the weighted sum of the first real-time impact index of silted-up roads for shared bicycles and the second real-time impact index of silted-up roads for shared bicycles.

[0150] If the comprehensive real-time impact index of shared bicycles on congested roads exceeds the preset comprehensive real-time impact threshold of congested roads, a dispatch management prompt message for the impact of shared bicycles on congested roads will be generated.

[0151] If the number of shared bicycles exceeds the preset upper limit, a dispatch management prompt message indicating a reduction in the number of shared bicycles will be generated.

[0152] If the number of shared bikes is less than the preset minimum number of shared bikes, a dispatch management prompt message indicating an increase in the number of shared bikes will be generated.

[0153] In an optional embodiment of this application, the parking feature prediction module 304 may also be used for:

[0154] A time series sequence of shared bicycle parking features is constructed based on shared bicycle identification information and shared bicycle parking feature identification information.

[0155] The time series sequence of shared bicycle parking features is input into the LSTM module of the shared bicycle parking feature prediction deep learning model to generate the hidden state of the shared bicycle parking feature prediction.

[0156] The hidden state of the predicted parking features of shared bicycles is input into the fully connected module of the deep learning model for predicting the parking features of shared bicycles, thereby generating prediction information on the impact of shared bicycles on road congestion and prediction information on the number of shared bicycles.

[0157] In an optional embodiment of this application, the management information generation module 305 may also be used for:

[0158] If the predicted impact of shared bicycles accumulating on roads exceeds the preset threshold for expected road accumulation impact, a scheduling and management prompt message for the expected impact of shared bicycles accumulating on roads will be generated.

[0159] If the predicted number of shared bicycles exceeds the preset upper limit for the expected number of shared bicycles, a scheduling management prompt message indicating a reduction in the expected number of shared bicycles will be generated.

[0160] If the predicted number of shared bicycles is less than the preset lower limit of the expected number of shared bicycles, a scheduling management prompt message indicating an increase in the expected number of shared bicycles will be generated.

[0161] In one embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the computer vision-based intelligent scheduling and management method for shared bicycles as described above.

[0162] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0163] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0164] The above-described embodiments are merely illustrative of several implementation methods of the embodiments of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the embodiments of this application, and these modifications and improvements all fall within the protection scope of the embodiments of this application.

Claims

1. A computer vision-based intelligent scheduling and management method for shared bicycles, characterized in that, The method includes: Acquire shared bicycle monitoring image data from shared bicycle dispatch monitoring points; The shared bicycle monitoring image data is input into the shared bicycle recognition deep learning model to generate shared bicycle recognition information, which includes shared bicycle recognition mask image data. The shared bicycle identification mask image data is input into the shared bicycle parking feature recognition deep learning model to generate shared bicycle parking feature recognition information; The shared bicycle identification information and the shared bicycle parking feature identification information are input into the shared bicycle parking feature prediction deep learning model to generate shared bicycle parking feature prediction information; Intelligent scheduling and management information for shared bicycles is generated based on the shared bicycle identification information, the shared bicycle parking feature identification information, and the shared bicycle parking feature prediction information.

2. The method according to claim 1, characterized in that, The shared bicycle identification information also includes shared bicycle quantity information and shared bicycle pose information. The step of inputting the shared bicycle monitoring image data into a shared bicycle identification deep learning model to generate shared bicycle identification information includes: The shared bicycle monitoring image data is input into the shared bicycle recognition deep learning model to generate shared bicycle recognition pre-selection box information and shared bicycle posture information; Based on the total number of shared bicycle preselection boxes corresponding to the shared bicycle identification preselection box information, the shared bicycle quantity information is set to obtain the shared bicycle quantity information; Based on the relative position information of the shared bicycle preselection box corresponding to the shared bicycle identification preselection box information, combined with the preset camera intrinsic parameter matrix and the preset homography matrix, the ground positioning information of the shared bicycle is generated. Integrate the shared bicycle's attitude information and the shared bicycle's ground positioning information to generate the shared bicycle's ground positioning information; Based on the shared bicycle identification preselection box information, the shared bicycle monitoring image data is segmented to generate the shared bicycle identification mask image data.

3. The method according to claim 2, characterized in that, The shared bicycle recognition deep learning model is a YOLO model. The shared bicycle posture information includes the overall angle information, the front angle information, and the overturned state information. The recognition model loss function of the shared bicycle recognition deep learning model includes an overall angle loss function, a front angle loss function, and an overturned state loss function. The expressions for the overall angle loss function, the front angle loss function, and the overturned state loss function are as follows: In the formula, , and These are the overall angle loss function, the vehicle front angle loss function, and the overturned state loss function, respectively. This represents the total number of all detected shared bicycle targets in the training sample image data for shared bicycle monitoring. For the first The overall angle information of the shared bicycle is output by the deep learning model for shared bicycle identification of the target shared bicycle. For the first A true overall perspective label for the shared bicycle target. For the first The shared bicycle target is the shared bicycle identification deep learning model outputting the shared bicycle's front angle information. For the first The actual front-end angle label of the target shared bicycle. For the first The actual collapsed state label of the shared bicycle target. The output of the deep learning model for shared bicycle recognition is the first... The probability that the shared bicycle in the target shared bicycle is in a fallen state.

4. The method according to claim 2, characterized in that, The shared bicycle intelligent dispatch management information includes dispatch management prompts for shared bicycles affected by road congestion, dispatch management prompts for a decrease in the number of shared bicycles, and dispatch management prompts for an increase in the number of shared bicycles. The shared bicycle parking feature identification information includes a first real-time shared bicycle road congestion impact index. The process of generating intelligent scheduling and management information for shared bicycles based on the shared bicycle identification information, the shared bicycle parking feature identification information, and the shared bicycle parking feature prediction information includes: The second real-time shared bicycle road congestion impact index is calculated based on the shared bicycle pose information. The difference between the first real-time shared bicycle road congestion impact index and the second real-time shared bicycle road congestion impact index is calculated to obtain the road congestion impact error index; If the silted road impact error index is greater than the preset silted road impact error threshold, an abnormal silted road impact error prompt message will be generated. If the impact error index of the silted road is greater than the preset impact error threshold of the silted road, a comprehensive real-time shared bicycle silted road impact index is generated based on the weighted sum of the first real-time shared bicycle silted road impact index and the second real-time shared bicycle silted road impact index. If the comprehensive real-time shared bicycle road congestion impact index is greater than the preset comprehensive real-time road congestion impact threshold, a shared bicycle road congestion impact scheduling management prompt message will be generated; If the number of shared bicycles exceeds the preset upper limit for the number of shared bicycles, a scheduling management prompt message indicating a reduction in the number of shared bicycles will be generated. If the number of shared bicycles is less than the preset minimum number of shared bicycles, a scheduling management prompt message indicating an increase in the number of shared bicycles will be generated.

5. The method according to claim 4, characterized in that, The deep learning model for shared bicycle parking feature recognition is a ResNet model, and the expression for the feature recognition loss function of the deep learning model for shared bicycle parking feature recognition is as follows: In the formula, The feature recognition loss function is... This refers to the batch size of the deep learning model for shared bicycle parking feature recognition during a single training iteration. For the first The actual road congestion level of shared bicycles in a sample The output of the deep learning model for shared bicycle parking feature recognition is the first... The first real-time shared bicycle road congestion impact index for each sample.

6. The method according to claim 1, characterized in that, The shared bicycle parking feature prediction information includes shared bicycle road congestion impact prediction information and shared bicycle quantity prediction information. The shared bicycle parking feature prediction deep learning model includes an LSTM module and a fully connected module. The step of inputting the shared bicycle identification information and the shared bicycle parking feature identification information into the shared bicycle parking feature prediction deep learning model to generate shared bicycle parking feature prediction information includes: A time-series sequence of shared bicycle parking features is constructed based on the shared bicycle identification information and the shared bicycle parking feature identification information; The shared bicycle parking feature time series is input into the LSTM module in the shared bicycle parking feature prediction deep learning model to generate the hidden state of the shared bicycle predicted parking feature; The hidden state of the predicted parking features of shared bicycles is input into the fully connected module in the deep learning model for predicting the parking features of shared bicycles to generate the prediction information of the impact of road congestion caused by shared bicycles and the prediction information of the number of shared bicycles.

7. The method according to claim 6, characterized in that, The intelligent dispatch and management information for shared bicycles also includes dispatch and management prompts for the expected impact of road congestion on shared bicycles, dispatch and management prompts for the expected decrease in the number of shared bicycles, and dispatch and management prompts for the expected increase in the number of shared bicycles. The process of generating intelligent scheduling and management information for shared bicycles based on the shared bicycle identification information, the shared bicycle parking feature identification information, and the shared bicycle parking feature prediction information further includes: If the predicted impact of the shared bicycle road congestion is greater than the preset expected impact threshold, a scheduling and management prompt message for the expected impact of the shared bicycle road congestion will be generated. If the predicted number of shared bicycles is greater than the preset upper limit of the expected number of shared bicycles, a scheduling management prompt message indicating a reduction in the expected number of shared bicycles will be generated. If the predicted number of shared bicycles is less than the preset lower limit of the expected number of shared bicycles, a scheduling management prompt message indicating an increase in the expected number of shared bicycles will be generated.

8. A computer vision-based intelligent dispatch and management system for shared bicycles, characterized in that, The system includes: The monitoring image acquisition module is used to acquire shared bicycle monitoring image data from shared bicycle dispatch monitoring points; The shared bicycle identification module is used to input the shared bicycle monitoring image data into the shared bicycle identification deep learning model to generate shared bicycle identification information, which includes shared bicycle identification mask image data. The parking feature recognition module is used to input the shared bicycle identification mask image data into the shared bicycle parking feature recognition deep learning model to generate shared bicycle parking feature recognition information; The parking feature prediction module is used to input the shared bicycle identification information and the shared bicycle parking feature identification information into the shared bicycle parking feature prediction deep learning model to generate shared bicycle parking feature prediction information; The management information generation module is used to generate intelligent scheduling and management information for shared bicycles based on the shared bicycle identification information, the shared bicycle parking feature identification information, and the shared bicycle parking feature prediction information.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 7.

Images