Park platform scheduling system based on multi-source data fusion and intelligent video algorithm

Abstract

The invention provides a park platform scheduling system based on multi-source data fusion and an intelligent video algorithm, and the system comprises the steps: obtaining multi-source data which comprises video monitoring data, sensor data and business system data; processing the video monitoring data through an optical flow method and a dynamic splicing algorithm to realize seamless integration and real-time monitoring of multiple paths of monitoring videos; performing fusion analysis on the multi-source data by adopting a machine learning algorithm to realize intelligent task allocation and resource optimization scheduling; according to a scheduling result, path planning, automatic obstacle avoidance and real-time adjustment are performed by using an optimal path algorithm, so that efficient execution of platform operation is ensured; data display is carried out through a data visualization technology and a BI tool, and visual data support is provided for scheduling decision making.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and particularly to a park platform scheduling system based on multi-source data fusion and intelligent video algorithms. Background Art

[0002] The core technical problem faced by the intelligent park vehicle and cargo scheduling system is how to achieve accurate and efficient vehicle and cargo scheduling in a complex and changeable park environment. This problem involves multiple aspects: First, there are a large number of heterogeneous data sources in the park, and how to effectively integrate and process these data in real time is a major challenge. Second, the park environment changes dynamically, and the states of vehicles and goods are constantly updated. How to timely capture these changes and make corresponding adjustments is crucial. Third, scheduling decisions need to consider many factors, including vehicle types, cargo characteristics, road conditions, time windows, etc. How to find the optimal solution under numerous constraints is a complex optimization problem. In addition, the handling of abnormal situations is also a difficult point, and the system needs to have the ability of rapid response and flexible adjustment. Finally, how to maximize resource utilization rate, reduce empty running and waiting time while ensuring scheduling efficiency is a long-term challenge faced by the entire system. These problems are interrelated and jointly constitute a complex system-level problem, which requires the comprehensive application of a variety of advanced technologies to be effectively solved. Summary of the Invention

[0003] The present invention provides a park platform scheduling system based on multi-source data fusion and intelligent video algorithms, which mainly includes: Obtain multi-source data, where the multi-source data includes video surveillance data, sensor data, and business system data; process the video surveillance data through the optical flow method and dynamic stitching algorithm to achieve seamless integration and real-time monitoring of multiple monitoring videos; use machine learning algorithms to perform fusion analysis on the multi-source data to achieve intelligent task allocation and resource optimization scheduling; according to the scheduling results, use the optimal path algorithm for path planning, automatic obstacle avoidance, and real-time adjustment to ensure the efficient execution of platform operations; perform data display through data visualization technology and BI tools to provide intuitive data support for scheduling decisions.

[0004] Further, the obtaining of multi-source data includes: collecting video surveillance data, sensor data, and business system data through high-throughput data acquisition technology and multi-source data access technology, and performing real-time data transmission and storage.

[0005] Further, the processing of video surveillance data through the optical flow method and dynamic stitching algorithm includes: using the optical flow method to analyze the motion information between video frames, and combining the dynamic stitching algorithm to achieve seamless integration of multiple videos, eliminate the viewing blind area, and improve the integrity and coherence of the monitoring screen.

[0006] Further, the fusion analysis of multi-source data using machine learning algorithms includes: using deep learning algorithms such as CNN, RNN, etc. to analyze video data to achieve license plate recognition, face recognition, behavior analysis, and anomaly detection; using machine learning algorithms to perform correlation analysis on various types of data to optimize task allocation and resource scheduling strategies.

[0007] Further, the optimization of task allocation and resource scheduling strategies using machine learning algorithms includes: analyzing data such as vehicle arrival time, cargo loading and unloading requirements, and platform idle periods through algorithms to automatically match platform resources, generate the optimal scheduling plan, reduce vehicle waiting time, and improve platform utilization rate.

[0008] Further, the path planning, automatic obstacle avoidance, and real-time adjustment using the optimal path algorithm include: comprehensively considering factors such as platform layout, obstacle distribution, and cargo storage location to plan the optimal driving path of the vehicle in the park; combining real-time environmental data to dynamically adjust the path to achieve the function of automatic obstacle avoidance and ensure the safety and efficiency of vehicle passage.

[0009] Further, the data display through data visualization technology and BI tools includes: using data visualization tools such as Tableau, Power BI, etc. to generate visual reports such as the execution progress of scheduling tasks, platform utilization rate, and key performance indicators; presenting the operating status of the platform in real time through a visual interface to provide intuitive data support for management decisions.

[0010] Further, the system further includes a data management module for cleaning, storing, and managing the collected multi-source heterogeneous data. The data management module includes a distributed database, a NoSQL database, and a data lake architecture, supporting the efficient storage and rapid retrieval of massive data.

[0011] The technical solution provided by the embodiment of the present invention may include the following beneficial effects: The present invention discloses an intelligent park vehicle and cargo scheduling system. The system obtains multi-source heterogeneous data in the park, uses the optical flow method and dynamic stitching algorithm to achieve panoramic monitoring, and uses deep learning algorithms to extract key business information. Based on real-time and historical data, the present invention uses machine learning algorithms for intelligent scheduling to allocate docking platforms and loading and unloading periods for vehicles. Considering various factors, the present invention uses the optimal path planning algorithm to plan routes for vehicles. In case of an abnormal situation, an emergency plan is activated to dynamically adjust the plan. Scheduling instructions are issued through a visual interface and a voice system, and the whole process is monitored and recorded. The present invention also uses big data analysis to continuously optimize the scheduling model to improve efficiency and accuracy. In summary, the present invention realizes the intelligent and precise scheduling of park vehicles and cargo, effectively improves the logistics operation efficiency, and reduces the labor cost. Description of the Drawings

[0012] Figure 1 This is a flowchart of a campus platform scheduling system based on multi-source data fusion and intelligent video algorithms of the present invention.

[0013] Figure 2 This is a schematic diagram of a campus platform scheduling system based on multi-source data fusion and intelligent video algorithms of the present invention.

[0014] Figure 3 This is another schematic diagram of a campus platform scheduling system based on multi-source data fusion and intelligent video algorithms of the present invention.

[0015] Figure 4 This is another schematic diagram of a campus platform scheduling system based on multi-source data fusion and intelligent video algorithms of the present invention.

[0016] Figure 5 This is another schematic diagram of a campus platform scheduling system based on multi-source data fusion and intelligent video algorithms of the present invention.

[0017] Figure 6 This is another schematic diagram of a campus platform scheduling system based on multi-source data fusion and intelligent video algorithms of the present invention.

[0018] Figure 7 This is another schematic diagram of a campus platform scheduling system based on multi-source data fusion and intelligent video algorithms of the present invention.

[0019] Figure 8 This is another schematic diagram of a campus platform scheduling system based on multi-source data fusion and intelligent video algorithms of the present invention.

[0020] Figure 9 This is another schematic diagram of a campus platform scheduling system based on multi-source data fusion and intelligent video algorithms of the present invention. Detailed implementation manners

[0021] To further understand the content of the present invention, the present invention will be described in detail in combination with the accompanying drawings and embodiments. The following further describes the present application in detail with reference to the accompanying drawings and embodiments. It can be understood that the specific embodiments described herein are only used to explain the relevant invention, rather than limiting the invention. In addition, it should be noted that for the sake of description, only the parts related to the invention are shown in the drawings.

[0022] As Figures 1-9 , a campus platform scheduling system based on multi-source data fusion and intelligent video algorithms in this embodiment may specifically include: S101. Obtain multi-source heterogeneous data in the campus, where the multi-source heterogeneous data includes video surveillance data, vehicle data, cargo data, and business data.

[0023] Deploy video surveillance devices to collect real-time video data within the park and obtain video surveillance data. Install sensor devices to collect information on vehicles entering and leaving the park and obtain vehicle data. Integrate business systems to extract business information such as goods storage, loading and unloading, and transportation, and obtain goods data. Connect to the business management platform to extract business information such as task allocation and resource scheduling, and obtain business data. Use a distributed database to store video surveillance data, vehicle data, goods data, and business data. Utilize data cleaning tools to clean multi-source heterogeneous data and remove noise and redundant information. Through data fusion algorithms, fuse the cleaned video surveillance data, vehicle data, goods data, and business data. Adopt a data lake architecture to store the fused multi-source heterogeneous data and form a unified data resource pool. Use big data analysis tools to deeply mine the multi-source heterogeneous data in the data resource pool and extract valuable information.

[0024] Specifically, deploy video surveillance devices to collect real-time video data within the park, obtain video surveillance data, and use high-definition cameras to capture dynamic images of the platform and its surrounding areas in real time at a resolution of 1080p. Install sensor devices to collect information on vehicles entering and leaving the park, obtain vehicle data, and use RFID sensors to read vehicle tag information at a frequency of 10 times per second, recording the time and location of vehicle entry and exit. Integrate the business system to extract business information such as goods storage, loading and unloading, and transportation, obtain goods data, and extract attributes such as the weight, volume, and loading and unloading status of goods in JSON format through the API interface. Connect to the business management platform to extract business information such as task assignment and resource scheduling, obtain business data, and use SQL query statements to extract task priorities, resource allocation plans, and scheduling status from the MySQL database. Use a distributed database to store video surveillance data, vehicle data, goods data, and business data, store video data in the Hadoop cluster in HDFS format, and store other data in the PostgreSQL database in a sharded manner. Use data cleaning tools to clean multi-source heterogeneous data, remove noise and redundant information, use Python scripts combined with regular expressions to filter out invalid characters, and use the mean imputation method to fill in missing values. Through data fusion algorithms, fuse the cleaned video surveillance data, vehicle data, goods data, and business data, use the Kalman filter algorithm to align time series data, and use a rule-based inference engine to achieve correlation analysis of multi-source data. Adopt a data lake architecture to store the fused multi-source heterogeneous data, form a unified data resource pool, store the fused data in Parquet format in AWS S3, and establish a metadata index to support fast query. Use big data analysis tools to deeply mine the multi-source heterogeneous data in the data resource pool, extract valuable information, use the Spark MLlib library to implement clustering analysis to identify peak vehicle entry and exit times, and use the decision tree algorithm to predict goods loading and unloading times.

[0025] S103. Use the optical flow method and dynamic stitching algorithm to perform real-time stitching and fusion processing on multiple surveillance videos, eliminate the viewing blind area, and obtain a panoramic surveillance image.

[0026] Obtain multiple surveillance video streams and parse the video frame sequences. Use the optical flow method to calculate the pixel motion vectors between adjacent frames. Estimate the overlapping areas between video frames based on the motion vectors. Apply the dynamic stitching algorithm to perform feature matching and image alignment on the overlapping areas. Fuse the overlapping areas of multiple videos to generate a preliminary panoramic surveillance image. Detect the viewing blind areas in the panoramic image and identify the uncovered areas. Adjust the stitching parameters of the video frames to optimize the coverage of the viewing blind areas. Perform color correction and brightness equalization on the optimized panoramic surveillance image. Output a seamlessly integrated panoramic surveillance image for real-time monitoring and elimination of viewing blind areas.

[0027] Specifically, obtain multiple monitoring video streams, parse the video frame sequence, adopt the H.264 encoding format, with a frame rate of 30fps and a resolution set to 1920x1080 to ensure the integrity and clarity of video data. Calculate the pixel motion vectors between adjacent frames using the optical flow method, use the Lucas-Kanade algorithm, set the window size to 15x15 pixels, calculate the displacement of each pixel point between two frames, and obtain the motion vector field. Estimate the overlapping area between video frames based on the motion vectors, determine the boundary of the overlapping area by calculating the average value of the motion vectors, set the overlapping width to 300 pixels to ensure smooth transition during stitching. Apply the dynamic stitching algorithm to perform feature matching and image alignment on the overlapping area, adopt the SIFT feature extraction algorithm to extract key points and calculate descriptors, use the RANSAC algorithm for feature matching, calculate the affine transformation matrix, and align the images. Fuse the overlapping areas of multiple videos to generate a preliminary panoramic monitoring image, adopt the multi-band fusion algorithm to decompose the image into low-frequency and high-frequency components, perform weighted fusion respectively, with the weights set to 0.6 and 0.4, to eliminate stitching traces. Detect the perspective blind spots in the panoramic image, identify the uncovered areas, use the Canny edge detection algorithm to extract the edge information of the image, set the threshold range to 50-150, and identify the boundaries of the blind spots. Adjust the stitching parameters of the video frames to optimize the coverage of the perspective blind spots, iteratively optimize the affine transformation matrix, adjust the rotation angle and translation distance, set the maximum number of iterations to 100, and optimize the blind spot coverage. Perform color correction and brightness equalization on the optimized panoramic monitoring image, adopt the histogram equalization algorithm to adjust the brightness and contrast of the image, and use the color transfer technique to unify the color style. Output the seamlessly integrated panoramic monitoring image for real-time monitoring and perspective blind spot elimination, adopt the H.265 encoding format, with a compression ratio of 2:1, to reduce storage and transmission costs while ensuring image quality.

[0028] S105. Analyze the panoramic monitoring image using deep learning algorithms to extract vehicle license plate information, personnel identity information, cargo information, etc., and obtain the real-time business data of the park.

[0029] Obtain the real-time video stream data of the panoramic surveillance video. Adopt the multi-source data access technology and input the video data into the data processing platform. Preprocess the video stream data, perform frame extraction and image enhancement to improve the picture clarity and provide high-quality input for subsequent analysis. Use the convolutional neural network (CNN) in the deep learning algorithm to detect the vehicles in the video frames and locate the vehicle positions. In the detected vehicle position area, use the license plate recognition algorithm to extract the license plate information and identify the license plate number. Adopt the face recognition algorithm to detect and extract the features of the people in the video frames and identify the personal identity information. Combine the three-dimensional geographic information and the real-time environmental data to detect and classify the goods, and extract the goods information, including the type and quantity of the goods. Associate the extracted license plate information, personal identity information and goods information to generate the real-time business data of the park. Clean and store the real-time business data, and use big data analysis tools to deeply mine the data to form structured data. Use BI tools to visually process the structured data to generate real-time business data reports and dashboards to display the operation status of the park.

[0030] Specifically, obtain the real-time video stream data of the panoramic surveillance video. Adopt the multi-source data access technology and input the video data into the data processing platform with a frame rate of 30 frames per second and a resolution of 1920×1080. Preprocess the video stream data, use Gaussian filtering and histogram equalization for image enhancement to improve the picture clarity to a signal-to-noise ratio greater than 30dB. Use the convolutional neural network (CNN) in the deep learning algorithm, such as the YOLOv5 model, to detect the vehicles in the video frames and locate the vehicle positions, with a detection accuracy of over 95%. In the detected vehicle position area, use the license plate recognition algorithm based on OpenCV to extract the license plate information and identify the license plate number, with an identification accuracy of over 98%. Adopt the face recognition algorithm, such as the FaceNet model, to detect and extract the features of the people in the video frames and identify the personal identity information, with an identification accuracy of 97%. Combine the three-dimensional geographic information and the real-time environmental data, and use the ResNet50 model to detect and classify the goods, extract the goods information, including the type and quantity of the goods, with a classification accuracy of 96%. Associate the extracted license plate information, personal identity information and goods information to generate the real-time business data of the park, with a data update frequency of once per second. Clean and store the real-time business data, use Spark for in-depth data mining to form structured data, and store it in HDFS. Use Tableau to visually process the structured data to generate real-time business data reports and dashboards to display the operation status of the park, including key indicators such as vehicle flow, personnel distribution and goods turnover.

[0031] S107. According to the real-time business data and historical data, use machine learning algorithms for intelligent scheduling of vehicles and goods, and allocate docking platforms and loading and unloading time periods for vehicles according to predetermined rules.

[0032] Obtain the real-time business data of vehicle arrival time, cargo loading and unloading requirements, and platform idle time periods in the platform management system. Extract historical scheduling data from the data management platform, including vehicle docking platform records, loading and unloading time period allocation, and resource utilization information. Use multi-source data fusion technology to integrate the real-time business data with the historical data to generate a scheduling analysis data set. Based on machine learning algorithms, train the scheduling analysis data set to construct an intelligent scheduling model for vehicles and goods. According to the intelligent scheduling model, generate a preliminary allocation plan for vehicle docking platforms and loading and unloading time periods. Combine three-dimensional geographic information and real-time environmental data to optimize the path planning and operation sequence for the preliminary allocation plan. Use an efficient scheduling algorithm to dynamically adjust the optimized allocation plan to determine whether there is a situation where vehicles wait too long or platforms are idle. If there is a situation where vehicles wait too long or platforms are idle, readjust the allocation plan to ensure maximum resource utilization. Push the finally determined allocation plan for vehicle docking platforms and loading and unloading time periods to the platform management system to execute the scheduling task.

[0033] Specifically, collect real-time business data such as vehicle arrival time, cargo loading and unloading requirements, and platform idle periods from the platform management system. For example, the vehicle arrival time is 9 am, the cargo loading and unloading requirement is 2 hours, and the platform idle period is from 10 am to 12 pm. Extract historical scheduling data from the data management platform, including vehicle docking platform records, loading and unloading period allocation, and resource utilization information. For example, the platform utilization rate in the past week was 85%. Use multi-source data fusion technology to integrate real-time business data with historical data to generate a scheduling analysis dataset. For example, match the vehicle arrival time with the historical loading and unloading efficiency. Based on machine learning algorithms, train the scheduling analysis dataset to build an intelligent scheduling model for vehicles and cargo. For example, use the random forest algorithm to predict the optimal loading and unloading period. According to the intelligent scheduling model, generate a preliminary allocation plan for vehicle docking platforms and loading and unloading periods. For example, allocate vehicle A to platform 1 with the loading and unloading period from 10 am to 12 pm. Combine three-dimensional geographic information and real-time environmental data to optimize the path planning and operation sequence for the preliminary allocation plan. For example, calculate the shortest path based on the distance between platforms. Use an efficient scheduling algorithm to dynamically adjust the optimized allocation plan to determine whether there is a situation where a vehicle waits too long or a platform is idle. For example, it is detected that vehicle B has waited for more than 30 minutes. If there is a situation where a vehicle waits too long or a platform is idle, readjust the allocation plan to ensure maximum resource utilization. For example, adjust vehicle B to platform 2 with the loading and unloading period from 11 am to 1 pm. Push the finally determined allocation plan for vehicle docking platforms and loading and unloading periods to the platform management system to execute the scheduling task. For example, update the platform status in real time through the system interface.

[0034] S109. Adopt the optimal path planning algorithm, and combine factors such as vehicle location, cargo storage location, and road conditions to plan the optimal path from the platform to the loading and unloading area for the vehicle.

[0035] Obtain vehicle location data, and use GPS positioning technology to collect the real-time coordinate information of the vehicle. Obtain the cargo storage location data, and extract the specific storage location of the cargo in the warehouse through the warehouse management system. Obtain road condition data, and combine the traffic monitoring system and historical data to analyze the current road congestion level and traffic conditions. Integrate the vehicle location, cargo storage location, and road condition data to construct a multi-dimensional path planning basic data set. Use a three-dimensional geographic information model to map the basic data set onto the three-dimensional map of the park to generate a visual path planning scenario. Based on machine learning algorithms, train a path planning model, and optimize the model by combining historical optimal path data and real-time environmental factors. Run the path planning model, and generate multiple candidate paths according to the vehicle location, cargo storage location, and real-time road conditions. Use an optimal path evaluation algorithm to comprehensively score the candidate paths based on path length, travel time, and energy consumption indicators. Determine the optimal path, and select the path with the highest score as the final driving route of the vehicle from the platform to the loading and unloading area.

[0036] Specifically, the generation steps are as follows: Use GPS positioning technology to collect the real-time coordinate information of the vehicle to obtain vehicle location data. For example, obtain the current longitude and latitude of the vehicle as (39.9042, 116.4074) through the GPS module. Extract the specific storage location of the cargo in the warehouse through the warehouse management system to obtain the cargo storage location data. For example, cargo A is located on the 3rd layer of shelf C in area B of the warehouse. Combine the traffic monitoring system and historical data to analyze the current road congestion level and traffic conditions to obtain road condition data. For example, identify the current road congestion index as 0.8 through traffic cameras. Integrate the vehicle location, cargo storage location, and road condition data to construct a multi-dimensional path planning basic data set. For example, integrate the vehicle location, cargo location, and road congestion index into a structured data table. Use a three-dimensional geographic information model to map the basic data set onto the three-dimensional map of the park to generate a visual path planning scenario. For example, use GIS software to construct a three-dimensional map including roads, warehouses, and platforms. Based on machine learning algorithms, train a path planning model, and optimize the model by combining historical optimal path data and real-time environmental factors. For example, use the random forest algorithm to train the model, input historical optimal path data and real-time environmental data, and output the path planning result. Run the path planning model, and generate multiple candidate paths according to the vehicle location, cargo storage location, and real-time road conditions. For example, generate path A with a length of 500 meters and path B with a length of 600 meters. Use an optimal path evaluation algorithm to comprehensively score the candidate paths based on path length, travel time, and energy consumption indicators. For example, the score of path A is 95 and the score of path B is 85. Determine the optimal path, and select the path with the highest score as the final driving route of the vehicle from the platform to the loading and unloading area. For example, select path A as the final driving route.

[0037] S1011. If abnormal situations such as vehicle failure or emergency tasks occur, the emergency dispatch plan will be activated and the dispatch plan will be dynamically adjusted according to the real-time road conditions and task priorities.

[0038] A real-time monitoring system is used to obtain the operating status of vehicles in the park and detect the occurrence of vehicle failures or emergency tasks. If a vehicle failure or emergency task is detected, the emergency dispatch plan is triggered and the exception handling process is started. According to the real-time road condition data, combined with the three-dimensional geographic information system, the current traffic conditions and available resources in the park are analyzed. The task priority information is obtained, and the urgency and importance of each task are evaluated by combining historical data and real-time data. An efficient scheduling algorithm is used to dynamically adjust task allocation and resource utilization to generate a preliminary emergency dispatch plan. Combined with real-time environmental data and automatic obstacle avoidance functions, the path planning and operation sequence in the dispatch plan are optimized. The adjusted dispatch plan is coordinated with the warehouse management, transportation management and other systems to ensure seamless connection of the entire chain. Through a unified data management platform, the dispatch information is updated in real time to ensure data consistency in each link. The big data analysis tool is used to monitor the execution effect of the emergency dispatch plan in real time and generate an execution report of the dispatch plan.

[0039] Specifically, a real-time monitoring system is used to collect vehicle operation status data in real time through cameras and sensor networks deployed in the park, and a fault detection threshold is set. For example, a vehicle speed below 5 km / h for 10 minutes is considered a fault. If a vehicle fault or an emergency task is detected, the emergency dispatch plan is triggered, the exception handling process is started, and the preset emergency dispatch rule library is called. According to the real-time road condition data and the three-dimensional geographic information system, the current traffic conditions and available resources in the park are analyzed. For example, the idle rate of each platform is calculated. If it is less than 20%, it is considered to be resource-constrained. Obtain task priority information, combine historical data and real-time data, evaluate the urgency and importance of each task, and use a weighted scoring model to set the weight of emergency tasks to 0.7 and the weight of ordinary tasks to 0.3. Use an efficient scheduling algorithm to dynamically adjust task allocation and resource utilization, and generate a preliminary emergency dispatch plan. For example, a genetic algorithm is used to optimize task allocation, and the objective function is to minimize the total waiting time. Combine real-time environmental data and automatic obstacle avoidance functions to optimize the path planning and operation sequence in the dispatch plan. For example, the Dijkstra algorithm is introduced in path planning to avoid congested areas. The adjusted dispatching plan is coordinated with the warehouse management, transportation management and other systems to ensure seamless connection of the entire chain, such as real-time synchronization of task status through API interfaces. Dispatching information is updated in real time through a unified data management platform to ensure data consistency in each link, such as using Kafka to achieve real-time data transmission and storage. Big data analysis tools are used to monitor the execution effect of the emergency dispatching plan in real time and generate execution reports of the dispatching plan, such as using Spark to analyze the dispatching execution data and calculate the task completion rate and resource utilization rate.

[0040] S1013. All scheduling instructions are sent to on-site workers and vehicle drivers in real time through a visual interface and a voice system, and the execution process is monitored and recorded.

[0041] The park platform scheduling system based on multi-source data fusion and intelligent video algorithms can obtain multi-dimensional data such as vehicle arrival time, cargo loading and unloading requirements, and platform idle periods in real time. An efficient scheduling algorithm is used to intelligently match and dynamically schedule vehicle and platform resources to generate scheduling instructions. Through three-dimensional geographic information and real-time environmental data, the optimal path and operation sequence are determined to improve the scheduling instructions. The scheduling instructions are transmitted to the visual interface and the voice system and sent to on-site workers and vehicle drivers in real time. Video stitching and video fusion technologies are used to monitor the execution process of the scheduling instructions in real time to obtain execution status information. Through license plate recognition and face recognition algorithms, the specific information of vehicles and personnel during the execution process is recorded. A unified data management platform is used to clean, store, and process the data during the execution process of the scheduling instructions. Big data analysis tools are used to deeply mine the execution process data to judge the difference between the execution efficiency and the expected goal. BI tools are used to generate visual reports and dashboards of the execution process of the scheduling instructions to display key performance indicators.

[0042] Specifically, the park platform scheduling system based on multi-source data fusion and intelligent video algorithms can obtain multi-dimensional data such as vehicle arrival time, cargo loading and unloading requirements, and platform idle periods in real time. A scheduling algorithm based on deep learning, such as reinforcement learning Q-Learning, is used to intelligently match and dynamically schedule vehicle and platform resources to generate scheduling instructions. Through three-dimensional geographic information and real-time environmental data, the A* algorithm is used to determine the optimal path and operation sequence, and the shortest path is calculated to be 150 meters to improve the scheduling instructions. The scheduling instructions are transmitted to the visual interface and the voice system and sent to the terminal devices of on-site workers and vehicle drivers in real time using the WebSocket protocol. Video stitching and video fusion technologies are used to monitor the execution process of the scheduling instructions in real time using the OpenCV library to obtain execution status information, and the frame rate is 30fps. Through license plate recognition and face recognition algorithms, the YOLOv5 model is used to record the specific information of vehicles and personnel during the execution process, and the recognition accuracy reaches 98%. A unified data management platform is used to clean, store, and process the data during the execution process of the scheduling instructions based on Hadoop, and the data storage capacity is 1TB. Big data analysis tools, such as Spark, are used to deeply mine the execution process data to calculate the difference between the execution efficiency and the expected goal, and the deviation value is controlled within 5%. BI tools, such as Tableau, are used to generate visual reports and dashboards of the execution process of the scheduling instructions to display key performance indicators, including platform utilization rate, vehicle waiting time, etc.

[0043] S1015. Use big data analysis tools to mine and analyze the turnover data of vehicles and goods, summarize the rules of platform operations, optimize the scheduling model and rules, and continuously improve the scheduling efficiency and accuracy.

[0044] Obtain the turnover data of vehicles and goods in the park platform management system, including vehicle arrival time, loading and unloading time, goods type, and platform usage. Use big data analysis tools to clean the turnover data of vehicles and goods, remove outliers and duplicate data, and standardize the data format. Based on the cleaned data, construct a time series model of vehicle and goods turnover, and analyze the time distribution and rules of platform operations. Use machine learning algorithms to train the time series model to identify peak and trough periods of vehicle arrival and goods loading and unloading. According to the training results, extract the key features of platform operations, including vehicle waiting time, loading and unloading efficiency, and platform utilization rate. Combine three-dimensional geographic information and real-time environmental data to conduct spatial analysis on the extracted key features to identify bottleneck areas in platform operations. Use optimization algorithms to iteratively optimize the scheduling model, adjust task allocation and resource utilization strategies, and reduce platform idle time and vehicle waiting time. According to the optimized scheduling model, generate new scheduling rules to dynamically adjust the loading and unloading sequence and operation time of vehicles. Use BI tools to visually display the optimized scheduling rules and generate efficiency reports and KPI indicators for platform operations.

[0045] Specifically, extract the turnover data of vehicles and goods from the park platform management system, including fields such as vehicle arrival time, loading and unloading time, goods type, and platform usage, with a data volume of approximately 1000 records per day. Use Spark to clean the data, use the Z-score method to identify and remove outliers, and standardize the time format to YYYY-MM-DD HH:MM:SS. Based on the cleaned data, construct a time series model, use the ARIMA algorithm to analyze the time distribution of platform operations, and identify 9:00 - 12:00 as the peak vehicle arrival period every day. Use the random forest algorithm to train the time series model to predict the time rules of vehicle arrival and goods loading and unloading, with an accuracy rate of 92%. Extract key features such as vehicle waiting time, loading and unloading efficiency, and platform utilization rate, where the average platform utilization rate is 75%. Combine three-dimensional geographic information and real-time environmental data, use spatial clustering algorithms to analyze the key features, and identify the east side area of the platform as a bottleneck area with a waiting time exceeding 30 minutes. Use the genetic algorithm to iteratively optimize the scheduling model, adjust the task allocation strategy, and reduce the platform idle time by 15%. According to the optimization results, generate dynamic scheduling rules, and the vehicle loading and unloading sequence is adjusted in real time according to the priority of goods type and the idle status of the platform. Use Tableau to visually display the scheduling rules and generate an efficiency report for platform operations. The KPI indicators include average waiting time, platform utilization rate, and operation completion rate.

[0046] The above are only the preferred specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention should cover within the protection scope of the present invention any equivalent replacement or change made according to the technical solution of the present invention and its inventive concept.

Claims

1. A campus platform scheduling system based on multi-source data fusion and intelligent video algorithms, characterized in that, Including: Obtain multi-source data, where the multi-source data includes video surveillance data, sensor data, and business system data; Process the video surveillance data through the optical flow method and dynamic stitching algorithm to achieve seamless integration and real-time monitoring of multiple surveillance videos; Adopt machine learning algorithms to conduct fusion analysis on the multi-source data to achieve intelligent task allocation and resource optimization scheduling; According to the scheduling results, use the optimal path algorithm for path planning, automatic obstacle avoidance, and real-time adjustment to ensure the efficient execution of platform operations; Conduct data display through data visualization technology and BI tools to provide intuitive data support for scheduling decisions.

2. The system according to claim 1, wherein The obtaining of the multi-source data includes: Collect video surveillance data, sensor data, and business system data through high-throughput data acquisition technology and multi-source data access technology, and perform real-time data transmission and storage.

3. The system according to claim 1, wherein The processing of the video surveillance data through the optical flow method and dynamic stitching algorithm includes: Adopt the optical flow method to analyze the motion information between video frames, and combine the dynamic stitching algorithm to achieve seamless integration of multiple videos, eliminate the viewing blind area, and improve the integrity and coherence of the monitoring screen.

4. The system according to claim 1, wherein The conducting of fusion analysis on the multi-source data by adopting machine learning algorithms includes: Use deep learning algorithms such as CNN and RNN to analyze video data to achieve license plate recognition, face recognition, behavior analysis, and anomaly detection; Utilize machine learning algorithms to conduct correlation analysis on various types of data and optimize task allocation and resource scheduling strategies.

5. The system according to claim 4, wherein The optimization of task allocation and resource scheduling strategies by utilizing machine learning algorithms includes: Analyze data such as vehicle arrival time, cargo loading and unloading requirements, and platform idle periods through algorithms, automatically match platform resources, generate the optimal scheduling plan, reduce vehicle waiting time, and improve platform utilization rate.

6. The system according to claim 1, wherein The path planning, automatic obstacle avoidance, and real-time adjustment by utilizing the optimal path algorithm include: Comprehensively consider factors such as platform layout, obstacle distribution, and cargo storage location to plan the optimal driving path of the vehicle in the park; Combine real-time environmental data to dynamically adjust the path, realize the automatic obstacle avoidance function, and ensure the safety and efficiency of vehicle passage.

7. The system according to claim 1, wherein The data display through data visualization technology and BI tools includes: Utilize data visualization tools such as Tableau and Power BI to generate visual reports such as the execution progress of scheduling tasks, platform utilization rate, and key performance indicators; Present the platform operation status in real time through a visual interface to provide intuitive data support for management decisions.

8. The system according to any one of claims 1-7, characterized in that, The system further includes a data management module for cleaning, storing, and managing the collected multi-source heterogeneous data. The data management module includes a distributed database, a NoSQL database, and a data lake architecture, supporting the efficient storage and rapid retrieval of massive data.

Images