A highway rescue command and dispatch decision system based on multi-source data

Abstract

The application relates to a highway rescue command scheduling decision system based on multi-source data, which comprises a fixed video acquisition unit, a mobile video acquisition unit and a central control unit. The fixed video acquisition unit is used for acquiring video and audio information of corresponding road sections of an expressway; the mobile video acquisition unit is used for confirming video and audio information of an accident road section of the expressway; the central control unit receives accident triggering information, extracts key frames containing accident images from the video and audio information sent by the mobile video acquisition unit, and identifies stake position marks to determine an accident calibration interval; based on the coordinates of the starting point and the ending point of the accident calibration interval, fixed video acquisition units overlapping with the accident calibration interval are screened, the optimal fixed video acquisition unit is selected, and picture relay is triggered, so that video and audio information within a preset time before the accident is traced back through a user interface. Through the complementation of the fixed video acquisition unit and the mobile video acquisition unit, rapid accident positioning and data tracing back are realized.

Description

Technical Field

[0001] This invention relates to the fields of traffic management and emergency rescue technology, and in particular to a highway rescue command and dispatch decision-making system based on multi-source data. Background Technology

[0002] With the continuous development of highway networks, traffic accidents and congestion have become increasingly prominent issues. Traditional rescue command systems suffer from problems such as untimely information transmission and low decision-making efficiency. Therefore, a highway rescue command and dispatch decision-making system based on multi-source data that can provide real-time monitoring and rapid response is needed.

[0003] In modern highway management, marker information is typically managed digitally and stored in databases for real-time querying and updates. Highway management departments regularly inspect and maintain the markers to ensure they are clear and intact, and to update the relevant data promptly. This basic knowledge is crucial for understanding highway traffic management, accident handling, and the operation of navigation systems.

[0004] In traffic accident management, accident classification is a key concept. Based primarily on factors such as the extent of vehicle damage, personal injury, and whether hazardous materials are involved, accidents are categorized into five severity levels. First, minor accidents involve only slight vehicle damage (such as scratches or minor collisions), with no personal injury or only minor bodily harm, and low property damage. Second, general accidents involve more severe vehicle damage requiring significant repair costs, usually accompanied by minor injuries requiring medical treatment, and moderate property damage. Third, serious accidents mean that vehicles may be completely destroyed or suffer significant damage, personnel may suffer serious injuries requiring emergency medical assistance, and the amount of property damage is high. Furthermore, fatal accidents are characterized by resulting in death, typically involving multiple vehicles or multiple victims, with extremely severe property damage and personal injury. Finally, special types of accidents include vehicle accidents involving hazardous material leaks, public transportation accidents, or major traffic accidents. These accidents, due to their complex risks (such as hazardous chemical spills or mass casualties) or wide impact, require the activation of special emergency response procedures to ensure effective control of the accident scene and efficient subsequent rescue efforts.

[0005] Although an accident classification system has been established (e.g., classifying accidents into five levels: minor, general, serious, fatal, and special, based on vehicle damage, personal injury, and hazardous materials), the existing system still relies on manual judgment of accident levels and cannot achieve automated classification through multi-source data fusion (e.g., onboard OBD, roadside unit (RSU), and video surveillance). The linkage between classification results and emergency response mechanisms (e.g., medical, fire, and hazardous materials handling resource scheduling) is inefficient, resulting in complex and time-consuming response processes for special types of accidents (e.g., hazardous chemical leaks or major traffic accidents).

[0006] The core technical problems of the current highway rescue command system include: (1) Insufficient accident location accuracy, lack of dynamic correlation between audio and video information and pile location marking information, resulting in location relying on manual labor and large errors; (2) Inefficient matching of video acquisition equipment, lack of intelligent correlation algorithm and real-time transmission technology, video retrieval relies on manual operation; (3) Significant delay in emergency response, video transmission relies on fixed monitoring points, and cannot achieve efficient coverage through the coordinated linkage of mobile and fixed equipment; (4) Lack of automated accident classification, relying on manual judgment, and unable to form a closed loop with the emergency resource dispatch of multiple departments.

[0007] These problems severely restrict the efficiency and reliability of emergency response to highway accidents, and there is an urgent need to optimize the entire process through technological innovation, including precise accident location, intelligent CCTV matching, automated video transmission, and tiered linkage response.

[0008] CN106971541A discloses a highway emergency monitoring system based on unmanned aerial vehicles (UAVs), including: a UAV, a UAV control unit, a relay station, a superior monitoring center, and network equipment. The UAV control unit is installed at toll stations, service areas, bridge and tunnel stations, and / or emergency command vehicles, and is used to acquire, forward, and store video information from the UAV and forward instructions from the superior monitoring center to the UAV. The network equipment is installed on lighting facilities near the UAV control unit and adopts a multi-network convergence mode.

[0009] CN105314122A discloses a drone for emergency command and road obstruction evidence collection. The drone is equipped with a forward-facing binocular camera and a rear-facing binocular camera, as well as a forward-facing binocular image stitching module and a forward-facing monocular vehicle recognition module corresponding to the forward-facing binocular camera, and a rear-facing binocular image stitching module corresponding to the rear-facing binocular camera. The forward-facing binocular camera includes a forward-facing left-eye camera and a forward-facing right-eye camera. The forward-facing binocular image stitching module stitches the video images from the forward-facing binocular camera, outputting a forward-stitched image and a forward-facing right-eye image. The rear-facing binocular image stitching module stitches the video images from the rear-facing binocular camera, outputting a rear-facing stitched image. The monocular vehicle recognition module identifies vehicle attributes based on the video image from the forward-facing right-eye camera.

[0010] As mentioned above, existing technologies can only monitor whether an accident has occurred, but cannot accurately classify traffic accidents. How to accurately assess information from a traffic accident scene and effectively deploy rescue efforts remains a current technological challenge.

[0011] Furthermore, on the one hand, there are differences in understanding among those skilled in the art; on the other hand, the applicant studied a large number of documents and patents when making this invention, but due to space limitations, not all details and contents were listed in detail. However, this does not mean that the present invention does not possess the features of these prior art. On the contrary, the present invention already possesses all the features of the prior art, and the applicant reserves the right to add relevant prior art to the background art. Summary of the Invention

[0012] The current highway rescue command system suffers from four major pain points: First, accident location has significant errors, primarily due to the lack of a dynamic correlation mechanism between audio-visual information and road marker information, requiring manual intervention and compromising accuracy. Second, video equipment retrieval is inefficient; existing systems lack intelligent matching algorithms and real-time transmission technology, relying on manual operation and limiting response speed. Third, emergency response is noticeably delayed; current video transmission systems heavily depend on fixed monitoring points, failing to achieve coordinated operation between mobile video acquisition units and fixed facilities, resulting in limited information acquisition channels. Fourth, accident classification lacks automated judgment capabilities, currently relying on manual assessment and failing to form a closed-loop management system with multi-department emergency resource dispatch, impacting overall coordination efficiency.

[0013] These problems collectively lead to efficiency bottlenecks and reliability deficiencies in emergency response to highway accidents. To overcome the current predicament, four major improvements are urgently needed through technological innovation: establishing a dynamic correlation model between mobile video and road markers to improve positioning accuracy; developing intelligent video matching algorithms and transmission technologies to achieve automated equipment deployment; constructing a collaborative transmission network for mobile and fixed equipment to shorten response time; and developing an intelligent hierarchical system and opening up emergency resource dispatch channels across multiple departments, thereby forming a closed-loop management mechanism for the entire process.

[0014] To address the shortcomings of existing technologies, this invention provides a highway rescue command and dispatch decision-making system based on multi-source data, comprising a fixed video acquisition unit, a mobile video acquisition unit, and a central control unit. The fixed video acquisition unit is used to collect audio and video information of the corresponding section of the highway; the mobile video acquisition unit, carried by patrol personnel or used by drones, is used to confirm the audio and video information of the accident section of the highway; the central control unit receives accident triggering information from the mobile video acquisition unit, extracts keyframes containing accident images from the audio and video information sent by the mobile video acquisition unit, and identifies pile markers to determine the accident calibration interval; based on the coordinates of the start and end points of the accident calibration interval, at least one fixed video acquisition unit whose coverage area overlaps with the accident calibration interval is selected, the optimal fixed video acquisition unit is selected, and the video is triggered for broadcast, thereby allowing the audio and video information collected by the fixed video acquisition unit within a preset time before the accident to be reviewed through the user interface.

[0015] According to a preferred embodiment, the central control unit includes a data processing module and an intelligent analysis module. The data processing module receives accident triggering information sent by the mobile video acquisition unit, extracts keyframes containing accident images from the audio-visual information sent by the mobile video acquisition unit, and identifies pile location markers. If no pile location marker is identified, the data processing module triggers an alarm and returns an audio-visual information re-acquisition command to the mobile video acquisition unit. If a pile location marker is identified, the data processing module determines the start and end points of the accident calibration interval and sends these points to the intelligent analysis module to select a fixed video acquisition unit. This step effectively solves the problem of time-consuming image feature recognition and reduces the time cost of manual operation. Automating the initial confirmation of the accident location greatly improves the system's response speed and accuracy, ensuring rapid location of the accident point even in complex environments, providing crucial support for subsequent rescue operations.

[0016] According to a preferred embodiment, the system also includes a database communicatively connected to the central control unit. The coordinates of the pile location markers, along with the coordinates and orientation data of the fixed video acquisition unit, are stored in the database. This allows the intelligent analysis module in the central control unit to quickly locate and confirm the position of the pile location markers. This significantly improves system query efficiency and shortens accident response time. Simultaneously, this also means that the system can quickly find the most suitable monitoring equipment to provide continuous footage before and after an accident, enhancing emergency decision-making support capabilities and enabling the command center to make faster response and dispatch decisions.

[0017] According to a preferred embodiment, the intelligent analysis module is configured to: upon receiving pile location markers corresponding to the start and end points of the accident calibration interval, send the pile location markers and extraction instructions to the database to receive the coordinates of the start and end points from the database; thereby filtering fixed video acquisition units whose coverage areas overlap with the accident calibration interval; calculating the overlap degree of the captured images between the filtered fixed video acquisition units; and prioritizing the fixed video acquisition units based on the overlap degree, thereby selecting the optimal fixed video acquisition unit to retrieve the audio-visual information of the accident. This step optimizes resource allocation, ensuring the best viewing angle and minimal overlapping coverage of the accident images. This not only improves the quality of video data but also accelerates the comprehensive assessment process of the accident situation, helping to coordinate emergency resources from multiple departments more efficiently and improving the speed and effectiveness of the overall emergency response.

[0018] According to a preferred embodiment, the data processing module is further configured to: identify pile location images from the audio-visual information acquired by the mobile video acquisition unit, and extract the locations of several pile location markers from the audio-visual information based on the pile location images. This function increases the system's flexibility and adaptability, enabling relatively accurate estimation of the accident location even when the full picture of the accident cannot be directly observed. This is crucial for rapid location and timely response, especially in emergency situations, effectively reducing response time and improving rescue efficiency.

[0019] According to a preferred embodiment, the data processing module is further configured to: extract keyframes containing accident images from the audio-visual information acquired by the mobile video acquisition unit, and determine the accident location using spatial interpolation; wherein, the accident location is determined as follows: in the audio-visual information acquired by the mobile video acquisition unit, the location of the pile marker with the timestamp closest to the accident trigger time is taken as the starting point, and the nearest downstream pile marker along the highway's direction of travel is taken as the ending point, with the accident location situated within the accident calibration interval between the starting and ending points. This method not only improves the accuracy of accident location but also simplifies subsequent information processing and accelerates the initiation of rescue operations. Furthermore, it enhances the understanding of the accident environment, contributing to the development of more effective rescue strategies.

[0020] According to a preferred embodiment, the intelligent analysis module dynamically matches fixed video capture units as follows: After filtering fixed video capture units whose coverage area overlaps with the accident calibration area, it retrieves the coordinates and orientation data of the fixed video capture units from the database, prioritizing fixed video capture units with consistent or similar orientation data, and calculating the overlap of image capture between the filtered fixed video capture units based on their coordinates and orientation data. This ensures the optimal viewing angle for image capture and avoids unnecessary duplicate coverage. This method not only improves the quality of video data but also enhances the efficiency of collaboration between different departments, promotes the opening of multi-department emergency resource dispatch channels, and forms an efficient collaborative working mode.

[0021] According to a preferred embodiment, the intelligent analysis module sends the coordinates and / or number of the selected fixed video capture unit to the data processing module. The data processing module determines the time range of the incident based on the incident triggering information and sends this time range information to the selected fixed video capture unit. This allows the selected fixed video capture unit to broadcast audio and video information from a preset time period before the incident to the user interface based on the time range information. This provides crucial data for incident analysis and significantly enhances the speed and effectiveness of emergency response. Through a closed-loop management mechanism, the system achieves seamless connection from incident discovery, location, assessment to dispatch, significantly improving overall collaborative efficiency and emergency response capabilities.

[0022] According to a preferred embodiment, the system also includes a drone equipped with a mobile video acquisition unit. The drone is communicatively connected to a central control unit and is used to fly to the accident location and transmit audio and video information of the accident scene to the central control unit in real time. The drone captures key details such as the positional relationships of accident vehicles and the distribution of debris from top-down and surround-view perspectives, assisting in accurately assessing the severity of the accident.

[0023] According to a preferred embodiment, the system also includes a user interface, which is connected to the central control unit via a mobile terminal. The user interface provides an intuitive graphical interface on the mobile terminal to display audio-visual information and command and dispatch plans, making it convenient for command personnel to view the command and dispatch plans and perform interactive operations anytime and anywhere.

[0024] With this setup, the mobile device and the central control unit synchronize data in real time, supporting remote collaboration among command teams, ensuring consistency of instructions, and avoiding information discrepancies caused by traditional walkie-talkie communication.

[0025] Beneficial technical effects of the present invention:

[0026] (1) This invention solves the problem of long image feature recognition time by using rule matching and efficient utilization of pile location data. At the same time, it combines the complementary mechanism of fixed and mobile video acquisition units to achieve fast and low-cost accident location and data backtracking.

[0027] (2) This invention filters out the fixed video acquisition units that actually capture the accident area by the overlap of the coverage area, avoiding the retrieval of irrelevant data. This invention prioritizes fixed video acquisition units with the same or similar orientation to ensure that the images captured by different fixed video acquisition units have spatiotemporal consistency in terms of viewing angle, lighting, and target direction, which facilitates subsequent analysis.

[0028] (3) This invention allows users to review audio and video information collected by a fixed video capture unit within a preset time period before an accident occurs via a user interface. By reviewing the data, key parameters such as traffic flow, vehicle speed changes, and vehicle spacing before the accident can be reconstructed, assisting in determining liability for the accident (such as speeding or illegal lane changes). Combined with continuous recordings from the fixed video capture unit, it can be determined whether the accident was caused by driver error or sudden road conditions. Attached Figure Description

[0029] Figure 1 This is a schematic diagram of the overall architecture of the highway rescue command and dispatch decision-making system based on multi-source data provided by the present invention.

[0030] Figure 2 This is a schematic diagram of another architecture of the highway rescue command and dispatch decision-making system based on multi-source data provided by the present invention.

[0031] Figure 3This is a flowchart illustrating the highway rescue command and dispatch decision-making system based on multi-source data provided by the present invention.

[0032] List of reference numerals in the attached diagram:

[0033] 100: Fixed video acquisition unit; 200: Mobile video acquisition unit; 300: Central control unit; 310: Data processing module; 320: Intelligent analysis module; 330: Command and dispatch module; 400: Database; 500: Communication module; 600: UAV; 700: User interface; 800: Mobile terminal. Detailed Implementation

[0034] The following is a detailed explanation with reference to the accompanying drawings.

[0035] The current highway rescue command system suffers from four major pain points: First, accident location has significant errors, primarily due to the lack of a dynamic correlation mechanism between audio-visual information and road marker information, requiring manual intervention and compromising accuracy. Second, video equipment retrieval is inefficient; existing systems lack intelligent matching algorithms and real-time transmission technology, relying on manual operation and limiting response speed. Third, emergency response is noticeably delayed; current video transmission systems heavily depend on fixed monitoring points, failing to achieve coordinated operation between mobile and fixed video acquisition units, resulting in limited information acquisition channels. Fourth, accident classification lacks automated judgment capabilities, currently relying on manual assessment and failing to form a closed-loop management system with multi-department emergency resource dispatch, impacting overall coordination efficiency.

[0036] These problems collectively lead to efficiency bottlenecks and reliability deficiencies in emergency response to highway accidents. To overcome the current predicament, four major improvements are urgently needed through technological innovation: establishing a dynamic correlation model between mobile video and road markers to improve positioning accuracy; developing intelligent video matching algorithms and transmission technologies to achieve automated equipment deployment; constructing a collaborative transmission network for mobile and fixed equipment to shorten response time; and developing an intelligent hierarchical system and opening up emergency resource dispatch channels across multiple departments, thereby forming a closed-loop management mechanism for the entire process.

[0037] To address the shortcomings of existing technologies, this invention provides a highway rescue command and dispatch decision-making system based on multi-source data. This invention can also provide a highway emergency management system based on pile location marking information, a method for managing highway pile locations and their marking positions, and an emergency rescue command system based on unmanned aerial vehicles (UAVs). This invention can also be an artificial intelligence-based emergency rescue command system.

[0038] Example 1

[0039] To address the shortcomings of existing technologies, this invention provides a highway rescue command and dispatch decision-making system based on multi-source data. The system includes at least one fixed video acquisition unit 100, at least one mobile video acquisition unit 200, and a central control unit 300. Figure 1 and Figure 3 As shown.

[0040] The fixed video acquisition unit 100 consists of closed-circuit television (CCTV) cameras deployed on key sections of the highway. Each fixed video acquisition unit 100 includes an optical sensor module, an infrared illumination device, and a protective housing, used for continuous video and audio data acquisition of a predetermined section of the highway around the clock. The fixed video acquisition unit 100 is used to acquire audio-visual information of the corresponding section of the highway. The fixed video acquisition unit 100 is connected to the central control unit 300 via wired or wireless means.

[0041] The mobile video acquisition unit 200 comprises two device forms. One is a portable digital video recording device carried by patrol personnel, with a built-in video encoding chip and removable storage media, equipped with a shockproof shell and waterproof interface. The other is a dedicated video acquisition device mounted on a drone 600, integrating a high-definition gimbal camera, digital video encoder, and wireless communication module, supporting aerial hovering shooting. The mobile video acquisition unit 200 is used to confirm audio and video information of accident sections on highways.

[0042] The central control unit 300 is connected to the fixed video acquisition unit 100 and the mobile video acquisition unit 200 via a communication module 500 in wired and wireless ways, respectively. It is used to receive, process, and analyze audio and video information from the fixed video acquisition unit 100 and the mobile video acquisition unit 200, and to coordinate the operation of the entire system. Preferably, the communication module 500 is equipped with a multi-channel network switching device, including a fiber optic transceiver and a cellular mobile communication base station interface module.

[0043] Preferably, the fixed video acquisition unit 100 is directly connected to the communication module 500 of the central control unit 300 via an optical fiber cable. The optical fiber cable comprises multiple sets of glass fiber transmission cores and a metal reinforcing layer to ensure stable transmission of audio and video information to the central control unit 300 via optical signals. Preferably, the fixed video acquisition unit 100 and the central control unit 300 are connected via single-mode optical fiber to meet the requirements of high-speed transmission of audio and video information and interference resistance on highways.

[0044] The mobile video acquisition unit 200 establishes a wireless connection with the central control unit 300 via a cellular mobile communication base station. Portable devices carried by patrol personnel upload data via a built-in fifth-generation mobile communication module. The drone 600 establishes a low-latency communication link with the mobile communication base station via a dedicated antenna array.

[0045] Preferably, the central control unit 300 receives accident triggering information from the mobile video acquisition unit 200, extracts keyframes containing accident images from the audio-visual information sent by the mobile video acquisition unit 200, and identifies pile location markers to determine the accident calibration interval. The central control unit 300 filters at least one fixed video acquisition unit 100 whose coverage area overlaps with the accident calibration interval based on the coordinates of the start and end points of the accident calibration interval. The central control unit 300 selects the optimal fixed video acquisition unit 100 and triggers video playback, thereby allowing the user interface 700 to review the audio-visual information acquired by the fixed video acquisition unit 100 within a preset time period before the accident occurred.

[0046] like Figure 1 and Figure 3 As shown, the central control unit 300 includes a data processing module 310 and an intelligent analysis module 320. The data processing module 310 is a server equipped with a graphics processing unit (GPU) accelerator card, connected to the communication module 500 via a high-speed data bus. It is responsible for receiving and parsing the audio-visual information transmitted from the fixed video acquisition unit 100 and the mobile video acquisition unit 200. The data processing module 310 receives accident triggering information from the mobile video acquisition unit 200, extracts keyframes containing accident images from the audio-visual information sent by the mobile video acquisition unit 200, and identifies pile location markers to determine the accident calibration range. A keyframe specifically refers to a frame type (i.e., an I-frame) in the video stream that carries complete image information. It uses intra-frame compression coding technology to independently store pixel matrix data. Compared to prediction frames, keyframes achieve a basic compression ratio of approximately 10:1 through discrete cosine transform quantization and entropy coding algorithms, while maintaining the independence of image decoding.

[0047] The intelligent analysis module 320 employs server hardware equipped with a high-performance computing chip and is connected to the data processing module 310 via a gigabit Ethernet cable. The intelligent analysis module 320 contains a preset analysis algorithm that, based on the coordinates of the start and end points of the accident calibration interval, filters at least one fixed video acquisition unit 100 whose coverage area overlaps with the accident calibration interval, and selects the optimal fixed video acquisition unit 100.

[0048] Preferably, the data processing module 310 is connected to the wired interface of the communication module 500 via a shielded twisted-pair cable to receive audio and video information transmitted from the fixed video acquisition unit 100 via a fiber optic network. The data processing module 310 also receives 4G / 5G data streams uploaded by the mobile video acquisition unit 200 via a wireless communication module. The intelligent analysis module 320 and the data processing module 310 exchange parsed keyframes and incident calibration parameters via a copper backplane, and simultaneously access the historical incident database in the data storage array via a gigabit Ethernet cable.

[0049] According to a preferred embodiment, such as Figure 3 As shown, the system also includes a database 400. The coordinates of the pile location markers, along with the coordinates and orientation data of the fixed video acquisition unit 100, are stored in the database 400 to allow the intelligent analysis module 320 in the central control unit 300 to quickly locate and confirm the position of the pile location markers. The positions of the pile location markers and the positions of the fixed video acquisition units 100 before and after them are stored in the database 400 in a mutually matching manner to improve the accuracy of the intelligent analysis module 320 in finding the position of the fixed video acquisition unit 100 based on the database 400.

[0050] The physical hardware of the Database 400 can include a distributed storage server cluster and a high-speed cache acceleration device. The storage servers are configured with dual-port Fibre Channel hard disk arrays, using high-speed NVMe solid-state drives as the primary storage medium, and are equipped with a tape library for offline backup. The cache acceleration device integrates a dynamic random access memory stack and a field-programmable gate array chip, directly connected to the storage servers via a backplane bus to form a tiered storage architecture. The Database 400 server is equipped with redundant power supply modules and hot-swappable fan assemblies to ensure continuous and stable operation in harsh highway environments.

[0051] Database 400 establishes a communication connection with the intelligent analysis module 320 of the central control unit 300 via a gigabit Ethernet cable, and is also directly connected to the data processing module 310 via a high-speed data bus. At the physical cabling level, the database 400 server is configured with redundant fiber optic channel interfaces, which are respectively connected to the core switching equipment of the central control unit 300 and the expansion ports of the data storage array, forming a dual-path data transmission guarantee mechanism.

[0052] In the system architecture of this invention, the data processing module 310 extracts keyframes based on the following two technical dimensions by real-time parsing of the video stream encoding structure (such as MPEG-TS container format):

[0053] (1) Content-aware extraction: The background subtraction algorithm is run through the edge computing device. When the detection value of moving targets in the video stream exceeds the preset threshold (such as pixel change rate ≥ 15%), the current key frame is forcibly extracted.

[0054] (2) Temporal rule extraction: Keyframes are extracted periodically at fixed intervals to ensure the integrity of the temporal dimension sampling. For example, the fixed interval can be 2 seconds, corresponding to a GOP length of 50 frames (25fps).

[0055] The metadata fields in the extracted keyframe audio-visual information include: timestamp (1ms precision), spatial coordinates (WGS-84 coordinate system), device identifier, and compression parameters (QP value). This structured data is directly written to the GPU memory buffer of the intelligent analysis module 320 via RDMA (Remote Direct Memory Access) technology, bypassing the PCIe switch, providing input tensors for subsequent convolutional neural networks (such as the ResNet-50 feature extractor). The system of this invention achieves a single-frame processing latency of ≤15ms at 1080P resolution, while reducing the transmission bandwidth requirement to 1 / 30th of that of continuous video streams, forming a dual optimization mechanism for data compression and event detection.

[0056] During data interaction, when performing pile location coordinate matching, the data processing module 310 sends a query command containing the pile number to the database 400 via the Ethernet protocol. Upon receiving the command, the database 400 retrieves the corresponding entry from its stored pile location and GPS coordinate mapping table, encapsulates the matched latitude and longitude coordinates and the associated front and rear fixed video acquisition unit 100 device numbers into a data packet, and returns it. When the inspection personnel initiate a pile location verification request through the user interface 700, the communication module 500 forwards the request command to the database 400. The database 400 simultaneously retrieves the geographical coordinates of the three sets of fixed video acquisition units 100 surrounding the target pile location and packages the coordinate set and real-time video stream thumbnails via a cellular mobile communication base station, sending them to the portable mobile video acquisition unit 200.

[0057] During the data update phase, the fixed video acquisition unit 100 periodically sends a self-test report to the database 400 via a fiber optic network. This report includes the coordinates of the fixed video acquisition unit 100 itself and calibration parameters for adjacent pile positions. Upon receiving the coordinates from the fixed video acquisition unit 100, the database 400 initiates a transaction processing mechanism to atomically update the matching table between the pile position markers and the video acquisition unit positions. It then sends a data version update notification to the data processing module 310 via the backplane bus, ensuring that the algorithm model always retrieves the latest version of the location data.

[0058] According to a preferred embodiment, such as Figure 1 and Figure 2 As shown, the system also includes a drone 600. The drone 600 is equipped with a mobile video acquisition unit 200. The drone 600 is communicatively connected to the central control unit 300, and is used to fly to the accident location and transmit audio and video information from the accident scene to the central control unit 300 in real time.

[0059] Specifically, the mobile video acquisition unit 200 mounted on the UAV 600 includes a high-definition gimbal camera, an embedded video encoder, and a multi-band wireless communication module. The gimbal camera integrates an optical image stabilization mechanism and a three-axis mechanically stabilized gimbal, while the video encoder has a built-in hardware decoding chip supporting H.265 compression. The mobile video acquisition unit 200 establishes a communication connection with the central control unit 300 through a directional antenna array integrated into the UAV 600 itself. This directional antenna array includes two operating frequency bands: a 2.4GHz band for transmitting control commands and a 5.8GHz band dedicated to high-definition video streaming. The communication connection between the UAV 600 and the central control unit 300 is achieved through a cellular mobile communication base station, specifically employing a fifth-generation mobile communication technology module to establish a low-latency transmission channel, while also being equipped with a redundant microwave communication link as a backup transmission path.

[0060] Compared to patrol personnel, the drone 600 can reach the accident site at high altitudes more quickly and communicate with the central control unit 300. Therefore, the drone 600 can proactively patrol the accident site. In contrast to existing technologies where the drone 600 is operated by patrol personnel to reach the accident site and can only collect images without directly communicating with the central control unit 300 or exchanging data, the drone 600 of this invention proactively reaches the accident site and actively sends data to the central control unit 300, thus gaining more time for analysis and dispatching, and improving the efficiency of accident handling.

[0061] According to a preferred embodiment, such as Figure 1 and Figure 2 As shown, the system also includes a user interface 700. The user interface 700 is communicatively connected to the central control unit 300. The user interface 700 provides an intuitive graphical interface on the mobile terminal 800 to display audio-visual information and command and dispatch plans, making it convenient for command personnel to view command and dispatch plans and perform interactive operations anytime and anywhere.

[0062] The physical hardware involved in the user interface 700 includes a touch display device of the mobile terminal 800 and a remote interaction terminal. The touch display device uses a capacitive multi-touch screen and integrates a fingerprint recognition module. The remote interaction terminal is equipped with a dedicated communication encryption chip and a dual-SIM dual-standby RF module. The user interface 700 establishes a communication connection with the communication module 500 of the central control unit 300 through a cellular mobile communication network, specifically using mobile base station backhaul technology to achieve bidirectional data transmission, and simultaneously establishing a high-speed auxiliary channel through a wireless access point of the 802.11ax protocol in a local area network environment. Preferably, the Bluetooth 5.2 module built into the touch display device can also establish a point-to-point connection with the portable printing device of the on-site inspection personnel for the visualization output of the dispatch plan.

[0063] The interaction process of this system is described below.

[0064] S1: Data processing module 310 receives accident triggering information sent by mobile video acquisition unit 200.

[0065] S2: The data processing module 310 extracts key frames containing accident images from the audio-visual information sent by the mobile video acquisition unit 200 and identifies pile location markers.

[0066] After receiving audio-visual information containing accident footage sent by the mobile video acquisition unit 200, the data processing module 310 extracts the two frames with timestamps closest to the accident trigger time from the audio-visual information (e.g., if the trigger time is 14:30:00, frames 14:29:59 and 14:30:01 are selected). The data processing module 310 selects the pile position with the timestamp closest to the trigger time from the pile position markers identified in the matching frames as the starting point (e.g., pile position KXX+XX). Following the highway travel direction (e.g., vehicle direction), the data processing module 310 identifies the downstream pile position marker closest to the starting point (e.g., pile position KXX+ZZ) in subsequent video frames as the ending point. The data processing module 310 retrieves the geographical coordinates and distance (in kilometers) of the starting and ending points from the database 400. The data processing module 310 calculates the time difference (e.g., -1 second) between the starting point pile position identification time and the accident trigger time. Based on the vehicle's average speed (e.g., assuming 80 km / h), the data processing module 310 calculates the vehicle's displacement from the starting point at the moment the accident was triggered (e.g., 80 × (1 / 3600) × 1 ≈ 0.022 km).

[0067] Preferably, the data processing module 310 identifies pile location images from the audio-visual information acquired by the mobile video acquisition unit 200. The data processing module 310 then extracts several pile location marker positions from the audio-visual information based on the pile location images.

[0068] Preferably, the data processing module 310 uses deep learning models such as YOLO or Faster R-CNN to detect the physical contours of the pile location markers (such as metal plaques or numerical markings for highway pile locations) in the video frame. Optical character recognition (OCR) is performed on the detected pile location marker areas to extract the pile location numbers (e.g., KXX+XX). The pixel coordinates of the pile location markers in the video frame are recorded and converted into geographic coordinates using the calibration parameters (focal length, field of view) of the fixed video acquisition unit 100, thereby determining the locations of several pile location markers.

[0069] If no pile location marker is detected, the data processing module 310 triggers an alarm and sends a command to the mobile video acquisition unit 200 to re-acquire audio and video information. If a pile location marker is detected, the data processing module 310 determines the start and end points of the accident calibration interval.

[0070] Determining location based on the identification of fixed pile location markers is reliable. This invention uses fixed pile location markers on highways as a geographical reference, avoiding the positioning error problem of the mobile video acquisition unit 200.

[0071] S3: Data processing module 310 calculates the accident calibration interval.

[0072] Preferably, the data processing module 310 calculates the accident calibration interval and displacement.

[0073] The data processing module 310 extracts key frames containing accident images from the audio and video information collected by the mobile video acquisition unit 200 and determines the accident location using spatial interpolation.

[0074] The data processing module 310 determines the accident location in the following way:

[0075] In the audio and video information collected by the mobile video acquisition unit 200, the position of the pile mark closest to the time of the accident triggering is taken as the starting point, and the position of the nearest downstream pile mark along the direction of highway traffic is taken as the ending point. The accident location is located in the accident calibration interval between the starting point and the ending point.

[0076] Spatial interpolation uses timestamps to match the nearest pile location markers as the starting and ending points, and combines this with the average vehicle speed to calculate the displacement. This method can quickly narrow down the accident location range with an error controlled within 20 meters, which meets the needs of practical applications.

[0077] First, determine the starting point pile location.

[0078] Let the time of the accident be t. acc The timestamps of the two frames extracted by the mobile video acquisition unit are t1 (previous frame) and t2 (later frame), which satisfy t1 <t acc ≤t2. The previous frame is the frame before the time of the accident triggering. The next frame is the frame after the time of the accident triggering.

[0079] The sets of pile location markers identified in the previous frame t1 and the subsequent frame t2 are as follows:

[0080] ; .

[0081] n represents the total number of pile location markers identified in the previous frame t1. That is, the total number of all pile location markers detected in the video frame t1 (e.g., if 3 pile locations are detected, then n=3). m represents the total number of pile location markers identified in the subsequent frame t2. That is, the total number of all pile location markers detected in the video frame t2 (e.g., if 2 pile locations are detected, then m=2). P1 represents the set of pile location markers detected by the image recognition algorithm in the previous frame. P2 represents the set of pile location markers detected in the subsequent frame.

[0082] The starting point pile location is:

[0083] .

[0084] This represents the timestamp of the frame corresponding to pile position p. This indicates the starting pile position. P represents the pile position.

[0085] Secondly, determine the location of the termination point pile.

[0086] Assuming the direction of travel along the highway is positive, the terminal point is at stake K. end for:

[0087] .

[0088] This represents the set of pile positions in the frame corresponding to time t. Indicates the location of the termination point pile.

[0089] The data processing module 310 sends pile location markers and extraction instructions to the database 400 to receive the coordinates of the starting point and the ending point from the database 400.

[0090] The coordinates of the starting point pile location are: C start =(x start ,y start ,z start (WGS-84).

[0091] The coordinates of the termination point pile location are: C end =(x end ,y end ,z end ).

[0092] (x start ,y start ,z start ), (x end ,y end ,z end ) represent the three-dimensional coordinates of the starting point pile position and the ending point pile position, respectively.

[0093] The distance between the starting point and the ending point is:

[0094] .

[0095] D represents the horizontal projected distance between the starting point and the ending point (ignoring elevation differences), in meters (m).

[0096] The time difference is: .

[0097] This indicates the timestamp of the starting point stake location in database 400.

[0098] This represents the difference between the time the accident was triggered and the timestamp of the starting point pile location, expressed in seconds (s).

[0099] Let the vehicle speed be: .

[0100] The offset of the accident point is: (Unit: kilometers)

[0101] The accident calibration range is: [K] start , K end ].

[0102] K start K represents the coordinates of the pile position at the starting point of the accident. end The coordinates of the pile position at the point where the accident terminates are indicated.

[0103] This invention utilizes fixed pile location markers and spatial interpolation to determine the accident location, which relies on accurate pile location coordinate data and precise timestamps. When this basic data is sufficiently accurate and updated in real time, the positioning accuracy is high.

[0104] The data processing module 310 needs to process the video stream quickly to identify the accident and calculate its location. Although the image feature recognition by the data processing module 310 takes a long time (more than 2 seconds), which may affect the immediacy of the accident response, the rule matching by the subsequent intelligent analysis module 320 takes less time (less than 50ms), which helps to improve the overall efficiency.

[0105] The data processing module 310 sends the start and end points of the accident calibration interval to the intelligent analysis module 320 to select the fixed video acquisition unit 100.

[0106] S4: After receiving the pile location markers corresponding to the start and end points of the accident calibration interval from the data processing module 310, the intelligent analysis module 320 sends the pile location markers and extraction instructions to the database 400 to receive the coordinates of the start and end points from the database 400 in order to filter the fixed video acquisition units 100 whose coverage intervals overlap with the accident calibration interval.

[0107] Alternatively, the intelligent analysis module 320 can directly receive the coordinates corresponding to the start and end points of the accident calibration interval from the data processing module 310.

[0108] S5: Intelligent analysis module 320 filters fixed video acquisition units 100 whose coverage area overlaps with the accident calibration area.

[0109] Coverage area of ​​each fixed video capture unit 100 for:

[0110] .

[0111] This represents the coverage area of ​​the i-th fixed video acquisition unit 100, that is, the range of highway mileage markers that the fixed video acquisition unit 100 can capture. This represents the coordinates of the starting pile position covered by the i-th fixed video acquisition unit 100. This represents the coordinates of the termination pile position covered by the i-th fixed video acquisition unit 100.

[0112] The accident calibration interval Acc is: Acc = [K] start , K end ].

[0113] Therefore, the overlap between the coverage area of ​​the fixed video acquisition unit 100 and the accident calibration area is defined as:

[0114] .

[0115] This represents the overlap length (in meters) between the coverage area of ​​the i-th fixed video acquisition unit 100 and the accident calibration area. The value represents the smaller of the accident termination point and the termination point of the fixed video acquisition unit 100, and represents the right boundary of the overlapping interval. The value representing the larger of the accident initiation point and the initiation point of the fixed video acquisition unit 100 is used to define the left boundary of the overlapping interval. Subtracting the left boundary from the right boundary results in a positive value if there is overlap, and a negative or zero value if there is no overlap.

[0116] This invention calculates the overlap between the coverage area of ​​each fixed video acquisition unit 100 and the accident calibration area, and selects the most relevant fixed video acquisition unit 100 (i.e. the fixed video acquisition unit 100 covering the most accident areas). This avoids retrieving data from all fixed video acquisition units 100 and only transmits video clips related to the accident, thereby reducing redundant data.

[0117] For example, the coverage area is: =[KX1+500, KX2+200]. X2-X1=1.

[0118] The accident calibration interval is: Acc = [KX1 + 800, KX2 + 100].

[0119] .

[0120] The coverage area of ​​fixed video acquisition unit 100 overlaps with the accident calibration area by 300 meters, so its video data needs to be retrieved.

[0121] S6: The intelligent analysis module 320 calculates the overlap of the video capture between the selected fixed video capture units 100, and sorts the fixed video capture units 100 by priority based on the overlap.

[0122] If only fixed video capture units 100 whose coverage area overlaps with the accident calibration area are selected, ignoring orientation data, it may lead to the selection of fixed video capture units 100 with opposite shooting directions, potentially resulting in duplicate target identification. For example, one fixed video capture unit 100 might capture the front of the vehicle, while another captures the rear, making trajectory reconstruction difficult. The same vehicle may be misidentified as multiple targets in different fixed video capture units 100, interfering with accident liability determination. Ignoring orientation data and failing to prioritize fixed video capture units 100 with consistent orientation may also result in missing the optimal viewing angle. This could lead to key accident frames not being captured (e.g., the moment of collision being recorded only by a fixed video capture unit 100 in one direction), requiring manual review and delaying insurance claims or legal evidence collection.

[0123] Relying on traditional algorithms to calculate the position and orientation of all fixed video acquisition units 100 in real time would result in high computational complexity, especially when the number of fixed video acquisition units 100 is large. The response time would be uncontrollable and would not meet the real-time requirements of traffic accident handling (e.g., locking the accident video within 5 minutes).

[0124] Therefore, to address these technical shortcomings, the intelligent analysis module 320 dynamically matches the fixed video acquisition units 100 as follows: after filtering fixed video acquisition units 100 whose coverage area overlaps with the accident calibration area, the intelligent analysis module 320 retrieves the coordinates and orientation data of the fixed video acquisition units 100 from the database 400, prioritizing fixed video acquisition units 100 with consistent or similar orientation data. Preferably, by utilizing the database 400 to retrieve data, the data matching method can reduce computation time compared to traditional intelligent algorithms. The intelligent analysis module 320 calculates the overlap of the video captures between the filtered fixed video acquisition units 100 based on their coordinates and orientation data.

[0125] This invention filters fixed video capture units 100 that actually capture images of the accident area by assessing the overlap of coverage areas, avoiding the retrieval of irrelevant data. This invention prioritizes fixed video capture units 100 with consistent or similar orientations, ensuring spatiotemporal consistency in viewpoint, lighting, and target direction across images captured by different fixed video capture units 100, facilitating subsequent analysis (such as vehicle trajectory reconstruction). Therefore, this invention can retrieve only data from fixed video capture units 100 covering the accident area and with matching viewpoints, avoiding the transmission of large amounts of irrelevant video and saving bandwidth and storage resources. This invention can also directly retrieve pre-stored coordinate and orientation data from a database, skipping complex algorithms for real-time calculation of positional relationships, reducing computational latency, and thus accelerating accident analysis. This invention can also utilize fixed video capture units 100 with consistent orientations to provide complementary or continuous images, avoiding misjudgments caused by differences in viewpoint (e.g., the same vehicle being misidentified as multiple vehicles in different fixed video capture units 100), improving decision-making reliability.

[0126] Specifically, the intelligent analysis module 320 filters fixed video acquisition units 100 that meet the following conditions: .

[0127] In other words, the two overlap if and only if the maximum value at the starting point of the coverage interval and the minimum value at the ending point are less than the minimum value at the ending point.

[0128] This represents the overlap length between the coverage area of ​​the i-th fixed video acquisition unit 100 and the accident calibration area. This indicates that the coverage area and the accident calibration area have an intersection (i.e., partial overlap). This means that if there is an intersection, the overlap length between the coverage area of ​​the i-th fixed video acquisition unit 100 and the accident area must be greater than zero.

[0129] The intelligent analysis module 320 sorts the fixed video acquisition units 100 according to priority.

[0130] The degree of overlap is ordered in descending order as follows: .

[0131] This represents the overlap length between the coverage area of ​​the j-th fixed video acquisition unit 100 and the accident calibration area. The greater the overlap, the higher the priority of the fixed video acquisition unit 100, and its image is retrieved first.

[0132] If the coverage direction of the fixed video capture unit 100 is opposite to the traffic direction, its footage may only capture the rear of the vehicle or key dynamics of the accident area that cannot be covered (e.g., the moment of vehicle collision, driver behavior). In footage captured by the fixed video capture unit 100 in the opposite direction, the vehicle's movement direction is opposite to the actual traffic direction, requiring additional algorithm correction (such as mirror flipping), increasing computational complexity and latency, and potentially introducing errors. If the three-dimensional orientation of the retrieved fixed video capture unit 100 is chaotic, traffic management personnel need to spend time identifying the image orientation, delaying decision-making.

[0133] Therefore, the directional consistency of the fixed video acquisition unit 100 is as follows: the coverage direction of the fixed video acquisition unit 100 is consistent with the traffic direction of the highway (assuming the traffic direction is the positive direction, and the directional parameter d). i ∈{+1,−1}, priority d i =+1). d i This indicates the coverage direction of the fixed video acquisition unit 100 (+1 indicates that it is consistent with the direction of passage).

[0134] Communication delay ascending order: .

[0135] This represents the communication delay of the i-th fixed video acquisition unit 100. This represents the communication delay of the j-th fixed video acquisition unit 100.

[0136] This invention calculates the overlap between the coverage area of ​​the fixed video acquisition unit 100 and the accident calibration area, and prioritizes fixed video acquisition units 100 with consistent direction and low latency to directly cover the traffic flow in the direction of travel. This allows for clear capture of key information such as vehicle front features, license plates, and driver actions, improving the accuracy of accident analysis. The images from fixed video acquisition units 100 with consistent direction do not require direction correction, reducing image processing complexity, minimizing computational resource consumption, and further improving real-time performance (especially under low-latency requirements). The images from the forward-facing fixed video acquisition unit 100 are consistent with the traffic flow direction, facilitating a direct understanding of the spatial relationships at the accident scene (e.g., the position of the accident vehicle in the lane, and the situation of vehicles approaching from behind), accelerating emergency response. Using both forward and reverse-facing fixed video acquisition units 100 simultaneously may lead to contradictory descriptions of the same event (e.g., conflicts between vehicle position coordinates and direction markings); unifying the direction eliminates logical ambiguity.

[0137] Sort by overlap only, which might select a high-latency fixed video capture unit 100, potentially causing image transmission delays; sort by delay only, which might select a fixed video capture unit 100 in the opposite direction, resulting in invalid images. This invention first filters by direction consistency and then sorts by delay, ensuring correct image orientation and timely transmission.

[0138] When there are no directional constraints, the system may need to simultaneously retrieve multiple fixed video acquisition units 100 with inconsistent orientations, consuming bandwidth and increasing the data processing burden. The directional consistency screening of this invention directly excludes low-value fixed video acquisition units 100, reducing redundant data transmission.

[0139] S7: The intelligent analysis module 320 selects the optimal fixed video acquisition unit 100 to retrieve the audio and video information of the accident.

[0140] Therefore, the optimal set of fixed video acquisition units 100 determined by the intelligent analysis module 320 is as follows: .

[0141] This refers to the fixed video acquisition unit 100. Represents a set.

[0142] S8: The intelligent analysis module 320 sends the coordinates and / or number of the selected fixed video acquisition unit 100 to the data processing module 310. The data processing module 310 determines the time range of the accident based on the accident triggering information and sends the time range information to the selected fixed video acquisition unit 100, so that the selected fixed video acquisition unit 100 can broadcast audio and video information within a preset time before the accident to the user interface 700 based on the time range information.

[0143] Preferably, the preset time is 10 minutes. That is, inspectors can review the data from 10 minutes before the accident through the user interface. The 10-minute data review is not only the "golden time window" for accident analysis, but also the key to improving emergency response efficiency, optimizing traffic management strategies, and strengthening the legal evidence chain. It has irreplaceable value for the daily operation and long-term planning of highway management.

[0144] Highway management needs to process massive amounts of monitoring data, and the special significance of "retrieving data from 10 minutes before an accident" lies in the following aspects:

[0145] (1) The mobile video acquisition unit 200 may miss the moment of the accident due to a delay in arriving at the scene (e.g., more than 10 minutes). The retrospective data from the fixed video acquisition unit 100 can fill this gap and form a complete "full-cycle" record of the accident. By retrospectively recording the data, key parameters such as traffic flow, vehicle speed changes, and vehicle spacing before the accident can be restored, which can help determine the responsibility for the accident (e.g., speeding, illegal lane changing). Combined with the continuous recording from the fixed video acquisition unit 100, it can be determined whether the accident was caused by driver error or sudden road conditions.

[0146] (2) By analyzing traffic flow changes before an accident, the central control unit 300 can also optimize traffic light timing or lane allocation. Traffic flow data before an accident can help predict the scope of the accident's impact and dispatch rescue resources (such as ambulances and breakdown vehicles) in advance. If retrospective data shows a sudden drop in traffic flow before an accident, it may indicate severe congestion or the risk of secondary accidents. The recordings from the fixed video acquisition unit 100 can serve as "third-party evidence" independent of the mobile video acquisition unit 200, enhancing the legal validity of accident handling.

[0147] (3) Reviewing the events 10 minutes prior to the incident can reduce legal disputes. Mobile video capture unit 200 may be unable to provide complete evidence due to limited viewpoint or damage, while the recording of fixed video capture unit 100 can make up for this deficiency.

[0148] (4) In addition, the long-term accumulated retrospective data can identify the high-incidence periods, road sections or weather conditions of accidents, promote preventive management (such as adding warning signs and adjusting speed limits); it is also helpful to analyze the commonalities before multiple accidents (such as excessive speed in rainy weather) and formulate targeted measures.

[0149] (5) The present invention can also use the fixed video acquisition unit 100 to store historical video recordings by pile position, which can quickly locate accident-related data and avoid the inefficiency of full data scanning; the video recordings on the pile position can be directly retrieved by the pile position mark, instead of traversing the data of all fixed video acquisition units 100, thereby improving the retrieval efficiency.

[0150] S9: End.

[0151] In this invention, the mobile video acquisition unit 200 (such as a patrol car or drone 600) can quickly respond to the accident scene, while the fixed video acquisition unit 100 provides continuous monitoring data before the accident, avoiding the loss of key images due to the delay of the mobile video acquisition unit 200, and realizing the complementarity and collaborative work between the coverage area of ​​the fixed video acquisition unit 100 and the mobile video acquisition unit 200.

[0152] This invention matches the fixed video acquisition unit 100 using rule matching (such as overlapping pile positions) rather than complex image feature recognition (such as bridges and tunnels), thus reducing the demand for computing power.

[0153] As described above, this invention improves the real-time performance of the response. The rule matching time of the central control unit 300 of this invention is only 50ms, which can quickly locate the accident location and shorten the emergency response time. This invention also reduces the hardware requirements, eliminating the need for high-performance GPUs or dedicated AI chips; ordinary servers can support it.

[0154] Mobile video acquisition units 200 may miss the exact moment of an accident due to delays, while fixed video acquisition units 100 provide continuous monitoring data, preventing information omissions. This invention utilizes mobile video acquisition units 200 (such as patrol vehicles) to quickly detect accidents and directly retrieves historical data (such as recordings from 10 minutes prior to the accident) from fixed video acquisition units 100 through pile location rule matching, achieving data integrity. Furthermore, this invention utilizes mobile video acquisition units 200 to provide real-time trigger signals and fixed video acquisition units 100 to provide historical evidence, forming a "trigger-backtracking" closed loop.

[0155] Compared with the method of training intelligent algorithms to determine the fixed video acquisition unit 100, the rule matching logic of the central control unit 300 of the present invention can quickly adapt to the addition of a fixed video acquisition unit 100 or pile location markers without retraining the model.

[0156] Preferably, for similar accident calibration intervals, the central control unit 300 of the present invention can also verify and support the pile position adjustment by using the historical data before the accident stored in the database 400.

[0157] The numbering of highway markers in a certain area is being adjusted, and inconsistencies in marker numbers may occur between sections during construction.

[0158] Before the pile position is adjusted, the central control unit 300 can analyze the traffic characteristics (such as accident hotspots and congestion hotspots) of the original pile position numbered road segment by reviewing historical accident data (such as traffic flow, speed and accident frequency in the 10 minutes before the accident).

[0159] For example, if an old marker post has a high accident rate, the design of the new road section can be optimized by combining historical data (such as adding warning signs).

[0160] After the pile positions are adjusted, the central control unit 300 can review the accident data of the road sections corresponding to the old and new pile positions to assess whether the adjustment has effectively reduced the accident rate or improved traffic efficiency.

[0161] For example, the central control unit 300 compares the accident frequency and traffic flow changes at the same physical location before and after the adjustment (such as the original GXX KXX and the new GYY KYY) to verify the rationality of the new numbering.

[0162] Adjustments to pile location numbers may lead to inconsistencies between historical data and current pile location identifiers (e.g., old pile location GXX KXX+XX corresponds to new pile location GYY KYY+YY).

[0163] Preferably, the database 400 stores a pile location mapping table. The pile location mapping table binds the geographical coordinates of old pile locations to new pile locations, ensuring that old pile locations in historical data can be converted to new numbers. The central control unit 300 maintains historical versions of pile location numbers and supports queries by time range (e.g., "old numbers were used before a certain year and month, and new numbers were used afterward").

[0164] Preferably, in response to the possibility of inconsistencies in the numbering of road sections before and after construction, when tracing back the data, if there is a conflict in the pile number, the central control unit 300 can cross-verify by combining place names or GPS coordinates to ensure the accuracy of the accident location marking.

[0165] For example, if an accident occurs in a construction section (old pile location GXX KXX+XX, new pile location GYY KYY+YY), the system needs to mark "original number: GXX KXX+XX, current number: GYY KYY+YY" and associate it with the corresponding historical data.

[0166] Adjusting the location of traffic cones may redefine the monitoring range of a road segment (such as the coverage area of ​​a fixed video acquisition unit 100), and the retrospective data needs to be adapted to the newly defined range. To address this, after the traffic cone adjustment, the central control unit 300 re-enters the coverage area corresponding to the new traffic cone (e.g., updating the coverage area of ​​the fixed video acquisition unit 100 for the old traffic cone GXX KXX to the new GYY KYY). For example, if it is necessary to retrieve accident data from a certain year and month (old traffic cone), the central control unit 300 converts it to a new number using a mapping table to ensure matching with the current coverage area of ​​the fixed video acquisition unit 100.

[0167] For example: if the accident calibration range is from the old pile position KXX+500 to KXX+700, the central control unit 300 will automatically switch to the new pile position GYY KXX+300 to GYY KYY+500 and match the newly numbered fixed video acquisition unit 100.

[0168] For example, adjusting the location of traffic markers may lead to changes in the function of a road segment (such as changing a truck-only lane to a mixed-use lane), and retrospective data can help analyze the changes in traffic characteristics after the adjustment.

[0169] For example, the central control unit 300 compares the traffic flow composition at the same physical location before and after the adjustment (such as changes in the proportion of trucks), providing a basis for subsequent management.

[0170] Example 2

[0171] This embodiment is a further improvement on embodiment 1, and repeated content will not be described again.

[0172] The central control unit 300 may also include a command and dispatch module 330, such as Figure 2As shown. The command and dispatch module 330 is directly connected to the geographic information system display terminal via a fiber optic transceiver, synchronizes the spatial correlation topology map in real time, and sends the command and dispatch plan to the user interface 700 of the mobile terminal 800 via an encrypted wireless local area network.

[0173] Preferably, the command and dispatch module 330 integrates a geographic information system display terminal and a task allocation engine, and is connected to the intelligent analysis module 320 via a fiber optic channel. It receives the accident calibration interval, the real-time video stream from the matched fixed video acquisition unit 100, and coordinate mapping relationships to establish spatial correlation data between the accident site and surrounding rescue stations. The command and dispatch module 330 establishes a wireless link with external rescue vehicle terminals via a cellular mobile communication base station, generates a command and dispatch plan based on the task allocation mechanism and spatial correlation data, and sends navigation command sets to the mobile terminal 800 in real time.

[0174] Preferably, the original audio-visual information is distributed to the data processing module 310 via the communication module 500 for parsing, keyframe extraction, and accident calibration interval calculation. The generated calibration interval data and the matching information from the fixed video acquisition unit 100 are sent to the command and dispatch module 330 via a high-speed data bus. The command and dispatch module 330 integrates the real-time video stream, GPS coordinate mapping relationship, and task allocation parameters to generate a dispatch instruction containing the optimal path, which is then sent to the rescue vehicle via a cellular communication base station. Simultaneously, it receives on-site update data transmitted back from the mobile video acquisition unit 200, achieving closed-loop control.

[0175] Meanwhile, the database 400 maintains real-time communication with the geographic information system of the command and dispatch module 330 via a dedicated data synchronization interface. When the coordinates of the pile location change, the database 400 immediately triggers an update operation on the electronic map layer of the geographic information system. With this setup, patrol personnel can obtain the pile location markers of surrounding fixed video acquisition units 100 from the database 400 via the mobile video acquisition unit 200 after arriving at the accident scene, thus narrowing down the selection range of fixed video acquisition units 100. The data processing module 310 stores the pile location markers and the arrival times of rescue units in the database 400 in descending order of speed to provide a priority reference for accident handling. Preferably, the pile location markers are matched with the coordinates of the service area's entrances and exits to ensure rapid acquisition of relevant information during accident handling.

[0176] Specifically, the data processing module 310 associates the collected pile location data with the corresponding rescue unit arrival time data and stores them in the database 400. The data processing module 310 creates a priority index table with time sorting attributes within the database 400. The data processing module 310 uses a time-dimensional algorithm to sort and calculate the response speed of rescue units, writing the calculation results into the priority reference field of the database 400 in descending order, forming a timestamped accident handling priority queue. During data storage, the coordinates of the pile location and the service area entrance / exit coordinates are spatially matched using a geographic coordinate system conversion module. A dynamic conversion algorithm between the WGS-84 coordinate system and the local plane coordinate system is used to achieve precise coordinate correspondence, and a bidirectional association index table is established within the database 400. When the accident handling process is initiated, the intelligent analysis module 320 synchronously retrieves the priority queue data and coordinate matching data through a joint query mechanism, ensuring that the pile location information of the entrance / exit coordinates of the associated service area and the optimal rescue path time parameters can be quickly obtained during the accident handling decision-making process.

[0177] For example, create a stake marker table in database 400 to record the unique identifier, geographic coordinates (latitude and longitude in WGS-84 coordinate system) of each highway stake, as well as the corresponding nearest service area entrance ID.

[0178] Database 400 contains a rescue response timeline, recording the average response time (in minutes) and data update time for each rescue unit (such as fire brigades and ambulances). For example, a rescue unit with a unique ID of XF-001 has an average response time of 12 minutes, and its last data update time is T1. Database 400 uses an algorithm to calculate the response speed ranking of rescue units and stores it in a priority list. The rescue response timeline includes the rescue unit ID, response speed ranking (the smaller the value, the faster), and a timestamp (recording the ranking update time). For example, it associates the rescue unit ID (e.g., XF-001), response speed ranking (e.g., 1 indicates the fastest), and ranking update time. Database 400 uses a geospatial library (e.g., GDAL) to perform coordinate system transformation, converting the latitude and longitude coordinates of the location points to a local plane coordinate system (e.g., a province's custom coordinate system).

[0179] When an accident is triggered, the central control unit 300 retrieves the priority queue and obtains the fastest-responding rescue unit (e.g., fire brigade A, ranked 1st) from the rescue response timeline. Based on the accident location markers, the central control unit 300 obtains the planar coordinates (X, Y) of the corresponding service area entrance from the priority queue list. Combining this with the rescue unit's current location, it calculates the estimated arrival time (e.g., 15 minutes) for the optimal route.

[0180] The decision reports generated by the central control unit 300 include the following information:

[0181] Accident pile location: KXX+XX.

[0182] Optimal rescue unit: Fire Brigade A (fastest response time).

[0183] Estimated arrival time: 15 minutes.

[0184] Route planning: Enter the G2 expressway through service area entrance B and proceed directly to the accident pile location.

[0185] Coordinate information: Pile location coordinates: latitude A, longitude B.

[0186] Service area entrance coordinates: X, Y.

[0187] After each rescue mission is completed, the system automatically updates the average response time of the rescue response timeline and recalculates the priority team list through triggers to ensure that priorities are effective in real time.

[0188] It should be noted that the specific embodiments described above are exemplary. Those skilled in the art can devise various solutions inspired by the disclosure of this invention, and these solutions all fall within the scope of this invention and its protection. Those skilled in the art should understand that this specification and its accompanying drawings are illustrative and not intended to limit the scope of the claims. The scope of protection of this invention is defined by the claims and their equivalents. This specification contains multiple inventive concepts; phrases such as "preferredly" or "according to a preferred embodiment" indicate that the corresponding paragraph discloses an independent concept. The applicant reserves the right to file divisional applications based on each inventive concept.

Claims

1. A highway rescue command and dispatch decision system based on multi-source data, characterized in that, The system comprises: a fixed video acquisition unit for acquiring video and audio information of a corresponding section of the expressway; a mobile video acquisition unit comprising two types of devices, carried by a patrol personnel or used by a drone, for quickly discovering an accident and confirming the video and audio information of the accident section of the expressway; a central control unit receiving the accident trigger information from the mobile video acquisition unit, extracting key frames containing accident images from the video and audio information sent by the mobile video acquisition unit, and identifying the stake markers, taking the position of the stake marker closest to the time stamp of the accident trigger time as the starting point, and taking the position of the nearest downstream stake marker along the direction of the expressway as the end point to determine the accident demarcation interval; based on the coordinates of the starting point and the end point of the accident demarcation interval, at least one fixed video acquisition unit overlapping the accident demarcation interval is screened, the optimal fixed video acquisition unit is selected, and picture relay is triggered, so that the user interface of the mobile terminal can backtrack the video and audio information collected by the fixed video acquisition unit within a preset time before the accident; Wherein, the accident trigger time is t acc , the front and back two frames of time stamp extracted by the mobile video acquisition unit are t1 and t2, and t1<t acc ≤t2, the set of stake markers identified in the front frame t1 and the back frame t2 are respectively: ； ； n represents the total number of stake markers identified in the front frame t1, m represents the total number of stake markers identified in the back frame t2, P1 represents the set of stake markers detected in the front frame by the image recognition algorithm, and P2 represents the set of stake markers detected in the back frame; the starting point stake is: ； a timestamp representing a frame corresponding to the stake position p, a start point stake position; The positive direction is along the highway traffic direction, and the terminal point pile position K end is: ； represents a set of peg locations in a frame corresponding to time t, represents an end point peg; The dynamic matching mode of the intelligent analysis module of the central control unit to the fixed video acquisition unit is: after screening the fixed video acquisition units overlapping the accident demarcation interval, the coordinates and orientation data of the fixed video acquisition units are called from the database, the fixed video acquisition units with consistent or similar orientation data are preferentially selected, the coincidence degree of picture acquisition between the selected fixed video acquisition units is calculated based on the coordinates and orientation data of the fixed video acquisition units, the fixed video acquisition units are prioritized based on the coincidence degree, and the optimal fixed video acquisition unit is selected to call the video and audio information of the accident, The coincidence degree of the coverage interval of the fixed video acquisition unit and the accident demarcation interval is defined as: ， wherein, represents the length of the overlap interval of the i-th fixed video capture unit and the accident demarcated interval; represents the smaller value between the accident end point and the end point of the fixed video capture unit, and represents the right boundary of the overlap interval; represents the larger value between the accident start point and the start point of the fixed video capture unit, and represents the left boundary of the overlap interval; represents the start stake coordinate covered by the i-th fixed video capture unit; represents the end stake coordinate covered by the i-th fixed video capture unit; the right boundary minus the left boundary, if the result is positive, it means there is an overlap; if it is negative or zero, it means there is no overlap.

2. The system of claim 1, wherein, The central control unit comprises: a data processing module receiving the accident trigger information sent by the mobile video acquisition unit, extracting key frames containing accident images from the video and audio information sent by the mobile video acquisition unit, and identifying the stake markers; if no stake marker is identified, the data processing module triggers an alarm and returns a video and audio information reacquisition instruction to the mobile video acquisition unit, if a stake marker is identified, the data processing module determines the starting point and the end point of the accident demarcation interval, and sends the starting point and the end point of the accident demarcation interval to the intelligent analysis module to select the fixed video acquisition unit.

3. The system of claim 2, wherein, The system further comprises a database in communication connection with the central control unit, The coordinates of the stake markers and the coordinates and orientation data of the fixed video acquisition units are stored in the database to allow the intelligent analysis module in the central control unit to quickly find and confirm the position of the stake markers.

4. The system of claim 3, wherein, The intelligent analysis module is configured to: After receiving the stake markers corresponding to the start point and the end point of the accident demarcated interval, the database is sent the stake markers and extraction instructions to receive the coordinates of the start point and the end point from the database, so as to screen the fixed video acquisition units whose coverage intervals overlap with the accident demarcated interval.

5. The system of claim 3, wherein, The data processing module is further configured to: identify stake images from the video and audio information collected by the mobile video acquisition unit, and extract a plurality of stake marker positions in the video and audio information based on the stake images.

6. The system of claim 5, wherein, The data processing module is further configured to: extract key frames containing accident images from the video and audio information collected by the mobile video acquisition unit, and determine the accident position by spatial interpolation.

7. The system of any one of claims 2-6, wherein, The intelligent analysis module sends the coordinates and / or number of the selected fixed video acquisition unit to the data processing module, The data processing module determines the time range of the accident based on the accident trigger information, and sends the time range information to the selected fixed video acquisition unit, so that the selected fixed video acquisition unit rebroadcasts the video and audio information within a preset time before the accident to the user interface based on the time range information.

8. The system of claim 7, wherein, When the mobile video acquisition unit is used by a drone, the drone is in communication connection with the central control unit, and the drone is used to fly to the accident position and transmit the video and audio information of the accident scene to the central control unit in real time.

9. The system of claim 7, wherein, The system further comprises a user interface, which is in communication connection with the central control unit through a mobile terminal, The user interface provides an intuitive graphical operation interface on the mobile terminal to display the video and audio information, so as to facilitate the command personnel to check the command and dispatch scheme and perform interactive operation at any time and any place.

Images