An automatic road flood control method and system based on edge collaboration

Abstract

This application discloses an automatic road flood control method and system based on edge collaboration. The method includes: generating flood depth and corresponding flood depth recognition credibility; determining a real-time control trigger threshold and a corresponding initial control level; realizing the interaction of flood depth, flood depth recognition credibility, and initial control level with the perception information of collaborative edge computing nodes, and performing consistency verification on the multi-node perception information obtained through interaction to generate collaborative credibility and a corrected flood depth; comparing the corrected flood depth with the real-time control trigger threshold, determining the control level based on the comparison result, and generating control commands; and sending the control commands to the control execution device to implement road control. Using the solution of this application, the entire process from flood monitoring and level determination to equipment execution is automated and precise, greatly improving the efficiency and reliability of road traffic safety management under severe weather conditions.

Description

Technical Field

[0001] This application generally relates to the fields of intelligent traffic management and IoT edge computing technology. More specifically, this application relates to a method and system for automatic road flood control based on edge collaboration. Background Technology

[0002] As urbanization continues to accelerate, coupled with the normalization of extreme rainfall due to global warming, key traffic nodes such as urban roads, culverts, and tunnels are frequently flooded or even suffer from waterlogging disasters, directly threatening the lives of drivers, passengers, and pedestrians, and seriously affecting the normal operation of urban traffic.

[0003] Existing road flooding monitoring and control solutions typically employ a centralized architecture where front-end cameras or water level sensors collect on-site data and upload it to a central platform. The platform then processes the data or provides manual assessment before issuing control commands. However, this existing technology suffers from several major drawbacks: First, it heavily relies on the central platform and communication links. In the event of extreme weather causing network outages or platform congestion / failures, on-site nodes cannot autonomously implement control measures, easily missing the optimal window for emergency response. Second, single-point judgment has weak anti-interference capabilities. The single-point identification of front-end sensors is susceptible to adverse environmental factors such as raindrop obstruction, nighttime reflections, and blurred images, leading to misjudgments and resulting in false or missed control measures. Third, it lacks a multi-node collaborative linkage mechanism. Multiple intersections and nodes are isolated from each other, making it difficult to form an effective coordinated control strategy for regional flooding disasters. Fourth, the overall system's resilience is insufficient. Existing centralized solutions do not fully utilize the decentralized and fault-tolerant advantages of distributed systems, lacking fault takeover capabilities between edge nodes, resulting in poor system disaster recovery.

[0004] In view of this, there is an urgent need to provide an automatic road flood control solution based on edge collaboration to solve the above problems. This solution should be able to operate autonomously on the edge side and have multi-node collaboration and high fault tolerance. Summary of the Invention

[0005] In order to at least address one or more of the technical problems mentioned above, this application proposes an automatic road flood control scheme based on edge collaboration in several aspects.

[0006] In a first aspect, this application provides an automatic road flood control method based on edge collaboration, comprising: edge computing nodes acquiring video data of a road monitoring area in real time, generating flood depth and corresponding flood depth recognition credibility based on the video data; acquiring meteorological data corresponding to the road monitoring area, determining a real-time control trigger threshold and a corresponding initial control level based on the meteorological data; realizing the interaction of the flood depth, the flood depth recognition credibility, and the initial control level with the perception information of the collaborative edge computing nodes based on a distributed communication mechanism, and performing consistency verification on the multi-node perception information acquired through interaction to generate collaborative credibility and a corrected flood depth; comparing the corrected flood depth with the real-time control trigger threshold, determining the control level based on the comparison result and generating a control command; and sending the control command to a control execution device to perform road control.

[0007] In a second aspect, this application provides an edge-cooperative automatic road flood control system, employing the edge-cooperative automatic road flood control method as described in any embodiment of the first aspect. The system includes: a video data processing module, used to acquire video data of a road monitoring area in real time using edge computing nodes, and generate flood depth and corresponding flood depth recognition credibility based on the video data; a meteorological data processing module, used to acquire meteorological data corresponding to the road monitoring area, and determine a real-time control trigger threshold and a corresponding initial control level based on the meteorological data; a multi-node collaborative verification module, used to realize the interaction of the flood depth, the flood depth recognition credibility, and the initial control level with the perception information of the collaborative edge computing nodes based on a distributed communication mechanism, and to perform consistency verification on the multi-node perception information acquired through interaction, so as to generate collaborative credibility and a corrected flood depth; a control command generation module, used to compare the corrected flood depth with the real-time control trigger threshold, determine the control level based on the comparison result, and generate a control command; and a control command issuing module, used to send the control command to a control execution device for road control.

[0008] The automatic road flood control scheme based on edge collaboration, as described above, achieves dynamic adaptive adjustment of the road closure trigger threshold by fusing video images and meteorological multimodal data. This makes closure decisions more consistent with real-time weather conditions, demonstrating greater rationality and foresight. Simultaneously, an edge collaboration mechanism is introduced. Through distributed interaction and consistency verification of sensing information among multiple nodes, it effectively overcomes misjudgments and omissions caused by limited field of view or equipment failure of a single node, calculating highly accurate corrected flood depth and collaborative reliability. This process not only significantly improves the accuracy of flood sensing and the overall robustness of the system but also achieves full-process automation and precision from flood monitoring and level determination to equipment execution, greatly improving the efficiency and reliability of road traffic safety management under severe weather conditions.

[0009] Furthermore, in some embodiments, by establishing a precise mapping relationship between image pixels and actual physical height, non-contact, high-precision quantitative measurement of water depth is achieved, effectively overcoming the shortcomings of traditional physical water level sensors that are easily blocked by sludge or damaged by water immersion. Simultaneously, when evaluating the reliability of the identification, the spatial clarity of a single-frame image and the temporal dynamic stability of continuous video frames are comprehensively considered, and environmental interference factors such as raindrop occlusion and nighttime reflections are specifically introduced. This multi-dimensional weighted fusion mechanism of spatiotemporal characteristics and environmental interference can effectively identify and filter out visual noise caused by vehicle splashes, sudden changes in lighting, and severe weather, scientifically and objectively quantifying the reliability of the current visual perception results, greatly enhancing the system's anti-interference capability and data confidence in complex and extreme visual scenarios.

[0010] Furthermore, in some embodiments, standardized perception data encapsulation and distributed communication mechanisms break down information silos between individual computing nodes, enabling secure and efficient interoperability of multi-dimensional water accumulation status information between adjacent nodes. Simultaneously, a multi-level, rigorous verification model is constructed, encompassing numerical deviation comparison, level consistency judgment, and collaborative credibility threshold verification. This multi-node cross-validation mechanism accurately identifies and intercepts abnormal conflicting data caused by single-point device failures or local environmental interference, and promptly issues alarms to prevent misjudgment risks when verification fails. Finally, the corrected perception result is output only when multiple nodes reach a high-confidence consensus, thereby significantly improving the system's fault tolerance and resilience in complex road network environments, ensuring extremely high accuracy and security in the final road closure decision.

[0011] Furthermore, in some embodiments, a duration-based delay confirmation mechanism is innovatively introduced in the lockdown trigger determination stage. This effectively filters out interference data caused by water waves stirred up by vehicles wading through water or occasional short-term water level fluctuations, avoiding false alarms or frequent, repeated triggering of lockdown actions. This significantly enhances the rigor of lockdown decisions and the system's anti-vibration capabilities. Simultaneously, after confirming that the triggering conditions are met, the system comprehensively considers the corrected water depth, the initial lockdown level, and the reliability of multi-node collaboration to ultimately determine the lockdown level. This multi-dimensional integrated decision-making mechanism ensures that the generated control commands can more accurately match the actual safety risk level of the current road, achieving refined grading and highly reliable execution of emergency control measures. Attached Figure Description

[0012] The above and other objects, features, and advantages of exemplary embodiments of this application will become readily understood by reading the following detailed description with reference to the accompanying drawings. In the drawings, several embodiments of this application are illustrated by way of example and not limitation, and the same or corresponding reference numerals denote the same or corresponding parts, wherein: Figure 1 An exemplary flowchart of an edge-cooperative automatic road flood control method according to an embodiment of this application is shown; Figure 2 An exemplary flowchart illustrating the generation of water depth based on video data according to an embodiment of this application is shown; Figure 3 An exemplary flowchart illustrating an embodiment of this application is provided for generating a confidence level for water depth recognition based on video data. Figure 4 An exemplary flowchart illustrating an embodiment of this application is shown, which determines the real-time lockdown trigger threshold and the corresponding initial lockdown level based on meteorological data. Figure 5 An exemplary flowchart illustrating the interaction of perception information between water depth, water depth recognition reliability, initial containment level, and collaborative edge computing nodes according to an embodiment of this application is shown. Figure 6 An exemplary flowchart illustrating the generation of collaborative credibility and the corrected water depth according to an embodiment of this application is shown; Figure 7 An exemplary flowchart illustrating how this application determines the lockdown level and generates control commands based on comparison results is shown in an embodiment of this application. Figure 8 An exemplary flowchart illustrating how control commands are sent to a road closure execution device according to an embodiment of this application is shown; Figure 9 An exemplary flowchart of the method for lifting road closures based on real-time acquired corrected water depth, according to an embodiment of this application, is shown. Figure 10 A schematic diagram of the water depth curve according to an embodiment of this application is shown; Figure 11 An exemplary structural block diagram of an edge-cooperative automatic road flood control system according to an embodiment of this application is shown. Detailed Implementation

[0013] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0014] It should be understood that the terms "comprising" and "including" used in the specification and claims of this application indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0015] It should also be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application. As used in this specification and claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this specification and claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations.

[0016] Figure 1 An exemplary flowchart of an edge-cooperative automatic road flood control method 100 according to an embodiment of this application is shown.

[0017] like Figure 1 As shown, in step S110, the edge computing node acquires video data of the road monitoring area in real time, and generates water depth and corresponding water depth recognition credibility based on the video data.

[0018] Specifically, road monitoring video data is acquired by video acquisition devices (such as high-definition cameras) set up in the road monitoring area, and the video data is sent to the corresponding edge computing nodes in real time via wired or wireless networks.

[0019] In the embodiments of this application, the specific process involved in generating water depth based on video data can be found in [reference needed]. Figure 2 .

[0020] Figure 2An exemplary flowchart illustrating the generation of water depth based on video data according to an embodiment of this application is shown.

[0021] like Figure 2 As shown, in step S210, the video data is preprocessed to obtain an enhanced image frame. In step S220, water surface features are extracted from the enhanced image frame to identify areas of road flooding. In step S230, a mapping relationship between image pixel positions and actual physical heights is established based on the pixel coordinates of fixed water level markers or preset reference objects in the enhanced image frame. In step S240, the water depth is obtained based on the pixel positions of the flooded areas in the enhanced image frame, combined with the mapping relationship between image pixel positions and actual physical heights.

[0022] In the embodiments of this application, the preprocessing operations performed by the edge computing node on the received video data include, but are not limited to: frame extraction, image denoising, and brightness or contrast enhancement. The enhanced image frames obtained through these preprocessing operations can effectively improve the visual recognizability of video data under complex weather conditions (such as heavy rain or fog) or poor lighting conditions, thereby providing high-quality and stable data input for subsequent road flooding identification.

[0023] In the embodiments of this application, the identification of water accumulation areas is achieved by matching and judging specific water surface features in the enhanced image frame. The water surface features include, but are not limited to: water surface reflection features, water surface texture smoothness features, or water area boundary continuity features.

[0024] In the embodiments of this application, after identifying the waterlogged area, the system constructs a transformation matrix between 2D pixels and 3D physical space using calibrated fixed water level markers on the road or preset reference objects of known height in the environment (such as curbs, road markings, utility poles, etc.). This allows the pixel coordinates of the waterlogged area in the image to be converted into its actual height in the physical world, thus calculating the precise water depth.

[0025] In the embodiments of this application, the specific process of generating the credibility of water depth identification based on video data can be found in [reference needed]. Figure 3 .

[0026] Figure 3 An exemplary flowchart illustrating an embodiment of this application is shown, illustrating the generation of credibility for water depth recognition based on video data.

[0027] like Figure 3As shown, in step S310, spatial evaluation features are extracted, including the image clarity of the video data and the sharpness of the road flooding area boundary in a single frame image. In step S320, temporal evaluation features are extracted from the historical recognition results of flooding depth, including the displacement stability of the road flooding area boundary in consecutive video frames and the variation amplitude of flooding depth within a preset time period. In step S330, environmental interference features are extracted, including the proportion of raindrop obstruction and the intensity of nighttime reflective noise. In step S340, the spatial evaluation features, temporal evaluation features, and environmental interference features are weighted and fused to generate the confidence level of flooding depth recognition.

[0028] Based on steps S310-S340, a multi-dimensional comprehensive evaluation is conducted based on the image quality of the current video data and the dynamic stability of the water accumulation recognition results. The system extracts features from three dimensions: spatial domain (single-frame image quality), temporal domain (temporal stability of consecutive video frames), and environmental interference factors (such as the proportion of raindrops obscuring the camera lens, road surface glare at night, or noise from vehicle headlights). A weighted fusion algorithm is then used to derive the final recognition credibility score. This credibility score intuitively characterizes the extent to which the current machine vision perception results accurately reflect the actual water accumulation state of the road, effectively preventing the system from outputting incorrect water depths due to occasional image interference (such as water splashing from passing vehicles or sudden bright light), thus avoiding subsequent incorrect road closure decisions.

[0029] After completing step S110, in step S120, meteorological data corresponding to the road monitoring area is obtained, and the real-time lockdown trigger threshold and the corresponding initial lockdown level are determined based on the meteorological data.

[0030] In the embodiments of this application, meteorological data directly related to the road monitoring area is obtained by connecting to the meteorological department's data interface or deploying meteorological sensors on-site. This meteorological data includes, but is not limited to, real-time rainfall intensity, future rainfall trend forecasts, or officially released meteorological warning levels. By using multi-dimensional meteorological data as important auxiliary parameters for determining road closures, the lag of relying solely on a single water depth is overcome.

[0031] In the embodiments of this application, the specific process involved in determining the real-time lockdown trigger threshold and the corresponding initial lockdown level based on meteorological data can be found in [reference needed]. Figure 4 .

[0032] Figure 4 An exemplary flowchart illustrating an embodiment of this application is shown, which determines the real-time lockdown trigger threshold and the corresponding initial lockdown level based on meteorological data.

[0033] like Figure 4As shown, in step S410, real-time rainfall intensity and meteorological warning characteristics are extracted from meteorological data, and a meteorological risk level characterizing the degree of precipitation impact is calculated. In step S420, the preset road closure benchmark threshold is dynamically adjusted downward based on the meteorological risk level to generate a real-time closure trigger threshold that matches the current meteorological data. In step S430, the water depth is compared with the real-time closure trigger threshold, and an initial closure level is obtained by matching the water depth to the real-time closure trigger threshold in a preset level mapping table.

[0034] In the embodiments of this application, for the extraction of real-time rainfall intensity features, the system not only records the current instantaneous rainfall, but also calculates the cumulative rainfall within a preset time period (such as the past 5 minutes, 15 minutes, 1 hour) using a sliding window algorithm, and analyzes the rainfall change rate (rainfall trend) accordingly. For example, if the current instantaneous rainfall increases sharply, and the cumulative rainfall within 15 minutes reaches the level of a rainstorm, then this feature value will receive a higher weight score.

[0035] In the embodiments of this application, the extraction of meteorological warning features automatically analyzes official warning signals (such as blue, yellow, orange, and red rainstorm warnings) issued by meteorological departments for the grid area. In addition to the warning level, the timeliness characteristics of the warning are also extracted, including the duration of the warning after its issuance and whether the warning is in an escalation state. This qualitative meteorological warning information is digitized into corresponding risk factors to characterize the macro-level precipitation risk background.

[0036] In the embodiments of this application, the system employs a multi-factor weighted evaluation model in calculating the meteorological risk level. This model uses extracted rainfall intensity (quantitative indicator), rainfall trend (dynamic indicator), and meteorological warning level (qualitative administrative indicator) as input variables. These input variables are weighted and summed to obtain a standardized meteorological risk level (e.g., divided into levels 1 to 4, with higher numbers indicating higher risk). This level not only reflects the current precipitation status but also predicts the possibility of rapid water accumulation in a short period, thus providing a scientific basis for adjusting the advance warning threshold for closure in step S420. For example, even if the current road surface water depth has not yet reached the standard, if the calculated meteorological risk level is the highest level, the system will proactively tighten the closure strategy.

[0037] In the embodiments of this application, an initial physical flood control baseline threshold is preset. Under extreme weather conditions, rigidly adhering to a fixed threshold often leads to delayed control actions. Therefore, this baseline threshold is dynamically adjusted in stages based on the risk level corresponding to different meteorological data. For example, when rainfall intensity is high or the weather warning level is high, the control trigger threshold is lowered by a preset proportion. This dynamic adjustment mechanism can significantly improve the system's sensitivity, providing sufficient buffer time for emergency management and thus proactively addressing the risk of rapidly rising floodwaters.

[0038] In the embodiments of this application, the current water depth calculated based on video in step S110 is compared with the real-time closure trigger threshold generated in step S420. Based on the percentage of the current water depth to the real-time closure trigger threshold or the difference between the two, a preset level mapping table is automatically queried, thereby assigning a preliminary closure control level to the current state of the road segment, providing a basic reference benchmark for subsequent multi-node collaborative decision-making.

[0039] After completing step S120, in step S130, the perception information of water depth, water depth recognition credibility and initial control level is realized by the distributed communication mechanism and the collaborative edge computing node. The consistency of the multi-node perception information obtained by the interaction is verified to generate collaborative credibility and corrected water depth.

[0040] Step S130 utilizes the collaborative computing capabilities of the edge side and cross-validates multi-source data to eliminate perception bias caused by occlusion, abnormal lighting, or equipment failure at a single detection point.

[0041] In the embodiments of this application, the specific process involved in the interaction of perception information between the water depth, the reliability of water depth identification, the initial control level, and the collaborative edge computing nodes based on the distributed communication mechanism can be found in [reference needed]. Figure 5 .

[0042] Figure 5 An exemplary flowchart illustrating the interaction of perception information between water depth, water depth recognition reliability, initial containment level, and collaborative edge computing nodes according to an embodiment of this application is shown.

[0043] like Figure 5As shown, in step S510, the water depth, water depth recognition confidence level, and initial containment level are combined and encapsulated into a local sensing data frame with node identification. In step S520, based on a distributed communication mechanism, the local sensing data frame is sent to the collaborative edge computing nodes with a connection relationship. In step S530, the collaborative sensing data frames sent by the collaborative edge computing nodes are received synchronously, and the collaborative water depth, collaborative water depth recognition confidence level, and collaborative initial containment level corresponding to the collaborative edge computing node are parsed from them to complete the interaction of sensing information.

[0044] In the embodiments of this application, each edge computing node joins the collaborative network and maintains a heartbeat based on the distributed discovery protocol of the HarmonyOS operating system. When combining and encapsulating the water depth, the credibility of water depth identification, and the initial control level, the local perception data frame not only contains the core water status data, but also carries a timestamp and time-to-live (TTL) field to ensure the real-time nature of the data and prevent expired data from participating in subsequent decisions.

[0045] In the embodiments of this application, during the execution of steps S520 and S530, nodes share critical data through distributed soft bus technology in a publish / subscribe or point-to-point synchronization manner. This distributed mechanism allows adjacent nodes (such as monitoring devices at the four corners of an intersection) to break down information silos and form a logically unified sensing space.

[0046] In the embodiments of this application, the specific process of performing consistency verification on the multi-node perception information obtained through interaction to generate collaborative credibility and corrected water depth can be found in [reference needed]. Figure 6 .

[0047] Figure 6 An exemplary flowchart illustrating the generation of collaborative credibility and the corrected water depth according to an embodiment of this application is shown.

[0048] like Figure 6 As shown, in step S610, the absolute value of the difference between the water depth and the coordinated water depth is calculated. In step S620, it is determined whether the absolute value of the difference is less than a preset error tolerance threshold, and whether the initial lockdown level and the coordinated initial lockdown level are consistent. In response to the absolute value of the difference not being less than the preset error tolerance threshold, or the initial lockdown level and the coordinated initial lockdown level being inconsistent, in step S630, it is determined that the multi-node perception information consistency verification has failed, and a first alarm message is generated. In response to the absolute value of the difference being less than the preset error tolerance threshold, and the initial lockdown level and the coordinated initial lockdown level being consistent, in step S640, it is determined that the multi-node perception information consistency verification has initially passed, a coordinated credibility is generated based on the water depth recognition credibility and the coordinated water depth recognition credibility, and a corrected water depth is generated based on the water depth and the coordinated water depth.

[0049] Next, in step S650, it is determined whether the collaborative credibility meets the collaborative credibility threshold range. If the collaborative credibility is within the collaborative credibility threshold range, in step S660, the consistency verification of the multi-node perception information is determined to be passed, and the collaborative credibility and the corrected water depth are output. In step S670, if the collaborative credibility is not within the collaborative credibility threshold range, a second alarm message is generated.

[0050] Based on step S620, if the difference in depth detected by the two nodes is too large (exceeding the error tolerance threshold), or if there is a serious disagreement in the judgment of the current lockdown level (such as one determining "complete lockdown" and the other determining "no lockdown required"), it indicates that there is a conflict in the perception results.

[0051] In the embodiments of this application, the first alarm information includes a conflict node identifier, data divergence details, preliminary investigation of the conflict cause, and emergency response suggestions. Specifically, the conflict node identifier is used to indicate the specific edge computing node ID that caused the data divergence (e.g., the node on the east side of the intersection and the node on the west side of the intersection). The data divergence details are used to record the original water depth value reported by each node, the recognition confidence level, and the corresponding initial control level. For example, the alarm information may include: "Node A detected a depth of 20cm (Level: Level 2), Node B detected a depth of 5cm (Level: No Control), the difference exceeds the preset range." The preliminary investigation of the conflict cause is used to mark possible conflict causes based on the self-check status of each node (e.g., a camera on a certain node is blocked by a foreign object, a section of road on one side experiences localized instantaneous water accumulation due to a blocked drain, or a node experiences abnormal light interference). The emergency response suggestions include a "manual verification request" sent to the management backend.

[0052] In the embodiments of this application, a collaborative credibility is generated by weighted calculation based on the recognition credibility of two nodes, and a fusion operation is performed based on the depth values ​​of the two nodes (e.g., a weighted average is performed using the recognition credibility of the two nodes as weights for their respective depth values) to generate a corrected water depth. This process utilizes data redundancy, significantly improving the accuracy of the final output.

[0053] In the embodiments of this application, the second alarm information mainly targets the overall perception environment in adverse scenarios. Even if the data detected by multiple nodes are numerically consistent (preliminary verification passed), if the calculated collaborative credibility is still lower than a preset threshold, it indicates that the current overall perception recognition environment is extremely unreliable. The second alarm information specifically includes an environmental interference factor report, a collaborative credibility score, a system degradation strategy prompt, and maintenance suggestions. The environmental interference factor report lists in detail the environmental factors that lead to low credibility, including but not limited to: raindrops obscuring the lens beyond a preset limit, excessive noise from strong nighttime light reflections, or extremely low image edge sharpness (affected by dense fog). The collaborative credibility score outputs the final credibility score after weighted fusion, used to quantify the unreliability of the perception result. The system degradation strategy prompt indicates that the system has entered a "low-confidence perception state," at which point the system will perform degradation operations according to preset logic. For example, the system may no longer automatically execute the fully enclosed command, but instead only issue low-risk control commands such as "speed limit reminder" or "road warning," and suggests secondary verification based on meteorological data and historical drainage models. Maintenance recommendations include reminding maintenance personnel to inspect the video capture equipment in the area (e.g., cleaning water stains from lenses, adjusting nighttime lighting parameters, etc.).

[0054] After step S130 is completed, in step S140, the corrected water depth is compared with the real-time lockdown trigger threshold, and the lockdown level is determined based on the comparison result and a control command is generated.

[0055] In the embodiments of this application, the specific details regarding determining the lockdown level and generating control commands based on the comparison results can be found in [reference needed]. Figure 7 .

[0056] Figure 7 An exemplary flowchart illustrating how this application determines the lockdown level and generates control commands based on comparison results is shown in an embodiment of this application.

[0057] like Figure 7 As shown, in step S710, it is determined whether the corrected water depth exceeds the real-time closure trigger threshold, and whether the duration for which the corrected water depth exceeds the real-time closure trigger threshold reaches a preset time. In response to the correction not exceeding the real-time closure trigger threshold, or the duration for which the corrected water depth exceeds the real-time closure trigger threshold not reaching the preset time, in step S720, it is determined that the road closure trigger condition is not met. In response to the correction exceeding the real-time closure trigger threshold, and the duration for which the correction exceeds the real-time closure trigger threshold reaches the preset time, in step S730, it is determined that the road closure trigger condition is met, the closure level is determined based on the corrected water depth and the coordination reliability, and a corresponding road closure control command is generated based on the closure level.

[0058] In the embodiments of this application, step S710 embodies the technical concept of anti-shake control. In real-world road scenarios, when a vehicle rapidly passes through a flooded area, it may cause instantaneous water splashing, resulting in sensors or video recognition outputting extremely high instantaneous depth values. By setting a preset time (e.g., 30 to 60 seconds), the system requires the water depth to continuously and stably exceed a threshold before triggering subsequent actions. This effectively filters out "false water accumulation" data caused by sudden interference, avoiding frequent and erroneous opening and closing of control facilities.

[0059] In the embodiments of this application, the lockdown level is divided into four levels: L0, L1, L2, and L3. The final determination of the level is jointly determined by the water depth, meteorological risk level, and collaborative credibility. L0 (Normal) indicates that the corrected water depth is below the safe level, there is no risk of road passage, and the closure device remains open or the display screen indicates "Passage is normal". L1 (Warning) indicates that the water depth is close to the threshold, and the meteorological risk level shows an increasing trend in rainfall, or the coordination reliability has slightly decreased due to environmental factors. At this time, the system generates a warning instruction, broadcasting "Water ahead, drive with caution" on the roadside LED screen and pushing the warning information to the backend. L2 (Partial Restriction) indicates that the water depth has exceeded the first-level threshold, but has not yet reached the full closure standard. The system generates a partial closure instruction, such as allowing only large vehicles with high chassis to pass, and controlling the entry of passenger cars through intelligent gates to achieve differentiated traffic flow management. L3 (Full Closure) indicates that the water depth significantly exceeds the preset closure benchmark threshold, and the coordination reliability is extremely high, indicating a serious risk of flooding. At this time, the system generates the highest-level closure instruction, forcibly lowering the physical gate or implementing complete interception through traffic lights to achieve full road closure.

[0060] After step S140 is completed, in step S150, control commands are sent to the road closure execution device to carry out road closure.

[0061] The specific process involved in step S150 in the embodiments of this application can be found in [reference needed]. Figure 8 .

[0062] Figure 8 An exemplary flowchart illustrating how control commands are sent to a road closure execution device to perform road closure, according to an embodiment of this application, is shown.

[0063] like Figure 8As shown, in step S810, the edge computing node with master control is determined as the master control node through a preset arbitration mechanism. In step S820, it is determined whether the master control node has malfunctioned. If the master control node has not malfunctioned, in step S830, the master control node directly sends control commands to the road closure execution device to implement the road closure operation. If the master control node malfunctions, in step S840, master control is allocated to the corresponding collaborative edge computing node through the arbitration mechanism, and the collaborative edge computing node with master control directly sends control commands to the road closure execution device to implement the road closure operation.

[0064] In the embodiments of this application, in the distributed collaborative network based on the HarmonyOS operating system, multiple edge nodes within the same monitoring area are in a peer-to-peer relationship. To avoid instruction conflicts (multiple concurrent instructions causing logical chaos in the execution mechanism), the system selects a unique master node through an arbitration mechanism. The arbitration mechanism determines the edge computing node with master control based on the health status, communication status, or collaborative consistency results of each edge computing node. For example, the health status, communication status, and collaborative consistency results of each edge computing node are weighted and summed, and the node with the highest weighted sum is selected as the master node, possessing exclusive authority to issue instructions to the control execution device.

[0065] In the embodiments of this application, a distributed discovery mechanism is used to maintain heartbeat signals between nodes in real time. Collaborating nodes continuously monitor the response status of the master node. If no heartbeat packet is received within a preset timeout period, or if the master node reports a fatal hardware error (such as video stream interruption, processor overheating, etc.), it is determined that the master node has malfunctioned.

[0066] In the embodiments of this application, when the master control node does not malfunction, the collaborative edge computing node is in a hot standby state, synchronously receiving the decision stream from the master control node, but does not repeatedly issue physical control commands to ensure the singleness of the control path.

[0067] In the embodiments of this application, when the master control node malfunctions, a collaborative edge computing node with master control sends control commands directly to the road closure execution device to implement road closure operations. This process demonstrates the system's automatic takeover and self-healing capabilities. Once the original master control node fails, the collaborative edge computing node will immediately and automatically upgrade to the master control node according to arbitration priority. Since the nodes have shared key data such as water depth, credibility, and closure status in real time through a distributed soft bus (and the data carries timestamps and is within the TTL validity period), the takeover node does not need to reinitialize calculations and can continue to issue closure or release commands based on the latest perception snapshot. This seamless switching mechanism ensures that the closure control remains uninterrupted, preventing the system from becoming paralyzed when severe weather causes damage to a single device.

[0068] In the embodiments of this application, after step S150 is completed, the system does not stop monitoring, but instead enters the unblocking assessment stage, that is, the road closure is lifted based on the corrected water depth obtained in real time.

[0069] In the embodiments of this application, the specific details regarding the lifting of road closures based on the real-time acquired corrected water depth can be found in the following documents. Figure 9 .

[0070] Figure 9 An exemplary flowchart illustrating the lifting of road closures based on real-time acquired corrected water depth, according to an embodiment of this application, is shown.

[0071] like Figure 9 As shown, in step S910, the corrected water depth corresponding to the road monitoring area is monitored in real time. In step S920, it is determined whether the corrected water depth is lower than the closure release threshold and whether the duration of the corrected water depth being lower than the closure release threshold reaches the release time threshold. In response to the corrected water depth being lower than the closure release threshold and the duration of the corrected water depth being lower than the closure release threshold reaching the release time threshold, in step S930, a road closure release command is generated and sent to the closure execution device to release the road closure status. In response to the corrected water depth not being lower than the closure release threshold or the duration of the corrected water depth being lower than the closure release threshold not reaching the release time threshold, in step S940, the current road closure status is maintained.

[0072] In the embodiments of this application, during the execution of step S910, the aforementioned multi-node collaborative sensing logic is still used to output the water depth data after verification and correction, so as to eliminate the interference caused by local water wave sloshing when the rainfall decreases.

[0073] In the embodiments of this application, a de-shaking control mechanism is introduced during the execution of step S920. Even if the water level drops below the control release threshold, the stability of the state still needs to be confirmed. Only when the water depth remains below the control release threshold for a preset release time threshold (e.g., for 5 to 10 minutes) is it determined that the water accumulation has indeed subsided and the drainage work has made substantial progress.

[0074] In the embodiments of this application, the lockdown lifting threshold is different from the aforementioned real-time lockdown trigger threshold.

[0075] Specifically, the lockdown release threshold is set lower than the real-time lockdown trigger threshold. Only when the water level not only falls below the real-time lockdown trigger threshold but also further drops to a safer lockdown release threshold will the system enter the pre-lockdown release state. The difference between the real-time lockdown trigger threshold and the lockdown release threshold is calculated as follows: Figure 10 The hysteresis interval shown effectively filters out the impact of small fluctuations in water level at the critical point on decision-making.

[0076] In summary, through the edge-collaboration-based automatic road flooding control scheme provided above, this application embodiment achieves dynamic adaptive adjustment of the road closure trigger threshold by fusing video images and meteorological multimodal data. This makes the closure decision more consistent with real-time weather conditions, exhibiting higher rationality and foresight. Simultaneously, the introduction of an edge-collaboration mechanism effectively overcomes misjudgments and omissions caused by limited field of view or equipment failure of a single node through distributed interaction and consistency verification of perception information among multiple nodes. This allows for the calculation of highly accurate corrected flooding depth and collaborative reliability. This process not only significantly improves the accuracy of flooding perception and the overall robustness of the system but also achieves full-process automation and precision from flooding monitoring and level determination to equipment execution, greatly enhancing the efficiency and reliability of road traffic safety management under severe weather conditions.

[0077] Furthermore, in some embodiments, by establishing a precise mapping relationship between image pixels and actual physical height, non-contact, high-precision quantitative measurement of water depth is achieved, effectively overcoming the shortcomings of traditional physical water level sensors that are easily blocked by sludge or damaged by water immersion. Simultaneously, when evaluating the reliability of the identification, the spatial clarity of a single-frame image and the temporal dynamic stability of continuous video frames are comprehensively considered, and environmental interference factors such as raindrop occlusion and nighttime reflections are specifically introduced. This multi-dimensional weighted fusion mechanism of spatiotemporal characteristics and environmental interference can effectively identify and filter out visual noise caused by vehicle splashes, sudden changes in lighting, and severe weather, scientifically and objectively quantifying the reliability of the current visual perception results, greatly enhancing the system's anti-interference capability and data confidence in complex and extreme visual scenarios.

[0078] Furthermore, in some embodiments, standardized perception data encapsulation and distributed communication mechanisms break down information silos between individual computing nodes, enabling secure and efficient interoperability of multi-dimensional water accumulation status information between adjacent nodes. Simultaneously, a multi-level, rigorous verification model is constructed, encompassing numerical deviation comparison, level consistency judgment, and collaborative credibility threshold verification. This multi-node cross-validation mechanism accurately identifies and intercepts abnormal conflicting data caused by single-point device failures or local environmental interference, and promptly issues alarms to prevent misjudgment risks when verification fails. Finally, the corrected perception result is output only when multiple nodes reach a high-confidence consensus, thereby significantly improving the system's fault tolerance and resilience in complex road network environments, ensuring extremely high accuracy and security in the final road closure decision.

[0079] Furthermore, in some embodiments, a duration-based delay confirmation mechanism is innovatively introduced in the lockdown trigger determination stage. This effectively filters out interference data caused by water waves stirred up by vehicles wading through water or occasional short-term water level fluctuations, avoiding false alarms or frequent, repeated triggering of lockdown actions. This significantly enhances the rigor of lockdown decisions and the system's anti-vibration capabilities. Simultaneously, after confirming that the triggering conditions are met, the system comprehensively considers the corrected water depth, the initial lockdown level, and the reliability of multi-node collaboration to ultimately determine the lockdown level. This multi-dimensional integrated decision-making mechanism ensures that the generated control commands can more accurately match the actual safety risk level of the current road, achieving refined grading and highly reliable execution of emergency control measures.

[0080] This application also provides an automatic road flood control system based on edge collaboration. It can use the aforementioned automatic road flood control method 100 based on edge collaboration to perform automatic road flood control based on edge collaboration, or other methods can be used to perform automatic road flood control based on edge collaboration. This application does not limit this.

[0081] Figure 11 An exemplary structural block diagram of an edge-cooperative automatic road flood control system according to an embodiment of this application is shown.

[0082] like Figure 11 As shown, the system 1100 includes a video data processing module 1110, a meteorological data processing module 1120, a multi-node collaborative verification module 1130, a lockdown instruction generation module 1140, and a lockdown instruction issuance module 1150.

[0083] Specifically, the video data processing module 1110 is used to acquire video data of the road monitoring area in real time using edge computing nodes, and generate water depth and corresponding water depth recognition credibility based on the video data.

[0084] Specifically, the meteorological data processing module 1120 is used to acquire meteorological data corresponding to the road monitoring area, and determine the real-time lockdown trigger threshold and the corresponding initial lockdown level based on the meteorological data.

[0085] Specifically, the multi-node collaborative verification module 1130 is used to realize the interaction of perception information between the water depth, the credibility of water depth identification and the initial control level and the collaborative edge computing nodes based on the distributed communication mechanism, and to perform consistency verification on the multi-node perception information obtained by the interaction, so as to generate collaborative credibility and the corrected water depth.

[0086] Specifically, the lockdown command generation module 1140 is used to compare the corrected water depth with the real-time lockdown trigger threshold, determine the lockdown level based on the comparison result, and generate control commands.

[0087] Specifically, the lockdown instruction issuing module 1150 is used to send control instructions to the lockdown execution device to carry out road lockdown.

[0088] When system 1100 uses the aforementioned edge-collaborative automatic road flood control method 100 to perform automatic road flood control based on edge collaboration, the video data processing module 1110 executes the aforementioned step S110, the meteorological data processing module 1120 executes the aforementioned step S120, the multi-node collaborative verification module 1130 executes the aforementioned step S130, the control command generation module 1140 executes the aforementioned step S140, and the control command issuance module 1150 executes the aforementioned step S150. The specific execution process can be found above and will not be repeated here.

[0089] While numerous embodiments of this application have been shown and described herein, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will arise for those skilled in the art without departing from the spirit and intent of this application. It should be understood that various alternatives to the embodiments of this application described herein may be employed in the practice of this application. The appended claims are intended to define the scope of protection of this application and therefore cover equivalents or alternatives within the scope of these claims.

Claims

1. A method for automatic road flood control based on edge collaboration, characterized in that, include: Edge computing nodes acquire video data of the road monitoring area in real time, and generate water depth and corresponding water depth recognition credibility based on the video data; Acquire meteorological data corresponding to the road monitoring area, and determine the real-time lockdown trigger threshold and the corresponding initial lockdown level based on the meteorological data; The system uses a distributed communication mechanism to realize the interaction of the water depth, the water depth recognition credibility, the initial control level, and the perception information of the collaborative edge computing nodes. It also performs consistency verification on the multi-node perception information obtained through the interaction to generate collaborative credibility and the corrected water depth. The corrected water depth is compared with the real-time lockdown trigger threshold, and the lockdown level is determined and control commands are generated based on the comparison results. The control command is sent to the road closure execution device to carry out road closure.

2. The automatic road flood control method based on edge collaboration according to claim 1, characterized in that, The following steps are performed in the process of generating water depth based on video data: The video data is preprocessed to obtain enhanced image frames; Extract water surface features from the enhanced image frames to identify areas of road flooding; Based on the pixel coordinates of fixed water level markers or preset reference objects in the enhanced image frame, a mapping relationship between image pixel positions and actual physical height is established; The water depth is obtained by combining the pixel position of the water accumulation area in the enhanced image frame with the mapping relationship between the image pixel position and the actual physical height.

3. The automatic road flood control method based on edge collaboration according to claim 2, characterized in that, In the process of generating a confidence level for water depth recognition based on video data, the following steps are performed: Extract spatial evaluation features, which include the image sharpness of the video data and the sharpness of the boundary of the road waterlogged area in a single frame image; Temporal evaluation features are extracted from the historical identification results of the water depth. The temporal evaluation features include the displacement stability of the road waterlogged area boundary in continuous video frames and the change range of the water depth within a preset time period. Extract environmental interference features, including the proportion of raindrop obstruction and the intensity of nighttime reflective noise; The spatial evaluation features, the temporal evaluation features, and the environmental interference features are weighted and fused to generate a water depth identification reliability.

4. The automatic road flood control method based on edge collaboration according to claim 1, characterized in that, In determining the real-time lockdown trigger threshold and the corresponding initial lockdown level based on the meteorological data, the following steps are performed: Real-time rainfall intensity and meteorological warning characteristics are extracted from meteorological data to calculate the meteorological risk level that characterizes the degree of impact of precipitation; The preset road closure benchmark threshold is dynamically adjusted and corrected based on the meteorological risk level to generate a real-time closure trigger threshold that matches the current meteorological data. The water depth is compared with the real-time lockdown trigger threshold, and the initial lockdown level is obtained by matching the water depth with the real-time lockdown trigger threshold in a preset level mapping table.

5. The automatic road flood control method based on edge collaboration according to claim 1, characterized in that, In the process of realizing the interaction of the water depth, the reliability of the water depth identification, the initial control level, and the perception information of the collaborative edge computing nodes based on the distributed communication mechanism, the following steps are performed: The water depth, the water depth recognition confidence level, and the initial control level are combined and encapsulated into a local perception data frame with node identity identifier; Based on the distributed communication mechanism, the local sensing data frame is sent to the collaborative edge computing node with a connection relationship; The system synchronously receives collaborative sensing data frames sent by the collaborative edge computing nodes and parses out the collaborative water depth, collaborative water depth recognition credibility, and collaborative initial containment level corresponding to the collaborative edge computing nodes to complete the interaction of the sensing information.

6. The automatic road flood control method based on edge collaboration according to claim 5, characterized in that, In the process of verifying the consistency of multi-node perception information obtained through interaction in order to generate collaborative credibility and corrected water depth, the following steps are performed: Calculate the absolute value of the difference between the water depth and the combined water depth; Determine whether the absolute value of the difference is less than a preset error tolerance threshold, and whether the initial lockdown level is consistent with the collaborative initial lockdown level; In response to the absolute value of the difference being not less than a preset error tolerance threshold, or the initial lockdown level being inconsistent with the collaborative initial lockdown level, it is determined that the consistency verification of the multi-node perception information has failed, and a first alarm message is generated. In response to the absolute value of the difference being less than a preset error tolerance threshold, and the initial lockdown level being consistent with the collaborative initial lockdown level, the consistency verification of the multi-node perception information is preliminarily passed. Based on the credibility of the water depth recognition and the credibility of the collaborative water depth recognition, the collaborative credibility is generated, and based on the water depth and the collaborative water depth, the corrected water depth is generated. Determine whether the collaboration credibility meets the collaboration credibility threshold range; In response to the fact that the collaborative credibility is within the collaborative credibility threshold range, the consistency verification of the multi-node perception information is determined to be passed, and the collaborative credibility and the corrected water depth are output. A second alarm message is generated in response to the fact that the collaboration trustworthiness is not within the collaboration trustworthiness threshold range.

7. The automatic road flood control method based on edge collaboration according to claim 1, characterized in that, In the process of determining the lockdown level and generating control instructions based on the comparison results, the following steps are performed: Determine whether the corrected water depth exceeds the real-time lockdown trigger threshold, and whether the duration of the corrected water depth exceeding the real-time lockdown trigger threshold reaches a preset time; If the corrected water depth does not exceed the real-time closure trigger threshold, or if the duration of the corrected water depth exceeding the real-time closure trigger threshold does not reach the preset time, it is determined that the road closure trigger condition is not met. If the corrected water depth exceeds the real-time closure trigger threshold and the duration of the corrected water depth exceeding the real-time closure trigger threshold reaches a preset time, it is determined that the road closure trigger condition is met. The closure level is determined based on the corrected water depth and the coordination reliability, and a corresponding road closure control command is generated based on the closure level.

8. The automatic road flood control method based on edge collaboration according to claim 1, characterized in that, During the process of sending the control command to the road closure execution device to close the road, the following steps are performed: The edge computing node with control is determined to be the master node through a preset arbitration mechanism. The master node is determined based on the health status, communication status or collaborative consistency results of each edge computing node. Determine whether the master control node has malfunctioned; In response to the absence of any abnormality in the master control node, the control command is directly sent to the road closure execution device through the master control node to implement the road closure operation; In response to an anomaly in the master control node, the arbitration mechanism allocates master control to the corresponding collaborative edge computing node, and the collaborative edge computing node with master control sends the control command directly to the blockade execution device to implement the road blockade operation.

9. The automatic road flood control method based on edge collaboration according to claim 1, characterized in that, The method further includes: lifting road closures based on the corrected water depth obtained in real time; During the process of lifting road closures based on the corrected water depth obtained in real time, the following steps are performed: Real-time monitoring of the corrected water depth corresponding to the road monitoring area; Determine whether the corrected water depth is lower than the lockdown release threshold and whether the duration of the corrected water depth being lower than the lockdown release threshold reaches the release time threshold; In response to the modified water depth being lower than the lockdown release threshold and the duration of the modified water depth being lower than the lockdown release threshold reaching the release time threshold, a road lockdown release command is generated and sent to the lockdown execution device to release the road lockdown status; If the corrected water depth is not lower than the closure release threshold or the duration of the corrected water depth being lower than the closure release threshold does not reach the release time threshold, the current road closure status will be maintained. The threshold for lifting the lockdown is different from the real-time lockdown trigger threshold.

10. An automatic road flood control system based on edge collaboration, characterized in that, The system employs the edge-cooperative automatic road flood control method as described in any one of claims 1-9, wherein the system comprises: The video data processing module is used to acquire video data of the road monitoring area in real time using edge computing nodes, and generate water depth and corresponding water depth recognition credibility based on the video data; The meteorological data processing module is used to acquire meteorological data corresponding to the road monitoring area, and determine the real-time lockdown trigger threshold and the corresponding initial lockdown level based on the meteorological data; The multi-node collaborative verification module is used to realize the interaction of the water depth, the water depth recognition credibility, and the initial control level with the perception information of the collaborative edge computing nodes based on the distributed communication mechanism, and to perform consistency verification on the multi-node perception information obtained by the interaction, so as to generate collaborative credibility and the corrected water depth. The lockdown command generation module is used to compare the corrected water depth with the real-time lockdown trigger threshold, determine the lockdown level based on the comparison result, and generate control commands. The road closure instruction issuing module is used to send the control instructions to the road closure execution device to carry out road closure.

Images