Inspection data monitoring method and server applied to early warning of roads and bridges in mountainous areas

Abstract

The invention provides an inspection data monitoring method and server applied to mountain road and bridge early warning, and the method comprises the steps: firstly obtaining road and bridge remote sensing monitoring data containing current topographic distribution characteristics, road and bridge structure state parameters and historical disaster records, and then, according to the current topographic distribution characteristics, calling a topographic feature extraction model to segment a target mountain road section, generating inspection subareas and topographic feature tags, generating a dynamic flight path containing inspection nodes and flight parameters for the unmanned aerial vehicle through a path planning model according to the topographic feature tags and road and bridge structure state parameters, and then controlling the unmanned aerial vehicle to inspect according to the path. According to the method, multi-dimensional monitoring data such as surface deformation data, road and bridge crack images and slope displacement tracks are collected, finally, the multi-dimensional monitoring data and historical disaster records are input into a risk early warning model, risk level signals and disposal suggestion information are generated, comprehensive intelligent monitoring and early warning of mountainous area roads and bridges are achieved, and the monitoring accuracy and efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to an inspection data monitoring method and a server for early warning of roads and bridges in mountainous areas. Background Art

[0002] During the construction and operation of mountain roads and bridges, ensuring the safety and stability of roads and bridges is crucial to ensuring smooth transportation and the safety of people's lives and property. Due to the complex geographical environment, undulating terrain, diverse geological conditions, and frequent natural disasters such as landslides, mudslides, and earthquakes, mountain roads and bridges face many potential risks. Therefore, effective monitoring and early warning of mountain roads and bridges is particularly critical.

[0003] At present, there are many ways to monitor roads and bridges in mountainous areas. The traditional manual inspection method relies on inspectors to check the condition of roads and bridges on site. However, due to the complex terrain in mountainous areas, manual inspections are not only inefficient, but also difficult to reach some rugged areas, prone to inspection blind spots, and unable to fully and accurately grasp the overall condition of roads and bridges.

[0004] With the development of technology, some monitoring systems based on fixed sensors have been applied. These monitoring systems obtain some data by installing fixed sensors at key locations of roads and bridges, such as simple displacement sensors to monitor the displacement changes of roads and bridges. However, such methods have obvious limitations. The installation position of sensors is fixed, and it is impossible to flexibly adjust the monitoring scope and focus according to the actual risk status and terrain characteristics of roads and bridges. The risk monitoring capabilities of complex terrain environments around roads and bridges and dynamic changes in different regions are insufficient.

[0005] At the same time, in terms of inspection route planning, most existing technologies use preset fixed routes. Whether it is manual inspection or inspection using equipment such as drones, fixed routes cannot adapt to the diversity of mountainous terrain and the dynamic changes in the state of road and bridge structures. This leads to insufficient inspection frequency in some high-risk areas, while too many resources are invested in some low-risk areas, and it is impossible to reasonably and efficiently allocate inspection forces. Summary of the invention

[0006] In view of the above-mentioned problems, in combination with the first aspect of the present invention, an embodiment of the present invention provides a patrol data monitoring method for early warning of roads and bridges in mountainous areas, the method comprising: Acquire remote sensing monitoring data of roads and bridges in target mountainous road sections, wherein the remote sensing monitoring data of roads and bridges includes current terrain distribution characteristics, road and bridge structure status parameters, and historical disaster records; Based on the current terrain distribution characteristics, calling the terrain feature extraction model to perform regional segmentation on the target mountainous road section, and generating multiple inspection sub-areas and corresponding terrain feature labels; According to the terrain feature labels and the road and bridge structure state parameters, a dynamic flight path of the UAV in the inspection sub-area is generated through a path planning model, wherein the dynamic flight path includes a plurality of inspection nodes and flight parameters between the inspection nodes; Based on the dynamic flight path, the UAV is controlled to perform real-time inspection tasks, and multi-dimensional monitoring data in the inspection sub-area is collected, wherein the multi-dimensional monitoring data includes surface deformation data, road and bridge crack images, and slope displacement trajectories; The multi-dimensional monitoring data and the historical disaster records are input into a risk warning model to generate a risk level signal and corresponding handling suggestion information for the target mountainous road section.

[0007] On the other hand, an embodiment of the present invention further provides a server, including a processor and a machine-readable storage medium, wherein the machine-readable storage medium is connected to the processor, the machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine-readable storage medium to implement the above method.

[0008] Based on the above aspects, the embodiment of the present application obtains the road and bridge remote sensing monitoring data including the current terrain distribution characteristics, the road and bridge structure state parameters and the historical disaster records, performs regional segmentation based on the current terrain distribution characteristics and generates terrain feature labels, and then generates the dynamic flight path of the UAV in combination with the road and bridge structure state parameters, and closely combines the terrain factors, the road and bridge structure and the UAV inspection path planning. Different from the traditional fixed path inspection method, the dynamic flight path generated by this method can intelligently adjust the inspection route and nodes according to the variability of the mountain terrain and the differences in the road and bridge structure state. Not only does it effectively cover all key areas and avoid inspection blind spots, but it can also reasonably allocate inspection resources according to the characteristics and risk levels of different regions, and improve the inspection efficiency and pertinence. At the same time, the flight parameter settings in the dynamic flight path ensure that the UAV performs the inspection task safely, stably and efficiently in a complex mountain environment, further improving the reliability and practicality of the entire monitoring system. In the real-time inspection task, the multi-dimensional monitoring data, including surface deformation data, road and bridge crack images and slope displacement trajectories, etc., can accurately capture the slight changes of the road and bridge under the influence of various factors, and provide a rich and accurate basis for timely discovery of potential risks. Finally, the multi-dimensional monitoring data and historical disaster records are input into the risk warning model to generate risk level signals and disposal suggestion information, realizing intelligent assessment and warning of the risks of the target mountainous road sections, taking into full account the evolution law and current actual conditions of mountainous road and bridge disasters, and being able to more accurately predict the possible disaster risks and provide targeted and highly operational disposal suggestions. Compared with traditional warning methods, it greatly improves the accuracy and timeliness of warnings, buys precious time for timely and effective preventive measures, effectively ensures the safe operation of mountainous roads and bridges, and reduces disaster losses. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 It is a schematic diagram of the execution flow of the inspection data monitoring method applied to mountainous road and bridge early warning provided by an embodiment of the present invention.

[0010] Figure 2 It is a schematic diagram of the server hardware architecture provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0011] The present invention will be described in detail below with reference to the accompanying drawings. Figure 1 It is a flow chart of a patrol data monitoring method for early warning of roads and bridges in mountainous areas provided by an embodiment of the present invention. The patrol data monitoring method for early warning of roads and bridges in mountainous areas is introduced in detail below.

[0012] Step S110, obtaining the road and bridge remote sensing monitoring data of the target mountainous road section, wherein the road and bridge remote sensing monitoring data includes current terrain distribution characteristics, road and bridge structure status parameters and historical disaster records.

[0013] For example, consider a section of highway in a mountainous area, which is located between mountains and has complex and changeable surrounding terrain. The multispectral image data and three-dimensional point cloud data of this target mountainous section are obtained through satellite remote sensing equipment. After the multispectral image data is processed by ground object classification, the sub-features of the current terrain distribution characteristics are obtained. For example, in a certain area, the vegetation is dense and the vegetation coverage density is high, while in other areas, there are more exposed rocks, and hydrological distribution features such as rivers can be clearly seen, which constitute the current terrain distribution characteristics. For the three-dimensional point cloud data, after elevation difference processing, an accurate terrain slope matrix and surface undulation index are generated. For example, on the hillside on one side of a certain section of road, the terrain slope matrix shows a large slope and a high surface undulation.

[0014] The road and bridge structure status parameters were extracted from the historical maintenance database of the highway section. Among them, a certain bridge pier had a certain degree of inclination, and its inclination angle was accurately recorded. The road surface had settlement in some sections, and the settlement rate was also measured and recorded. The slope support structure integrity parameters can also be checked in the database.

[0015] In addition, historical disaster records show that this mountainous section of road has experienced multiple disaster events over the past few years. For example, a landslide occurred at a specific location, causing damage to part of the roadbed; a bridge crack expansion event was found on a bridge, which has been repaired but still needs close attention; and a roadbed collapse event occurred on another section of the road. These disaster events are classified by type, and their corresponding geographic coordinates and timestamps are recorded in detail.

[0016] Step S120 , based on the current terrain distribution characteristics, calling a terrain feature extraction model to perform regional segmentation on the target mountainous road section, and generating a plurality of inspection sub-areas and corresponding terrain feature labels.

[0017] Taking the mountain highway section mentioned above as an example, the terrain feature extraction model can segment the entire mountain section based on the current terrain distribution characteristics obtained previously, such as vegetation coverage density, rock exposure areas, and hydrological distribution in different areas. For example, an area with dense vegetation, gentle slopes, and close to water sources is divided into an inspection sub-area, and the terrain feature label of this area is "dense vegetation-low slope-near water". For those sections with exposed rocks and large slopes, they are divided into another inspection sub-area, and its terrain feature label is "exposed rocks-high slope". In this way, the entire mountain section is divided into multiple inspection sub-areas, each of which has its own unique terrain feature label. These terrain feature labels accurately reflect the terrain characteristics of the sub-area, providing an important basis for the subsequent inspection path planning of drones.

[0018] Step S130, generating a dynamic flight path of the UAV in the inspection sub-area through a path planning model according to the terrain feature labels and the road and bridge structure state parameters, wherein the dynamic flight path includes a plurality of inspection nodes and flight parameters between the inspection nodes.

[0019] Still taking the mountain highway section as an example, for the different inspection sub-areas previously divided and their corresponding terrain feature labels and road and bridge structure status parameters, the path planning model is used to plan the flight path of the UAV.

[0020] For example, in the inspection sub-area of ​​dense vegetation, low slope and near water, due to the high density of vegetation coverage, the terrain slope matrix shows a low slope. According to relevant calculations, the minimum safe flight altitude of the drone in this area is relatively low, and the obstacle avoidance buffer distance is also relatively small. At the same time, according to the road and bridge structure status parameters, if the inclination angle of a bridge pier is small and the integrity of the slope support structure is good, then when determining the priority order of the inspection nodes, these nodes have a relatively low priority, and the data collection time of each inspection node is also short.

[0021] For example, for the inspection node of a bridge, due to its good structural condition and low priority, the data collection time is set to a shorter time. Then, the path optimization algorithm is used to spatially sort these inspection nodes. First, the corresponding flight altitude lower limit and obstacle avoidance buffer boundary range are assigned to each inspection node. For example, for an inspection node close to vegetation, its flight altitude lower limit is set to be slightly higher than the vegetation height plus a certain safety margin, and the obstacle avoidance buffer boundary range takes into account the interference factor of vegetation. Then, the inspection nodes are sorted from high to low according to priority. When the priority is the same, the data collection time is sorted from short to long to form a node access sequence with time weight. Then, multiple candidate flight paths are generated between two adjacent inspection nodes. For example, from the inspection node at one end of the bridge to the inspection node at the other end, there are multiple different flight paths to choose from. Each candidate flight path must meet the obstacle avoidance buffer boundary range requirements and flight altitude requirements of the starting node and the end node.

[0022] For each candidate flight path, the comprehensive cost value is calculated. For example, if a candidate flight path is close to an area where a landslide event has occurred, and the overlap ratio between the candidate flight path and the spatial distribution density of disaster points in the historical disaster records is high, then its safety risk cost is high, and the comprehensive cost value increases accordingly. The path with the lowest comprehensive cost value is selected from the candidate path set as the reference path. If there are nodes in the reference path where the data collection time exceeds the preset threshold, such as the pier of a complex structure, the optimized target path is generated by exchanging the node order or inserting an intermediate transition node. Finally, the flight speed of the drone and the sensor trigger frequency are dynamically matched according to the flight altitude lower limit value and the obstacle avoidance buffer boundary range of each path segment in the target path. For example, in the path segment with a higher flight altitude, the flight speed can be appropriately increased, while in the area with a smaller obstacle avoidance buffer boundary range, the sensor trigger frequency is increased to ensure the accuracy of data collection.

[0023] At the same time, the flight parameters are dynamically adjusted based on real-time meteorological data. If the real-time wind speed of the mountainous road section exceeds the first preset threshold during the inspection, the UAV reduces the flight speed according to the preset deceleration strategy, and increases the obstacle avoidance buffer distance according to the preset obstacle avoidance buffer strategy to ensure flight safety. If the rainfall intensity exceeds the second preset threshold, the anti-interference mode of the UAV is enabled, and the sensor trigger frequency is increased according to the first preset compensation parameter, because rainfall may affect the data collection accuracy, and increasing the sensor trigger frequency can compensate for this loss of accuracy. If the visibility is lower than the third preset threshold, the infrared imaging module of the UAV is activated to replace the visible light camera, and the data collection time of the corresponding inspection node is extended according to the second preset compensation parameter to obtain more comprehensive and accurate monitoring data. And predict the flight risk in the future time window based on the changing trend of meteorological data. If the predicted risk level exceeds the critical value, send a path replanning instruction to the navigation control unit of the UAV.

[0024] In the entire flight path planning, emergency avoidance nodes will also be inserted into the initial flight path. For example, according to historical disaster records, in areas where multiple landslides have occurred, the spatial distribution density is high. Emergency avoidance nodes are adaptively configured in these areas to generate an adjusted flight path. Finally, the adjusted flight path and flight parameters are integrated into a dynamic flight path and synchronized to the navigation control unit of the drone. A path deviation detection mechanism is preset in the navigation control unit. When the lateral offset between the actual flight trajectory of the drone and the dynamic flight path continues to exceed the preset ratio of the obstacle avoidance buffer boundary range, local path replanning is triggered, and only the remaining nodes that have not been visited are re-sorted in space, etc., to ensure that the drone can complete the inspection task according to the predetermined path.

[0025] Step S140, based on the dynamic flight path, the UAV is controlled to perform a real-time inspection task to collect multi-dimensional monitoring data in the inspection sub-area, wherein the multi-dimensional monitoring data includes surface deformation data, road and bridge crack images, and slope displacement trajectories.

[0026] In the above-mentioned mountain highway section, the drone begins to perform inspection tasks according to the dynamic flight path. When the drone flies to a certain area, it scans the surface deformation area through the lidar sensor. For example, in a section where there have been signs of minor landslides, the drone continuously emits laser pulses during flight, receives reflected signals and calculates the pulse round-trip time difference to generate an initial point cloud data set. Then, the initial point cloud data is denoised to remove abnormal points caused by surrounding vegetation occlusion or meteorological interference (such as scattering interference caused by foggy days) to generate denoised point cloud data. Next, the denoised point cloud data is elevation corrected based on the previously obtained terrain slope matrix to eliminate the error effect of terrain undulation on deformation calculation and obtain the current point cloud data. The time series difference algorithm is used to compare the current point cloud data with the historical benchmark data to calculate the three-dimensional displacement and displacement direction of each monitoring point. If the displacement of a certain area exceeds the warning threshold, the area is marked as a risk area, and the risk level of the risk area is distinguished by color gradient in the deformation displacement vector map, thereby obtaining surface deformation data.

[0027] At the same time, an optical camera is used to capture images of the surface of the bridge. During the inspection of a bridge, the surface image of the bridge is captured, and the crack length, width and direction characteristics corresponding to the image are extracted through the edge detection algorithm. For example, a crack is found in a certain supporting part of the bridge, the crack length is several centimeters, the width is gradually increasing, and the direction is along a certain direction. These structured data constitute the crack image of the bridge.

[0028] The multi-band radar of the drone is used to continuously monitor the slope. In a certain slope area, the multi-band radar captures the slope displacement trajectory of the rock and soil inside the slope, including the displacement rate curve and sliding direction indicators. For example, the displacement rate curve shows that the displacement rate gradually increases within a certain period of time, and the sliding direction is towards the roadbed.

[0029] In addition, the drone also collects environmental temperature and humidity, wind speed and direction, and vibration frequency parameters simultaneously. For example, at a certain inspection time, the ambient temperature is how many degrees Celsius, the humidity is how many percentages, the wind speed and direction, and the vibration frequency are all accurately collected. These auxiliary verification information are aligned with the surface deformation data, road and bridge crack images, and slope displacement trajectories in time series, and are bound to the corresponding geographic coordinates and stored in a distributed database.

[0030] Step S150, inputting the multi-dimensional monitoring data and the historical disaster records into a risk warning model to generate a risk level signal and corresponding handling suggestion information for the target mountainous road section.

[0031] Continue to use the risk warning model for analysis based on the previously collected multi-dimensional monitoring data and the historical disaster records of this mountain road section.

[0032] For surface deformation data, if the displacement rate in a certain area continues to increase and the direction is consistent with the direction of historical landslides, for example, the surface displacement rate under a roadbed continues to increase over a period of time, and the displacement direction is the same as the direction of a previous landslide, then a first-level warning signal is triggered.

[0033] When analyzing the crack images of road bridges, if it is found that the extension direction of the cracks in a key part of the bridge coincides with the main stress direction of the bridge, and the width expansion rate exceeds the critical value, for example, the crack width grows rapidly in a short period of time and reaches a level that may affect the safety of the bridge structure, a secondary warning signal will be triggered.

[0034] The slope stability coefficient is calculated based on the sliding direction and speed of the slope displacement trajectory combined with real-time rainfall intensity data. For example, if the displacement trajectory of a certain slope shows that its sliding direction is toward the road, the displacement speed is fast, and the real-time rainfall intensity is large, the slope stability coefficient is lower than the safety threshold through calculation (such as vector decomposition of the sliding direction, comparison with historical disaster records, and searching the correlation mapping table between rainfall accumulation and displacement acceleration interval, etc.), and a level 3 warning signal is triggered.

[0035] Differentiated handling suggestions are generated based on the warning signal level. If a level 1 warning signal is triggered, emergency road closure measures may be taken; if a level 2 warning signal is triggered, speed limit measures may be taken; if a level 3 warning signal is triggered, a support structure reinforcement plan may be initiated.

[0036] Finally, the risk level signal and disposal suggestion information are pushed to the road and bridge management terminal, and the corresponding monitoring data traceability link is associated for manual review. For example, the road and bridge management department receives a risk level signal of a certain section of road as level 2 on the terminal, and the disposal suggestion is speed limit. At the same time, detailed monitoring data, such as bridge crack images, surface deformation data, etc., can be viewed through the traceability link for manual review to ensure the accuracy of the disposal decision.

[0037] In addition, a feedback closed-loop mechanism between multi-dimensional monitoring data and disposal suggestion information can be established throughout the process. When the disposal suggestion information is adopted and executed by the road and bridge management terminal, the structural response data of the corresponding area is continuously monitored. For example, after the support structure reinforcement plan is implemented on a certain slope, the drone continues to monitor the deformation data of the reinforcement material and the stress distribution data of the support structure in the area. By comparing the difference in slope displacement trajectory before and after the treatment, the effective radius and attenuation coefficient of the disposal measures corresponding to the disposal suggestion information are calculated. If the implementation attenuation coefficient exceeds the standard deviation of the preset multiple of the historical average value, it is determined that the current disposal plan has design defects. For example, if the displacement of the slope after reinforcement is still large and exceeds expectations, an optimization suggestion for material replacement or structural redesign is generated. The optimization suggestion is input into the path planning model to increase the monitoring frequency and sensor accuracy level of the defective area, and when the new disposal plan is implemented, the feedback closed-loop mechanism is restarted until the actual improvement rate reaches the standard, so as to continuously optimize the monitoring and disposal strategies and ensure the safe operation of roads and bridges in mountainous sections.

[0038] Based on the above steps, the embodiment of the present application obtains the road and bridge remote sensing monitoring data including the current terrain distribution characteristics, the road and bridge structure state parameters and the historical disaster records, performs regional segmentation based on the current terrain distribution characteristics and generates terrain feature labels, and then generates the dynamic flight path of the UAV in combination with the road and bridge structure state parameters, and closely combines the terrain factors, the road and bridge structure and the UAV inspection path planning. Different from the traditional fixed path inspection method, the dynamic flight path generated by this method can intelligently adjust the inspection route and nodes according to the variability of the mountain terrain and the differences in the road and bridge structure state. Not only does it effectively cover all key areas and avoid inspection blind spots, but it can also reasonably allocate inspection resources according to the characteristics and risk levels of different regions, and improve the inspection efficiency and pertinence. At the same time, the flight parameter settings in the dynamic flight path ensure that the UAV performs the inspection task safely, stably and efficiently in a complex mountain environment, further improving the reliability and practicality of the entire monitoring system. In the real-time inspection task, the collection of multi-dimensional monitoring data, including surface deformation data, road and bridge crack images and slope displacement trajectories, can accurately capture the slight changes of the road and bridge under the influence of various factors, and provide a rich and accurate basis for timely discovery of potential risks. Finally, the multi-dimensional monitoring data and historical disaster records are input into the risk warning model to generate risk level signals and disposal suggestion information, realizing intelligent assessment and warning of the risks of the target mountainous road sections, taking into full account the evolution law and current actual conditions of mountainous road and bridge disasters, and being able to more accurately predict the possible disaster risks and provide targeted and highly operational disposal suggestions. Compared with traditional warning methods, it greatly improves the accuracy and timeliness of warnings, buys precious time for timely and effective preventive measures, effectively ensures the safe operation of mountainous roads and bridges, and reduces disaster losses.

[0039] In a possible implementation, step S110 includes: Step S111, obtaining multispectral image data and three-dimensional point cloud data of the target mountainous road section through satellite remote sensing equipment.

[0040] Step S112, performing ground object classification processing on the multispectral image data, extracting vegetation coverage density, rock exposed area and hydrological distribution characteristics as sub-features of the current terrain distribution characteristics.

[0041] For example, satellites conduct all-round scanning and imaging of the mountain highway sections mentioned above from space, and the multispectral image data contains rich information. Then, the multispectral image data is processed for ground object classification to extract sub-features such as vegetation coverage density, rock exposure area and hydrological distribution characteristics. For example, in some sections of mountain sections, vegetation grows lushly, and the vegetation coverage density value of the area is obtained through precise analysis and calculation. These vegetation may have potential impacts on the road and bridge structure, such as the roots of vegetation may damage the roadbed. In other parts, there are rock exposure areas, which may indicate that the geological structure of the area is less stable and prone to disasters such as landslides. At the same time, the analysis can also clarify the hydrological distribution characteristics, such as the direction of the river and its relative position to the road and bridge. If the river is too close to the road and bridge, it may cause erosion and other hazards to the road and bridge foundation.

[0042] Step S113, performing elevation difference processing on the three-dimensional point cloud data to generate a terrain slope matrix and a surface undulation index, and associating the terrain slope matrix and the surface undulation index with corresponding inspection sub-areas.

[0043] The acquired three-dimensional point cloud data needs to be processed by elevation difference. In this process, the terrain slope matrix and the surface undulation index are generated. For example, in a certain section of a mountain highway, the terrain slope matrix shows that the slope is large, which means that there are greater challenges when driving or building and maintaining this section. The surface undulation index also reflects the complexity of the terrain in the area, such as some hilly areas with large undulations. Then, the terrain slope matrix and the surface undulation index are associated with the corresponding inspection sub-areas, which will help to formulate targeted inspection plans based on different terrain characteristics.

[0044] Step S114, calling up the historical maintenance database of the target mountainous road section, extracting the pier inclination angle, road surface settlement rate and slope support structure integrity parameters from the road bridge structure status parameters.

[0045] In the historical maintenance database of the target mountainous road section, many parameters related to the road and bridge structural status are recorded in detail. For example, the piers under a bridge on the mountain highway are regularly measured by professional measuring instruments to obtain the data of the pier inclination angle. The value of the pier inclination angle reflects the structural stability of the pier. If the inclination angle is too large, it may affect the safety of the entire bridge. The pavement settlement rate is also an important parameter. In some sections, due to the long-term influence of factors such as vehicle loads and geological changes, the pavement will settle. By analyzing historical data, an accurate pavement settlement rate can be obtained to evaluate the pavement condition. There are also slope support structure integrity parameters. In order to prevent the impact of disasters such as landslides on roads and bridges, slope support structures are set up on the slopes of mountainous sections. The detection data of their integrity is recorded in the database. Poor integrity means that the slope's protective capacity may be reduced and there is a risk of landslides.

[0046] Step S115, classifying the historical disaster records according to disaster types, which include landslide events, bridge crack expansion events and roadbed collapse events, and associating the geographical coordinates and occurrence timestamps corresponding to each event.

[0047] For example, this mountain highway section has experienced many types of disasters, such as landslides, bridge crack expansion events, and roadbed collapse events. For landslide events, the geographic coordinates of the event are clearly identified, such as on a hillside near a bend, with accurate longitude and latitude values, and the timestamp of the event is recorded to analyze the cause of the landslide in the area under specific time and geological conditions. For bridge crack expansion events, the specific part of the bridge where the event occurred and the corresponding geographic coordinates are determined, and the rate and trend of crack expansion are analyzed through the timestamp. Roadbed collapse events also record the exact location and time of the event.

[0048] In a possible implementation, step S130 includes: Step S131, calculating the minimum safe flight altitude and obstacle avoidance buffer distance of the UAV in the corresponding inspection sub-area according to the vegetation coverage density and terrain slope matrix in the terrain feature label.

[0049] Step S132, based on the pier inclination angle and slope support structure integrity parameters in the road and bridge structure status parameters, determine the priority order of the inspection nodes and the data collection time of each inspection node.

[0050] Among them, the minimum safe flight altitude H_min=H_base+α×V+β×S̄ The obstacle avoidance buffer distance D_buffer=γ×(V / V_max)+δ×(ΔS / S_ref) The priority of the inspection node P = θ × w1 + (1-I) × w2 The data collection duration T = T_base × (1 + P / P_max) H_base is the benchmark safety height, α is the weight coefficient of vegetation coverage density V, β is the correction coefficient of the average slope S̄ of the terrain slope matrix, γ is the influencing factor of vegetation density, V_max is the maximum vegetation coverage density in the region, δ is the adjustment coefficient of the slope change rate ΔS, S_ref is the slope reference threshold, θ is the inclination angle of the pier, I is the normalized value of the integrity parameter of the slope support structure, w1 and w2 are the weight coefficients of the inclination angle and integrity respectively, T_base is the basic collection time, and P_max is the maximum priority value.

[0051] In this embodiment, a certain inspection sub-area in a mountain highway section is taken as an example. The vegetation coverage density in this area is at a high level according to the previous analysis, and the terrain slope matrix shows that the average slope is also large. For the calculation of the minimum safe flight altitude, the base safety altitude H_base is a basic setting value, which is assumed to be a fixed value here, such as 100 meters. The vegetation coverage density V is relatively high, and its weight coefficient α is set according to experience and relevant technical standards, and is assumed to be 0.5. The average slope S̄ of the terrain slope matrix is ​​relatively large, and its correction coefficient β is assumed to be 0.3. According to the formula of minimum safe flight altitude H_min = H_base + α×V + β×S̄, due to the influence of vegetation coverage density and average slope, it is calculated that the minimum safe flight altitude of the area may reach 120 meters to ensure that the drone will not collide with vegetation or be in danger due to terrain undulations during flight.

[0052] For the obstacle avoidance buffer distance, the maximum vegetation coverage density V_max in the area is assumed to be a certain value, such as 80%, the vegetation density influence factor γ is assumed to be 0.2, the slope change rate ΔS is obtained based on the terrain slope matrix analysis, the slope reference threshold S_ref is a fixed reference standard value, assumed to be 15 degrees, and the slope change rate adjustment coefficient δ is assumed to be 0.1. According to the formula of obstacle avoidance buffer distance D_buffer = γ×(V / V_max)+δ×(ΔS / S_ref), due to the influence of vegetation coverage density and slope change rate, the obstacle avoidance buffer distance of the area is 5 meters, which can provide sufficient safety buffer space for the drone to avoid obstacles.

[0053] Next, at a bridge on a mountain highway, the pier inclination angle θ was measured to be 2 degrees, and the slope support structure integrity parameter was evaluated and the normalized value I was 0.8. The weight coefficient w1 of the inclination angle was assumed to be 0.6, and the weight coefficient w2 of the integrity was assumed to be 0.4. According to the formula of the inspection node priority P = θ×w1+(1 - I)×w2, the priority of the inspection node was calculated to be 0.52. Assuming that the basic collection time T_base is 10 seconds and the maximum priority P_max is 1, according to the formula of data collection time T = T_base×(1 + P / P_max), the data collection time of the inspection node is 15.2 seconds. In the entire inspection sub-area, different inspection nodes determine the priority order and data collection time according to the above calculation method based on their corresponding pier inclination angles and slope support structure integrity parameters. For those inspection nodes with larger pier inclination angles or poor slope support structure integrity, higher priority and longer data collection time will be given to ensure more detailed monitoring of key parts.

[0054] Step S133, based on the minimum safe flight altitude and obstacle avoidance buffer distance of the UAV in the corresponding inspection sub-area, as well as the priority order of the inspection nodes and the data collection time of each inspection node, a path optimization algorithm is used to spatially sort the inspection nodes, generate an initial flight path, and dynamically adjust the flight parameters based on real-time meteorological data, the flight parameters including flight speed, hovering time and sensor trigger frequency.

[0055] Step S134, inserting emergency avoidance nodes into the initial flight path to generate an adjusted flight path, wherein the emergency avoidance nodes are adaptively configured according to the spatial distribution density of disaster types in historical disaster records.

[0056] For example, in mountainous highway sections, areas where multiple landslides have occurred have a high spatial distribution density, and in these areas, it is necessary to adaptively configure emergency avoidance nodes. For example, on the flight path near a section where landslides often occur, an emergency avoidance node is inserted so that the drone can avoid these dangerous areas during the inspection process and generate an adjusted flight path.

[0057] Step S135, integrating the adjusted flight path and the flight parameters into the dynamic flight path, and synchronizing the dynamic flight path to the navigation control unit of the UAV, and presetting a path deviation detection mechanism in the navigation control unit. When the lateral offset between the actual flight trajectory of the UAV and the dynamic flight path continues to exceed a preset ratio of the obstacle avoidance buffer boundary range, local path replanning is triggered, and only the remaining nodes that have not been visited are returned to execute the step of spatially sorting the inspection nodes using a path optimization algorithm to generate an initial flight path.

[0058] For example, when the lateral offset between the actual flight trajectory of the UAV and the dynamic flight path continues to exceed the preset ratio of the obstacle avoidance buffer boundary range, assuming the preset ratio is 20%, if the lateral offset of the UAV continues to exceed this ratio, it will trigger local path replanning, and only return to execute the spatial sorting of the inspection nodes using the path optimization algorithm for the remaining nodes that have not been visited, and generate the steps of the initial flight path to ensure that the UAV can complete the inspection task according to the predetermined path and ensure the effective monitoring of the road and bridge structures of mountain highway sections.

[0059] In a possible implementation, step S133 includes: Step S1331, based on the minimum safe flight altitude and obstacle avoidance buffer distance of each inspection sub-area, a corresponding flight altitude lower limit value and obstacle avoidance buffer boundary range are assigned to each inspection node, and a set of inspection node safety parameters including altitude constraints and spatial obstacle avoidance ranges are generated.

[0060] In this embodiment, a specific inspection sub-area of ​​a mountain highway is taken as an example, and there are multiple inspection nodes in the area. One of the inspection nodes is close to the hillside and has more vegetation. According to the previously calculated minimum safe flight altitude of 120 meters, the lower limit of the flight altitude assigned to the inspection node is 120 meters, so as to ensure that the drone will not collide with the mountain or vegetation due to too low altitude when flying over the node. At the same time, considering that the obstacle avoidance buffer distance of the area is 5 meters, the obstacle avoidance buffer boundary range of the inspection node is set as the area with a radius of 5 meters centered on the node. In the same way, each inspection node in the inspection sub-area is operated, and finally a set of inspection node safety parameters including height constraints and spatial obstacle avoidance ranges is generated.

[0061] Step S1332, according to the priority order of the inspection nodes and the data collection time, the inspection nodes are sorted from high to low according to the priority, and if the priority is the same, they are secondary sorted from short to long according to the data collection time, to form a node access sequence with time weight.

[0062] For example, at a patrol node near a bridge on a mountain highway section, the pier has a large inclination angle and the integrity of the slope support structure is relatively poor. According to calculations, the patrol node has a higher priority. Another patrol node that is far away from the bridge and has a better structural condition has a relatively low priority. For the data collection time, the patrol node with a high priority is calculated to have a data collection time of 15.2 seconds, and the patrol node with a low priority has a data collection time of 10 seconds. According to the above sorting rules, all patrol nodes are first arranged from high to low according to priority. If there are multiple patrol nodes with the same priority, such as two patrol nodes far away from the main structure with the same priority, but one of them has a data collection time of 8 seconds and the other has a data collection time of 10 seconds, then the patrol node corresponding to the data collection time of 8 seconds is arranged in front according to the data collection time from short to long, thereby forming a node access sequence with time weight.

[0063] Step S1333, based on the order of the node access sequence and the inspection node safety parameter set, a candidate path set consisting of multiple candidate flight paths is generated between two adjacent inspection nodes, and each candidate flight path satisfies that the obstacle avoidance buffer boundary range of the starting node and the obstacle avoidance buffer boundary range of the end node do not overlap, and the flight altitude is always not lower than the higher minimum safe flight altitude of the two nodes.

[0064] For example, in the above mountain highway section, there are two adjacent inspection nodes in the node access sequence, one is the bridge support structure inspection node, and the other is the road surface inspection node close to the bridge. For these two nodes, multiple candidate flight paths are generated between them according to their respective obstacle avoidance buffer boundary ranges. One of the candidate flight paths starts from the bridge support structure inspection node and flies to the road surface inspection node along a certain angle. This path must ensure that the obstacle avoidance buffer boundary range of the starting node (bridge support structure inspection node) and the obstacle avoidance buffer boundary range of the end node (road surface inspection node) do not overlap, and during the flight process, the flight altitude is always not lower than the higher minimum safe flight altitude of the two nodes. For example, the minimum safe flight altitude of the bridge support structure inspection node is 120 meters, and the minimum safe flight altitude of the road surface inspection node is 100 meters, then the flight altitude of this candidate flight path cannot be lower than 120 meters. According to this rule, multiple candidate flight paths are generated between all adjacent inspection nodes to form a candidate path set.

[0065] Step S1334, for each candidate flight path, calculate the comprehensive cost value of the candidate flight path, the comprehensive cost value is obtained by weighted summation of the flight distance cost, the data collection time cost and the safety risk cost, wherein the safety risk cost is calculated based on the overlap ratio of the candidate flight path and the spatial distribution density of disaster points in the historical disaster records.

[0066] Take a candidate flight path from a slope inspection node at one end of a mountain highway section to a bridge inspection node at the other end as an example. First, calculate the flight distance cost. Assuming that the length of the path is 500 meters, according to the set unit distance cost coefficient (for example, the cost per meter is 0.1 yuan), the flight distance cost is 50 yuan. For the data collection time cost, since the total data collection time of the inspection nodes passed by this path is 30 seconds, according to the time cost coefficient per second (for example, the cost per second is 0.5 yuan), the data collection time cost is 15 yuan. Looking at the safety risk cost, this candidate flight path has a certain overlap with the spatial distribution density of landslide events in historical disaster records. After detailed calculation, the overlap ratio is 30%. According to the calculation method of the safety risk cost (for example, setting a coefficient relationship proportional to the overlap ratio, and each 10% overlap ratio corresponds to a cost of 10 yuan), the safety risk cost is 30 yuan. Finally, according to the weighted summation method (assuming that the flight distance cost weight is 0.4, the data collection time cost weight is 0.3, and the safety risk cost weight is 0.3), the comprehensive cost value of this candidate flight path is calculated to be 36 yuan. In the same way, the comprehensive cost value of each candidate flight path in the candidate path set is calculated.

[0067] Step S1335, select the path with the lowest comprehensive cost value from the candidate path set as the reference path, rearrange the paths between adjacent nodes for nodes in the reference path whose data collection time exceeds a preset threshold, generate an optimized target path by exchanging the node order or inserting intermediate transition nodes, and re-evaluate the comprehensive cost value of the target path until the convergence condition is reached. In the currently optimized target path, dynamically match the flight speed and sensor trigger frequency of the drone according to the lower limit value of the flight altitude of each path segment in the target path and the obstacle avoidance buffer boundary range, wherein the flight speed is positively correlated with the minimum safe flight altitude of the path segment, and the sensor trigger frequency is negatively correlated with the area of ​​the obstacle avoidance buffer boundary range.

[0068] Assuming that the reference path selected from the candidate path set has a data collection node with a data collection time of 20 seconds, and the preset threshold is 15 seconds, then it is necessary to rearrange the paths between adjacent nodes of this node. If the adjacent node is a node with a data collection time of 10 seconds and a lower priority, the comprehensive cost value of this new path is re-evaluated by exchanging the order of the two nodes. In this process, the node order is continuously adjusted or intermediate transition nodes are inserted, and the comprehensive cost value is recalculated after each adjustment until the comprehensive cost value no longer decreases and the convergence condition is met. In the optimized target path, for the path segment with a higher flight altitude lower limit, since the flight speed is positively correlated with the minimum safe flight altitude, the drone can fly at a relatively high speed in this path segment, for example, in the path segment with a minimum safe flight altitude of 150 meters, the flight speed is set to 20 meters / second; and for the area with a smaller obstacle avoidance buffer boundary range, since the sensor trigger frequency is negatively correlated with the area of ​​the obstacle avoidance buffer boundary range, the sensor trigger frequency in this area should be increased, for example, in the area with a radius of 3 meters of the obstacle avoidance buffer boundary range, the sensor trigger frequency is set to 5 times per second to ensure accurate data collection.

[0069] Step S1336, detect whether there is a target area in the target path where the flight direction turning angles of a preset number of consecutive inspection nodes exceed the maximum steering capability of the drone. If so, insert an auxiliary correction node in the target area. The position of the auxiliary correction node is determined according to the intersection of the extended tangents of the obstacle avoidance buffer boundary range, and recalculate the comprehensive cost value of the path after inserting the auxiliary correction node until the target path optimization is completed and the initial flight path is generated.

[0070] On the target path of a complex section of a mountain highway, assuming that the preset number is 3 inspection nodes, it is found that the flight direction turning angles of 3 consecutive inspection nodes exceed the maximum steering capability of the drone. At this time, the position of the auxiliary correction node is determined according to the intersection of the extended tangents of the obstacle avoidance buffer boundary range of these 3 inspection nodes, and the auxiliary correction node is inserted at this intersection. After insertion, the comprehensive cost value of the path containing the auxiliary correction node is recalculated, and then the target path is continuously checked to see if there are similar problem areas. If so, the auxiliary correction nodes are continuously inserted and the comprehensive cost value is recalculated until the target path optimization is completed and the initial flight path is generated.

[0071] Step S1337, obtaining real-time wind speed, rainfall intensity and visibility data of the target mountainous road section.

[0072] For example, meteorological monitoring equipment is installed on mountain highway sections, which can accurately obtain real-time meteorological information of the section. For example, the meteorological monitoring equipment shows that the real-time wind speed is 15 meters per second, the rainfall intensity is 50 millimeters per hour, and the visibility is 50 meters.

[0073] Step S1338, if the real-time wind speed exceeds the first preset threshold, the flight speed of the UAV is reduced according to the preset deceleration strategy and the obstacle avoidance buffer distance is increased according to the preset obstacle avoidance buffer strategy; if the rainfall intensity exceeds the second preset threshold, the anti-interference mode of the UAV is enabled, and the sensor trigger frequency is increased according to the first preset compensation parameter to compensate for the loss of data collection accuracy; if the visibility is lower than the third preset threshold, the infrared imaging module of the UAV is activated to replace the visible light camera, and the data collection time of the corresponding inspection node is extended according to the second preset compensation parameter.

[0074] If the real-time wind speed exceeds the first preset threshold, for example, the first preset threshold is set to 10 meters / second, since the real-time wind speed of 15 meters / second exceeds the threshold, the flight speed of the drone is reduced according to the preset deceleration strategy and the obstacle avoidance buffer distance is increased according to the preset obstacle avoidance buffer strategy. Assuming that the original flight speed of the drone is 20 meters / second, according to the preset deceleration strategy (for example, the flight speed is reduced by 1 meter / second for every wind speed exceeding 1 meter / second), the flight speed is reduced to 15 meters / second. At the same time, the original obstacle avoidance buffer distance is 5 meters, and according to the preset obstacle avoidance buffer strategy (for example, the obstacle avoidance buffer distance is increased by 1 meter for every wind speed exceeding 1 meter / second), the obstacle avoidance buffer distance is increased to 10 meters.

[0075] If the rainfall intensity exceeds the second preset threshold, assuming that the second preset threshold is 30 mm / hour, since the rainfall intensity of 50 mm / hour exceeds the threshold, the anti-interference mode of the drone is enabled, and the sensor trigger frequency is increased according to the first preset compensation parameter to compensate for the loss of data collection accuracy. For example, the first preset compensation parameter is to increase the sensor trigger frequency by 2 times / second for every rainfall intensity exceeding 10 mm / hour. The original sensor trigger frequency is 3 times / second, so now the sensor trigger frequency is increased to 7 times / second to ensure that more accurate data can be collected in a rainy environment.

[0076] If the visibility is lower than the third preset threshold, assuming that the third preset threshold is 100 meters, since the visibility is 50 meters lower than the threshold, the infrared imaging module of the drone is activated to replace the visible light camera, and the data collection time of the corresponding inspection node is extended according to the second preset compensation parameter. For example, the second preset compensation parameter is to extend the data collection time by 5 seconds for every visibility lower than 50 meters. For an inspection node whose original data collection time is 10 seconds, its data collection time is now extended to 15 seconds, so that sufficient monitoring data can be obtained in a low visibility environment.

[0077] Step S1339, predicting the flight risk in the future time window according to the changing trend of the meteorological data. If the predicted risk level exceeds the critical value, a path replanning instruction is sent to the navigation control unit of the UAV.

[0078] For example, based on the historical change trends and current data of meteorological data, it is predicted that the wind speed will continue to increase and the rainfall intensity will increase in the next hour. The risk level calculated according to the risk assessment model will exceed the critical value. At this time, a path replanning instruction is sent to the navigation control unit of the drone to ensure that the drone can fly safely and effectively and collect data in subsequent inspection tasks.

[0079] In a possible implementation, step S140 includes: Step S141, scanning the surface deformation area by using the laser radar sensor of the drone to generate a deformation displacement vector map as the surface deformation data.

[0080] Step S142, using an optical camera to capture a road bridge surface image, and using an edge detection algorithm to extract crack length, width and direction features corresponding to the road bridge surface image as structured data of the road bridge crack image.

[0081] For example, when a drone flies over a bridge on a mountain highway, an optical camera captures an image of the bridge surface. The bridge surface image captured by the camera contains various information about the bridge surface, including possible defects such as cracks. Then, the captured image is processed by an edge detection algorithm. The edge detection algorithm can identify the edges of objects in the bridge surface image, and for cracks on the bridge surface, the boundaries of the cracks can be accurately determined. By analyzing these boundary information, the length of the crack is calculated, for example, the distance from one end of the crack to the other end is measured to be several centimeters. At the same time, the width of the crack can also be determined. By analyzing the pixels of the crack edge and combining the known image scale, the width value of the crack can be obtained. In addition, according to the direction of the crack in the image, its direction characteristics on the actual bridge surface are determined, such as extending along the horizontal or longitudinal direction of the bridge. These crack length, width and direction characteristics constitute the structured data of the road bridge crack image.

[0082] Step S143, calling the multi-band radar of the UAV to continuously monitor the slope and capture the slope displacement trajectory of the rock and soil body inside the slope, wherein the slope displacement trajectory includes a displacement rate curve and a sliding direction index.

[0083] In the slope area of ​​the mountain highway section, the multi-band radar continuously transmits radar waves to the slope and receives reflected waves. Due to the movement of the rock and soil inside the slope, the reflected waves will change. By analyzing these reflected waves, the multi-band radar can capture the displacement information of the rock and soil inside the slope. As time goes by, this displacement information is continuously collected, and a displacement rate curve can be drawn. For example, over a period of time, the displacement rate of the rock and soil at a certain depth may start low and then gradually increase. This change process is reflected in the displacement rate curve. At the same time, based on information such as the change in the direction of the radar wave reflection, the sliding direction index of the rock and soil can be determined, such as whether it slides toward the road or toward the inside of the hillside. The above displacement rate curve and sliding direction index constitute the slope displacement trajectory.

[0084] Step S144, synchronously collect environmental temperature and humidity, wind speed and direction, and vibration frequency parameters as auxiliary verification information of the multi-dimensional monitoring data.

[0085] Step S145, aligning the surface deformation data, the road and bridge crack images and the slope displacement trajectory in time series, binding them with the corresponding geographic coordinates and storing them in a distributed database.

[0086] During the entire inspection process, the temperature and humidity sensors, wind speed and direction meters, and vibration sensors carried by the drone work simultaneously. In the inspection sub-area of ​​the mountain highway, the temperature and humidity sensors can accurately measure the temperature and humidity of the environment. For example, at a certain moment, the ambient temperature is measured to be 25 degrees Celsius and the humidity is 60%. The wind speed and direction meter can detect the current wind speed and wind direction, such as the wind speed is 10 meters per second and the wind direction is southeast. The vibration sensor can detect the vibration frequency of the drone itself or the surrounding environment. Assume that the detected vibration frequency is 100 Hz. These environmental temperature and humidity, wind speed and direction, and vibration frequency parameters are aligned with the surface deformation data, road and bridge crack images, and slope displacement trajectories in time series. For example, when the surface deformation data at a certain moment is collected, the environmental parameters at that moment are also recorded to ensure that all data are corresponding in time. Then these data are bound to the corresponding geographic coordinates. For example, a certain surface deformation data corresponds to the coordinates of a specific section of a mountain highway. These bound data are stored in a distributed database for subsequent query, analysis, and processing.

[0087] In a possible implementation, step S141 includes: Step S1411, continuously emit laser pulses during the flight of the drone, receive reflected signals and calculate the pulse round-trip time difference to generate an initial point cloud data set.

[0088] In a certain inspection sub-area of ​​a mountain highway, there is a surface area where there are signs of a slight landslide. When the drone flies over the area, the lidar continuously emits laser pulses during the flight. These laser pulses are reflected back after being shot at the surface. The sensor receives the reflected signal and accurately calculates the round-trip time difference of the pulse. For example, a laser pulse returns after a period of time after being emitted. According to the propagation speed of the laser in the air, the distance information from the monitoring point corresponding to the pulse to the drone can be obtained by calculating this round-trip time difference. By emitting and receiving multiple laser pulses in the entire surface deformation area, a large amount of such distance information is collected to generate an initial point cloud data set.

[0089] Step S1412, denoising the initial point cloud data, removing abnormal points caused by vegetation occlusion or meteorological interference, and generating denoised point cloud data.

[0090] In this embodiment, this initial point cloud data set contains information on a large number of points in the surface area, but there may be some abnormal points. These abnormal points may be caused by vegetation obstruction or meteorological interference. For example, there are some tall trees in the area, and the laser pulse may be scattered or reflected multiple times when passing through the branches and leaves of the trees, resulting in inaccurate calculated distance information. In addition, under some meteorological conditions, such as fog or rain, the propagation of the laser pulse will also be interfered. Therefore, it is necessary to denoise the initial point cloud data to remove these abnormal points caused by vegetation obstruction or meteorological interference, thereby generating denoised point cloud data.

[0091] Step S1413, performing elevation correction on the denoised point cloud data based on the terrain slope matrix to obtain current point cloud data, thereby eliminating the error effect of terrain undulation on deformation calculation.

[0092] Step S1414, using a time difference algorithm to compare the current point cloud data with the historical benchmark data, and calculating the three-dimensional displacement and displacement direction of each monitoring point.

[0093] Step S1415: Mark the area where the displacement exceeds the warning threshold as a risk area, and distinguish the risk level of the risk area by color gradient in the deformation displacement vector map.

[0094] In detail, the historical benchmark data is the data previously collected and processed in the same area, which serves as a benchmark for comparison. By comparing the coordinate information of each monitoring point in the current point cloud data and the historical benchmark data, the three-dimensional displacement can be calculated. For example, the coordinates of a monitoring point in the current data have certain differences in the x, y, and z directions compared with the coordinates in the historical benchmark data. This difference is the three-dimensional displacement of the monitoring point. At the same time, the displacement direction can be determined according to the change trend of the coordinates, such as towards a specific geographical direction. If the displacement of a certain area exceeds the warning threshold, the area is marked as a risk area, and the risk level of the risk area is distinguished by color gradient in the deformation displacement vector diagram. For example, areas with smaller displacements are represented by light blue, areas with larger displacements and close to the danger value are represented by yellow, and areas with displacements exceeding the danger value by more are represented by red, thus generating a deformation displacement vector diagram as surface deformation data.

[0095] For example, step S1413 includes: Step S1413-1, obtaining denoised point cloud data, and extracting three-dimensional coordinate information of each monitoring point in the denoised point cloud data, wherein the three-dimensional coordinate information includes a horizontal coordinate and a vertical elevation value.

[0096] Step S1413-2, retrieve the terrain slope matrix, the terrain slope matrix contains the terrain slope distribution information of the target mountainous road section, the terrain slope distribution information is composed of a plurality of slope units, each slope unit corresponds to a geographical area.

[0097] Step S1413-3, spatially matching the horizontal coordinates of each monitoring point in the denoised point cloud data with the slope units in the terrain slope matrix to determine the slope unit corresponding to each monitoring point.

[0098] Step S1413-4, calculating the terrain undulation correction value of each monitoring point according to the terrain slope distribution information of the slope unit, wherein the terrain undulation correction value is used to eliminate the influence of the terrain slope on the vertical elevation value.

[0099] Step S1413-5, applying the terrain undulation correction amount to the vertical elevation value of each monitoring point in the denoised point cloud data to generate a corrected vertical elevation value, wherein the corrected vertical elevation value eliminates the error effect of terrain undulation on deformation calculation.

[0100] Step S1413-6, recombining the corrected vertical elevation value with the horizontal coordinate in the denoised point cloud data to generate corrected current point cloud data.

[0101] In this embodiment, although the denoised point cloud data is relatively accurate, the existence of terrain undulation may cause errors in deformation calculation. In order to eliminate this effect, it is necessary to perform elevation correction on the denoised point cloud data based on the terrain slope matrix. First, the denoised point cloud data is obtained, and the three-dimensional coordinate information of each monitoring point is extracted. This three-dimensional coordinate information includes horizontal coordinates and vertical elevation values. Then the terrain slope matrix obtained previously is retrieved. The terrain slope matrix contains the terrain slope distribution information of the target mountainous road section. This information is composed of multiple slope units, and each slope unit corresponds to a geographical area. The horizontal coordinates of each monitoring point in the denoised point cloud data are spatially matched with the slope units in the terrain slope matrix to determine the slope unit corresponding to each monitoring point. For example, if the horizontal coordinates of a certain monitoring point are within a certain slope unit range, then it is determined that the monitoring point corresponds to this slope unit. Then, according to the terrain slope distribution information of this slope unit, the terrain undulation correction amount of each monitoring point is calculated. This terrain undulation correction amount is specifically used to eliminate the influence of terrain slope on vertical elevation value. For example, if the slope of a slope unit is large, the vertical elevation value of the corresponding monitoring point may have a large deviation due to the terrain slope. The terrain undulation correction calculated can correct this deviation. The terrain undulation correction is applied to the vertical elevation value of each monitoring point in the denoised point cloud data, thus generating a corrected vertical elevation value, which has eliminated the error effect of terrain undulation on deformation calculation. Finally, the corrected vertical elevation value is recombined with the horizontal coordinates in the denoised point cloud data to generate the corrected current point cloud data.

[0102] In a possible implementation, step S150 includes: Step S151, calling the risk warning model, matching the displacement rate in the surface deformation data with the landslide events in the historical disaster records, and triggering a first-level warning signal if the displacement rate continues to increase and the direction is consistent with the historical landslide direction.

[0103] When analyzing surface deformation data, we focus on the displacement rate and match it with landslide events in historical disaster records. For example, the surface deformation data of a section of a mountain highway shows that the displacement rate in a specific area shows a trend of continuous increase. Through the query of historical disaster records in the area, it is known that a landslide event has occurred here before, and the displacement direction monitored this time is consistent with the historical landslide direction. This indicates that there is an extremely high risk of landslides in the area, which meets the conditions for triggering a first-level warning signal. This situation may be due to the relatively fragile geological structure of the area, coupled with the recent impact of factors such as rain erosion or groundwater level changes, resulting in a gradual decrease in the stability of the surface soil, and then the displacement change characteristics similar to historical landslides appear.

[0104] And, step S152, analyzing the trend characteristics of the road bridge crack image, if the crack extension direction coincides with the main stress direction of the bridge and the width expansion rate exceeds a critical value, a secondary warning signal is triggered.

[0105] During the inspection of highway bridges in mountainous areas, the images of road and bridge cracks captured by optical cameras are processed to obtain relevant characteristic information of the cracks. If it is found that the extension direction of a crack coincides with the direction of the main stress of the bridge, this means that the development direction of the crack is consistent with the direction in which the bridge structure is most unfavorable, posing a serious threat to the structural safety of the bridge. At the same time, if the crack width expansion rate exceeds the critical value, for example, during several consecutive inspections, it is found that the crack width increases at a faster rate, exceeding the critical value set according to the bridge structure design and safety standards, this indicates that the damage to the bridge structure is accelerating, and a secondary warning signal will be triggered at this time. The development of such cracks may be caused by factors such as long-term vehicle loads, temperature changes, or material aging, and the coincidence of cracks with the main stress direction will accelerate the destruction of the bridge structure.

[0106] And, step S153, according to the sliding direction and speed of the slope displacement trajectory, combined with the real-time rainfall intensity data, the slope stability coefficient is calculated. If the slope stability coefficient is lower than the safety threshold, a third-level warning signal is triggered.

[0107] Step S154, generating differentiated handling suggestion information based on the warning signal level, wherein the handling suggestion information includes emergency closure of road sections, speed limit or initiation of support structure reinforcement plan.

[0108] For example, when a Level 1 warning signal is triggered, due to the extremely high risk of landslides, emergency road closure measures need to be taken to ensure driving safety and road facility safety on mountain highways. This means immediately preventing vehicles from entering the dangerous area, setting up roadblocks and warning signs, and notifying relevant departments for emergency treatment, such as geological surveys and reinforcement of the landslide area.

[0109] If the second-level warning signal is triggered, it indicates that there are major safety hazards in the bridge structure, but it has not yet reached the level of immediate traffic interruption. At this time, the recommended information for speed limit is adopted. Speed ​​limit signs are set up at both ends of the bridge to reduce the speed of vehicles passing through the bridge and the impact of vehicle loads on the bridge. At the same time, professionals are arranged to conduct further detailed inspections and assessments of the bridge and formulate maintenance and reinforcement plans.

[0110] When the third-level warning signal is triggered, the support structure reinforcement plan is initiated to address the slope stability issue. This may include strengthening the retaining wall of the slope, increasing the number and length of anchor rods or anchor cables, and protecting the slope surface to improve the stability of the slope and prevent slope landslide accidents.

[0111] Step S155: Push the risk level signal and disposal suggestion information to the road and bridge management terminal, and associate the corresponding monitoring data traceability link for manual review.

[0112] For example, after receiving information that the risk level of a certain road section is level 1 and the disposal recommendation is to close the road section urgently, the road and bridge management terminal can also view detailed monitoring data through the associated monitoring data traceability link, such as the displacement rate change curve and displacement direction information in the surface deformation data, as well as relevant records of historical landslide events. In this way, road and bridge management personnel can manually review the warning information and disposal recommendations to ensure the accuracy and rationality of the decision, and if necessary, adjust or supplement the disposal recommendations according to the actual situation.

[0113] In a possible implementation, step S153 includes: Step S1531, performing vector decomposition on the sliding direction of the slope displacement trajectory to generate time series data of horizontal displacement components and vertical displacement components.

[0114] For example, in the monitoring of a slope on a mountain highway, the multi-band radar captured the displacement trajectory of the rock and soil inside the slope, and through vector decomposition technology, the displacement changes in the horizontal and vertical directions over a period of time were obtained. The horizontal displacement component reflects the lateral movement trend of the rock and soil in the slope plane, and the vertical displacement component reflects the settlement or rise trend of the rock and soil.

[0115] Step S1532, performing similarity matching between the horizontal displacement component and the landslide event displacement pattern of the corresponding geographic coordinates in the historical disaster records, and screening out a reference event set whose displacement trend similarity exceeds a matching threshold.

[0116] In the historical disaster records of mountain highways, there are detailed landslide event records for each slope area, including the displacement pattern of the rock and soil when the landslide occurred. By comparing the horizontal displacement component of the current slope with the displacement pattern in the historical records, a set of reference events with displacement trend similarity exceeding the matching threshold is screened out. This matching threshold is determined based on a large amount of historical data and engineering experience to ensure that the screened reference events have a high reference value.

[0117] Step S1533: extracting the rainfall intensity time series data before each landslide event in the reference event set, and establishing a correlation mapping table between rainfall accumulation and displacement acceleration interval.

[0118] The rainfall intensity time series data records the changes in rainfall intensity in the period before the landslide event. By analyzing these data, a correlation mapping table between rainfall accumulation and displacement acceleration interval is established. For example, in a certain reference landslide event, it is found that when the rainfall accumulation reaches a certain value, the displacement rate of the slope rock and soil body begins to accelerate significantly and enters the displacement acceleration interval. By analyzing multiple reference events, the relationship between different rainfall accumulations and displacement acceleration intervals can be established.

[0119] Step S1534, calculating the current rainfall accumulation according to the real-time rainfall intensity data, and searching the association mapping table for a critical rainfall threshold that overlaps with the current displacement acceleration interval.

[0120] During the inspection process, the real-time rainfall intensity data is obtained through the meteorological monitoring equipment installed on the mountain highway section, and the current rainfall accumulation is calculated at a certain time interval. Then, the critical rainfall threshold that overlaps with the current displacement acceleration interval is searched in the association mapping table. Assuming that the current slope displacement is in a certain acceleration interval, the corresponding critical rainfall threshold is 100 mm when the association mapping table is searched.

[0121] Step S1535, generating a dynamic correction factor of the slope stability coefficient based on the ratio of the change rate of the vertical displacement component to the critical rainfall threshold.

[0122] For example, if the rate of change of the vertical displacement component is 5 mm / hour and the critical rainfall threshold is 100 mm, then the dynamic correction factor is 5 / 100 = 0.05. This dynamic correction factor reflects the degree of change of the vertical displacement of the slope under the current rainfall conditions relative to the critical rainfall conditions.

[0123] Step S1536, weighted fusion of the dynamic correction factor and the normalized value of the slope support structure integrity parameter is performed to output a comprehensive slope stability coefficient.

[0124] Assuming that the normalized value of the slope support structure integrity parameter after evaluation is 0.8, a weighted fusion calculation is performed according to the set weighting coefficient (for example, the dynamic correction factor weight is 0.3, and the slope support structure integrity parameter weight is 0.7), and the comprehensive slope stability coefficient is 0.8×0.7 + 0.05×0.3 = 0.575. If this comprehensive slope stability coefficient is lower than the safety threshold, a level 3 warning signal is triggered. This indicates that under the current slope state, due to the influence of factors such as rainfall and rock and soil displacement, the stability of the slope is already in a dangerous range, and corresponding measures need to be taken.

[0125] In a possible implementation, the method further includes: Step S210, establishing a feedback closed-loop mechanism for the multi-dimensional monitoring data and disposal suggestion information.

[0126] Step S220, when the treatment suggestion information is adopted and executed by the road and bridge management terminal, the structural response data of the corresponding area is continuously monitored, and the structural response data is compared and analyzed with the expected treatment effect. If the comparison and analysis result indicates that the actual improvement rate is lower than the set ratio of the expected value, the treatment plan optimization instruction is triggered.

[0127] For example, after a certain slope is treated with a support structure reinforcement solution, the deformation data of the reinforcement material and the stress distribution data of the support structure in the treatment area are collected simultaneously when the drone performs the inspection task. In the reinforced slope area, the deformation data of the reinforcement material, such as the elongation of the anchor rod or the change in the inclination of the retaining wall, are obtained through the sensors installed in the reinforcement material. At the same time, the stress distribution data of the support structure is obtained through the stress sensor installed in the support structure to understand the size and distribution of the internal stress of the support structure.

[0128] These structural response data are then compared and analyzed with the expected treatment effect. The expected treatment effect is pre-set according to the treatment plan. For example, it is expected that the displacement rate of the slope should be reduced to a certain safe range within a certain period of time, the deformation of the reinforcement material should be kept within the design allowable range, and the stress distribution of the support structure should comply with the principles of structural mechanics. If the comparative analysis results indicate that the actual improvement rate is lower than the set ratio of the expected value, for example, the slope displacement rate is expected to decrease by 80%, but it is actually only reduced by 50%, then the treatment plan optimization instruction is triggered.

[0129] Step S230, after adjusting the parameter weights of the risk warning model based on the disposal plan optimization instructions and updating the emergency avoidance node configuration in the path planning model, the optimized model parameters and configuration information are synchronized to the edge computing units of all in-service drones to realize online iterative upgrades of the monitoring strategy.

[0130] For example, if it is found that the current disposal plan has a poor control effect on slope displacement, it may be that the weight setting of the relationship between slope displacement rate and rainfall intensity in the risk warning model is unreasonable. Adjust the weights of relevant parameters in the risk warning model, such as increasing the weight of the impact of rainfall intensity on slope stability. At the same time, since the unstable state of the slope may affect the inspection path planning of the drone, update the emergency avoidance node configuration in the path planning model, such as adding more emergency avoidance nodes near the slope or adjusting the position of existing avoidance nodes. Then synchronize these optimized model parameters and configuration information to the edge computing units of all in-service drones, so that drones can perform more effective inspections according to the new monitoring strategy in subsequent inspection tasks.

[0131] For example, the implementation of the feedback closed loop mechanism includes: Step S211, when the UAV performs the inspection task, the deformation data of the reinforcement materials and the stress distribution data of the supporting structure in the treatment area are synchronously collected.

[0132] Step S212, by comparing the difference in slope displacement trajectories before and after the treatment, the effective action radius and attenuation coefficient of the treatment measure corresponding to the treatment suggestion information are calculated.

[0133] Step S213: If the implemented attenuation coefficient exceeds the standard deviation of a preset multiple of the historical average value, it is determined that the current disposal plan has a design defect, and an optimization suggestion for material replacement or structural redesign is generated.

[0134] Step S214: input the optimization suggestion into the path planning model to increase the monitoring frequency of the defective area and the sensor accuracy level.

[0135] And, in step S215, after the new treatment plan is implemented, the feedback closed-loop mechanism is restarted until the actual improvement rate reaches the target.

[0136] For example, the slope displacement trajectory data are obtained before and after the slope is reinforced. By comparison, it is found that the slope displacement is well controlled within a certain range from the reinforced structure. This range is the effective radius of the treatment measures. As time goes by, the control effect of the treatment measures on the slope displacement may gradually weaken. The attenuation coefficient is calculated by analyzing the changing trend of the displacement trajectory data. If the implemented attenuation coefficient exceeds the standard deviation of a preset multiple of the historical average, for example, the historical average attenuation coefficient is 0.1, the standard deviation is 0.02, and the currently calculated attenuation coefficient is 0.15, which exceeds 1.5 times the standard deviation, it is determined that the current treatment plan has design defects, and optimization suggestions for material replacement or structural redesign are generated.

[0137] Input the optimization suggestions into the path planning model to increase the monitoring frequency and sensor accuracy level of the defective area. For example, if it is determined that the problem is with the reinforcement material, the optimization suggestions for material replacement are input into the path planning model, so that the drone can increase the monitoring frequency of the reinforcement materials in the area in subsequent inspections, such as from once a day to three times a day. At the same time, improve the accuracy level of the sensor to more accurately obtain the deformation data of the reinforcement material and the stress distribution data of the support structure. When the new disposal plan is implemented, restart the feedback closed-loop mechanism until the actual improvement rate reaches the standard, to ensure that the slopes of the mountain highway section are in a safe and stable state, and to ensure the safe operation of the entire road.

[0138] Figure 2 The schematic diagram shows exemplary hardware and software components of the server 100 that can implement the concept of the present application provided by some embodiments of the present application. For example, the processor 120 can be used on the server 100 and used to execute the functions in the present application.

[0139] The server 100 can be a general-purpose server or a special-purpose server, both of which can be used to implement the inspection data monitoring method for mountain road and bridge early warning of the present application. Although only one server is shown in the present application, for convenience, the functions described in the present application can be implemented in a distributed manner on multiple similar platforms to balance the processing load.

[0140] For example, the server 100 may include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and storage media 140 in different forms, such as a disk, ROM, or RAM, or any combination thereof. Exemplarily, the server 100 may also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. The method of the present application can be implemented according to these program instructions. The server 100 also includes an input / output (I / O) interface 150 between the computer and other input / output devices.

[0141] For ease of explanation, only one processor is described in the server 100. However, it should be noted that the server 100 in the present application may also include multiple processors, so the steps performed by one processor described in the present application may also be performed jointly or individually by multiple processors. For example, if the processor of the server 100 performs steps A and B, it should be understood that steps A and B may also be performed jointly by two different processors or individually in one processor. For example, the first processor performs step A, the second processor performs step B, or the first processor and the second processor perform steps A and B together.

[0142] In addition, an embodiment of the present invention further provides a readable storage medium, in which computer executable instructions are preset. When a processor executes the computer executable instructions, the inspection data monitoring method applied to mountain road and bridge early warning as described above is implemented.

[0143] It should be noted that in order to simplify the description of the present invention and thus help understand one or more embodiments of the invention, in the foregoing description of the embodiments of the present invention, various features are sometimes combined into one embodiment, drawing or description thereof.

Claims

1. A patrol data monitoring method for early warning of roads and bridges in mountainous areas, characterized in that: The method comprises: Acquire remote sensing monitoring data of roads and bridges in target mountainous road sections, wherein the remote sensing monitoring data of roads and bridges includes current terrain distribution characteristics, road and bridge structural status parameters, and historical disaster records; Based on the current terrain distribution characteristics, calling the terrain feature extraction model to perform regional segmentation on the target mountainous road section, and generating multiple inspection sub-areas and corresponding terrain feature labels; According to the terrain feature labels and the road and bridge structure state parameters, a dynamic flight path of the UAV in the inspection sub-area is generated through a path planning model, wherein the dynamic flight path includes a plurality of inspection nodes and flight parameters between the inspection nodes; Based on the dynamic flight path, the UAV is controlled to perform real-time inspection tasks, and multi-dimensional monitoring data in the inspection sub-area is collected, wherein the multi-dimensional monitoring data includes surface deformation data, road and bridge crack images, and slope displacement trajectories; The multi-dimensional monitoring data and the historical disaster records are input into a risk warning model to generate a risk level signal and corresponding handling suggestion information for the target mountainous road section.

2. The inspection data monitoring method for early warning of mountainous roads and bridges according to claim 1 is characterized in that: The acquisition of the road and bridge remote sensing monitoring data of the target mountainous road section includes: Acquire multispectral image data and three-dimensional point cloud data of the target mountainous road section through satellite remote sensing equipment; Performing ground object classification processing on the multispectral image data, extracting vegetation coverage density, rock exposed area and hydrological distribution characteristics as sub-features of the current terrain distribution characteristics; Performing elevation difference processing on the three-dimensional point cloud data to generate a terrain slope matrix and a surface undulation index, and associating the terrain slope matrix and the surface undulation index with corresponding inspection sub-areas; Retrieving the historical maintenance database of the target mountainous road section, extracting the pier inclination angle, road surface settlement rate and slope support structure integrity parameters from the road and bridge structure status parameters; The historical disaster records are classified according to disaster types, including landslide events, bridge crack expansion events and roadbed collapse events, and the geographical coordinates and occurrence timestamps corresponding to each event are associated.

3. The inspection data monitoring method for early warning of mountainous roads and bridges according to claim 1 is characterized in that: The generating a dynamic flight path of the UAV in the inspection sub-area through a path planning model according to the terrain feature labels and the road and bridge structure state parameters includes: According to the vegetation coverage density and terrain slope matrix in the terrain feature label, the minimum safe flight altitude and obstacle avoidance buffer distance of the UAV in the corresponding inspection sub-area are calculated; Based on the pier inclination angle and slope support structure integrity parameters in the road and bridge structure status parameters, determine the priority order of the inspection nodes and the data collection time of each inspection node; Based on the minimum safe flight altitude and obstacle avoidance buffer distance of the drone in the corresponding inspection sub-area, as well as the priority order of the inspection nodes and the data collection time of each inspection node, a path optimization algorithm is used to spatially sort the inspection nodes, generate an initial flight path, and dynamically adjust the flight parameters based on real-time meteorological data, including flight speed, hovering time and sensor trigger frequency; Inserting emergency avoidance nodes into the initial flight path to generate an adjusted flight path, wherein the emergency avoidance nodes are adaptively configured according to the spatial distribution density of disaster types in historical disaster records; Integrate the adjusted flight path and the flight parameters into the dynamic flight path, and synchronize the dynamic flight path to the navigation control unit of the UAV, preset a path deviation detection mechanism in the navigation control unit, and trigger local path replanning when the lateral offset between the actual flight trajectory of the UAV and the dynamic flight path continues to exceed a preset ratio of the obstacle avoidance buffer boundary range, and return to execute the step of spatially sorting the inspection nodes using a path optimization algorithm to generate an initial flight path only for the remaining nodes that have not been visited; Wherein, the minimum safe flight altitude H_min=H_base+α×V+β×S̄; The obstacle avoidance buffer distance D_buffer=γ×(V / V_max)+δ×(ΔS / S_ref); The priority of the inspection node P = θ × w1 + (1-I) × w 2； The data collection duration T=T_base×(1+P / P_max); H_base is the benchmark safety height, α is the weight coefficient of vegetation coverage density V, β is the correction coefficient of the average slope S̄ of the terrain slope matrix, γ is the influencing factor of vegetation density, V_max is the maximum vegetation coverage density in the region, δ is the adjustment coefficient of the slope change rate ΔS, S_ref is the slope reference threshold, θ is the inclination angle of the pier, I is the normalized value of the integrity parameter of the slope support structure, w1 and w2 are the weight coefficients of the inclination angle and integrity respectively, T_base is the basic collection time, and P_max is the maximum priority value.

4. The inspection data monitoring method for early warning of mountainous roads and bridges according to claim 3 is characterized in that: The step of spatially sorting the inspection nodes using a path optimization algorithm based on the minimum safe flight altitude and obstacle avoidance buffer distance of the drone in the corresponding inspection sub-area, the priority order of the inspection nodes and the data collection time of each inspection node to generate an initial flight path includes: Based on the minimum safe flight altitude and obstacle avoidance buffer distance of each inspection sub-area, a corresponding flight altitude lower limit and obstacle avoidance buffer boundary range are assigned to each inspection node, and a set of inspection node safety parameters including altitude constraints and spatial obstacle avoidance ranges is generated; According to the priority order of the inspection nodes and the data collection time, the inspection nodes are sorted from high to low according to the priority, and when the priorities are the same, they are sorted from short to long according to the data collection time to form a node access sequence with time weight; Based on the order of the node access sequence and the inspection node safety parameter set, a candidate path set consisting of multiple candidate flight paths is generated between two adjacent inspection nodes, and each candidate flight path satisfies that the obstacle avoidance buffer boundary range of the starting node and the obstacle avoidance buffer boundary range of the end node do not overlap, and the flight altitude is always not lower than the higher minimum safe flight altitude of the two nodes; For each candidate flight path, a comprehensive cost value of the candidate flight path is calculated, wherein the comprehensive cost value is obtained by weighted summation of flight distance cost, data collection time cost and safety risk cost, wherein the safety risk cost is calculated according to the overlap ratio between the candidate flight path and the spatial distribution density of disaster points in historical disaster records; The path with the lowest comprehensive cost value is selected from the candidate path set as the reference path, and the paths between adjacent nodes are rearranged for nodes whose data collection time exceeds a preset threshold in the reference path, and an optimized target path is generated by exchanging the node order or inserting an intermediate transition node, and the comprehensive cost value of the target path is re-evaluated until the convergence condition is reached. In the currently optimized target path, the flight speed and sensor trigger frequency of the UAV are dynamically matched according to the flight altitude lower limit value and obstacle avoidance buffer boundary range of each path segment in the target path, wherein the flight speed is positively correlated with the minimum safe flight altitude of the path segment, and the sensor trigger frequency is negatively correlated with the area of ​​the obstacle avoidance buffer boundary range; Detect whether there is a target area in the target path where the flight direction turning angles of a preset number of consecutive inspection nodes exceed the maximum steering capability of the drone. If so, insert an auxiliary correction node in the target area, the position of the auxiliary correction node is determined according to the intersection of the extended tangent of the obstacle avoidance buffer boundary range, and recalculate the comprehensive cost value of the path after inserting the auxiliary correction node until the target path optimization is completed to generate the initial flight path; The dynamically adjusting the flight parameters based on real-time meteorological data includes: Obtain real-time wind speed, rainfall intensity and visibility data for target mountainous road sections; If the real-time wind speed exceeds the first preset threshold, the flight speed of the UAV is reduced according to the preset deceleration strategy and the obstacle avoidance buffer distance is increased according to the preset obstacle avoidance buffer strategy; If the rainfall intensity exceeds the second preset threshold, the anti-interference mode of the drone is enabled, and the sensor trigger frequency is increased according to the first preset compensation parameter to compensate for the loss of data collection accuracy; If the visibility is lower than the third preset threshold, the infrared imaging module of the drone is activated to replace the visible light camera, and the data collection time of the corresponding inspection node is extended according to the second preset compensation parameter; The flight risk in the future time window is predicted based on the changing trend of meteorological data. If the predicted risk level exceeds the critical value, a path replanning instruction is sent to the navigation control unit of the drone.

5. The inspection data monitoring method for early warning of mountainous roads and bridges according to claim 2 is characterized in that: The controlling of the UAV to perform real-time inspection tasks based on the dynamic flight path and collecting multi-dimensional monitoring data in the inspection sub-area includes: Scanning the surface deformation area by using a laser radar sensor of the drone to generate a deformation displacement vector map as the surface deformation data; Using an optical camera to capture a road bridge surface image, and using an edge detection algorithm to extract crack length, width, and direction features corresponding to the road bridge surface image as structured data of the road bridge crack image; The multi-band radar of the UAV is used to continuously monitor the slope and capture the slope displacement trajectory of the rock and soil inside the slope, which includes the displacement rate curve and the sliding direction index; Synchronously collect environmental temperature and humidity, wind speed and direction, and vibration frequency parameters as auxiliary verification information for the multi-dimensional monitoring data; The surface deformation data, road and bridge crack images and slope displacement trajectories are aligned in time series, bound with corresponding geographic coordinates and stored in a distributed database.

6. The inspection data monitoring method for early warning of mountainous roads and bridges according to claim 5 is characterized in that: The laser radar sensor of the drone is used to scan the surface deformation area to generate a deformation displacement vector diagram, including: The UAV continuously emits laser pulses during flight, receives reflected signals and calculates the round-trip time difference of the pulses to generate an initial point cloud data set; De-noising the initial point cloud data to remove abnormal points caused by vegetation occlusion or meteorological interference, thereby generating de-noised point cloud data; Based on the terrain slope matrix, the denoised point cloud data is subjected to elevation correction to obtain current point cloud data, thereby eliminating the error effect of terrain undulation on deformation calculation; A time series difference algorithm is used to compare the current point cloud data with the historical benchmark data to calculate the three-dimensional displacement and displacement direction of each monitoring point; The area where the displacement exceeds the warning threshold is marked as a risk area, and the risk level of the risk area is distinguished by color gradient in the deformation displacement vector diagram.

7. The inspection data monitoring method for early warning of mountainous roads and bridges according to claim 1 is characterized in that: The multi-dimensional monitoring data and the historical disaster records are input into the risk warning model to generate a risk level signal and corresponding disposal suggestion information for the target mountainous road section, including: Calling the risk warning model, matching the displacement rate in the surface deformation data with the landslide events in the historical disaster records, and triggering a first-level warning signal if the displacement rate continues to increase and the direction is consistent with the historical landslide direction; And, analyzing the trend characteristics of the road bridge crack image, if the crack extension direction coincides with the main stress direction of the bridge and the width expansion rate exceeds a critical value, a secondary warning signal is triggered; And, according to the sliding direction and speed of the slope displacement trajectory, combined with the real-time rainfall intensity data, the slope stability coefficient is calculated. If the slope stability coefficient is lower than the safety threshold, a third-level warning signal is triggered; Generate differentiated disposal suggestion information based on the warning signal level, including emergency road closure, speed limit or initiation of support structure reinforcement plan; The risk level signal and treatment suggestion information are pushed to the road and bridge management terminal, and the corresponding monitoring data traceability link is associated for manual review.

8. The inspection data monitoring method for early warning of mountainous roads and bridges according to claim 7 is characterized in that: The slope stability coefficient is calculated based on the sliding direction and speed of the slope displacement trajectory in combination with real-time rainfall intensity data, including: Performing vector decomposition on the sliding direction of the slope displacement trajectory to generate time series data of horizontal displacement components and vertical displacement components; Performing similarity matching between the horizontal displacement component and the landslide event displacement pattern of the corresponding geographic coordinates in the historical disaster records, and screening out a reference event set whose displacement trend similarity exceeds a matching threshold; Extracting the rainfall intensity time series data before each landslide event in the reference event set, and establishing a correlation mapping table between rainfall accumulation and displacement acceleration interval; Calculate the current rainfall accumulation according to the real-time rainfall intensity data, and search the association mapping table for a critical rainfall threshold value overlapping with the current displacement acceleration interval; generating a dynamic correction factor for the slope stability coefficient based on a ratio of the rate of change of the vertical displacement component to a critical rainfall threshold; The dynamic correction factor is weightedly fused with the normalized value of the slope support structure integrity parameter to output a comprehensive slope stability coefficient.

9. The inspection data monitoring method for early warning of mountainous roads and bridges according to claim 1 is characterized in that: The method further comprises: Establish a closed-loop feedback mechanism for the multi-dimensional monitoring data and disposal suggestion information; When the disposal suggestion information is adopted and executed by the road and bridge management terminal, the structural response data of the corresponding area is continuously monitored, and the structural response data is compared and analyzed with the expected disposal effect. If the comparative analysis result indicates that the actual improvement rate is lower than the set ratio of the expected value, the disposal plan optimization instruction is triggered; After adjusting the parameter weights of the risk warning model based on the disposal plan optimization instructions and updating the emergency avoidance node configuration in the path planning model, the optimized model parameters and configuration information are synchronized to the edge computing units of all in-service drones to realize online iterative upgrades of the monitoring strategy.

10. A server, characterized in that: The server includes a processor and a memory, the memory is connected to the processor, the memory is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the memory to implement the inspection data monitoring method for mountain road and bridge early warning as described in any one of claims 1 to 9 above.

Images