Soil erosion characteristic data processing system based on filtering algorithm

Abstract

The application relates to the technical field of geographic information systems, navigation and environment monitoring, in particular to a soil and water loss characteristic data processing system based on a filtering algorithm, which comprises a data acquisition module, an main navigation module, a shadow monitoring module and a residual error topology generation module; the data acquisition module collects real-time bottom sensor data streams of a mobile carrier through integrated inertial perception and positioning hardware; the main navigation module receives the bottom sensor data streams and outputs a current geographic information system space mileage; the shadow monitoring module receives the bottom sensor data streams and the current geographic information system space mileage, generates kinematic innovation residual errors on space domain discrete points; and the residual error topology generation module generates a disturbance characteristic map for correcting navigation parameters of the main navigation module based on navigation kinematic state abnormality evaluation results and the current geographic information system space mileage; the application realizes pure navigation type monitoring which does not depend on laser surveying or visual modeling.

Description

Technical Field

[0001] This invention relates to the fields of geographic information systems, navigation and environmental monitoring technologies, specifically a water and soil erosion characteristic data processing system based on filtering algorithms. Background Technology

[0002] Currently, soil and water loss monitoring in scenarios such as mountain orchards is usually achieved through laser mapping, visual modeling, or manual inspection. When inspection vehicles repeatedly travel along fixed operation routes, existing methods struggle to stably align and continuously compare surface disturbances of different periods under a unified geographic information system spatial coordinate system. This leads to inaccurate identification of navigation environment changes caused by local scour, subsidence, or soft zones, reducing the real-time performance and reliability of inspection and monitoring. Summary of the Invention

[0003] The purpose of this invention is to provide a water and soil erosion characteristic data processing system based on filtering algorithms, and to solve the following technical problems: This transforms changes in surface support caused by soil erosion into kinematic anomalies that can be identified, aligned, and cumulatively analyzed within the navigation link. This enables pure navigation-based monitoring that does not rely on laser mapping or visual modeling, while simultaneously ensuring stable tracking of inspection vehicles and the ability to identify abnormal locations.

[0004] The objective of this invention can be achieved through the following technical solutions: A soil erosion characteristic data processing system based on filtering algorithms, mounted on a mobile platform, includes: The data acquisition module is configured to: acquire the underlying sensor data stream of the mobile carrier in real time through integrated inertial sensing and positioning hardware; The main navigation module is configured to: receive data streams from the underlying sensors, perform time-domain updates using an error state Kalman filter configured with the first observation noise covariance to suppress terrain disturbances, generate a smooth navigation trajectory, and output the current geographic information system spatial mileage. The shadow monitoring module is configured to: receive the underlying sensor data stream and the current geographic information system spatial mileage, trigger the spatial domain state transition based on the current geographic information system spatial mileage, extract terrain disturbances using a filter configured with the second process noise covariance, and generate kinematic information residuals at discrete points in the spatial domain. The residual topology generation module is configured to: extract the feature components related to surface elevation and attitude from the kinematic innovation residuals to construct the current period kinematic perturbation feature vector; under the same geographic information system spatial coordinates, spatially align and differentially compare the current period kinematic perturbation feature vector with the historical baseline period kinematic perturbation feature vector of the same spatial segment obtained in advance to generate a navigation kinematic state anomaly assessment result; and based on the navigation kinematic state anomaly assessment result combined with the current geographic information system spatial mileage, map and generate a perturbation feature map for correcting the navigation parameters of the main navigation module.

[0005] Furthermore, the data acquisition module includes: An inertial measurement unit is configured to acquire triaxial acceleration and triaxial angular velocity data of a moving vehicle. Wheel-mounted odometer, configured to collect relative motion displacement data of a moving vehicle; A satellite navigation receiver, configured to collect absolute spatial coordinate data of a mobile vehicle; The underlying sensor data stream is generated by aggregating triaxial acceleration data, triaxial angular velocity data, relative motion displacement data, and absolute spatial coordinate data.

[0006] Furthermore, the main navigation module includes: The time-domain filtering unit is configured to update the state and measurement of the underlying sensor data stream by using a preset time interval as the trigger condition. The smoothing control unit is configured to use the first observation noise covariance to filter and suppress high-frequency transient inputs contained in the underlying sensor data stream to generate a smooth navigation trajectory; The mileage calculation unit is configured to calculate the current geographic information system spatial mileage of the mobile vehicle in the geographic information system coordinate system based on a smooth navigation trajectory.

[0007] Furthermore, the shadow detection module includes: The spatial domain triggering unit is configured to trigger the status update pipeline when the current geographic information system spatial mileage reaches a preset spatial interval. The spatial mapping unit is configured to convert the underlying sensor data stream from a time-domain integral model to a spatial-domain integral model based on a preset spatial interval. The high-sensitivity residual extraction unit is configured to amplify the high-frequency transient input using the second process noise covariance under the spatial domain integral model, calculate the deviation between the predicted state derived from the spatial domain integral model and the actual observation value provided by the underlying sensor data stream, and generate kinematic innovation residuals.

[0008] Furthermore, the shadow detection module also includes: The data resampling unit is configured to resample the underlying sensor data stream acquired based on discrete time into a data sequence based on equal spatial intervals using a spline interpolation algorithm before the spatial domain triggering unit is triggered, and then input the data sequence into the spatial mapping unit.

[0009] Furthermore, the residual topology generation module includes: The feature extraction unit is configured to separate the Z-axis acceleration component and pitch angle component of the moving carrier in the body coordinate system from the kinematic information residual. The vector splicing unit is configured to combine the Z-axis acceleration component with the pitch angle component to construct the kinematic perturbation feature vector for the current cycle. The differential comparison unit is configured to calculate the difference between the same components of the current periodic kinematic perturbation feature vector and the historical baseline periodic kinematic perturbation feature vector on a unified geographic information system spatial coordinate axis, and then perform dimensionless weighted fusion of the difference results to generate a one-dimensional risk intensity difference sequence.

[0010] Furthermore, the system also includes: The state determination module is configured to perform conditional determinations on the difference sequence based on a preset threshold. When the difference in the difference sequence is less than the preset safety threshold, it is determined that the navigation environment of the current spatial segment has not changed and has not affected the navigation accuracy, and the current inspection route of the mobile vehicle is maintained. When the difference in the difference sequence is greater than or equal to the safety threshold and less than the preset danger threshold, it is determined that there is a navigation environment disturbance in the current spatial road segment, and local trajectory deviation warning information in the navigation kinematic state abnormality assessment result is generated and recorded. When the difference in the difference sequence is greater than or equal to the danger threshold, it is determined that there is a risk of navigation failure in the current spatial segment, triggering an emergency avoidance command and replanning the subsequent geographic information system route of the mobile vehicle.

[0011] Furthermore, the system also includes: The cloud mapping module is configured to map the navigation kinematic state anomaly assessment results to a navigation environment disturbance intensity heatmap. The feature feedback module is configured to extract the current periodic kinematic perturbation feature vector and the navigation environment perturbation intensity heatmap at discrete points in the spatial domain, and upload them to a remote server.

[0012] Furthermore, the inertial measurement unit employs a six-axis microelectromechanical system (MEMS) inertial sensor; Wheel-type odometers use photoelectric encoders; The satellite navigation receiver uses a real-time dynamic differential global navigation satellite system module; The data acquisition module, main navigation module, and shadow detection module are all integrated into the navigation computer of the mobile vehicle.

[0013] The beneficial effects of this invention are: 1. This invention separates the main navigation smooth control from the shadow monitoring high-sensitivity residual extraction, without relying on laser or vision equipment; through the pure navigation link, the changes in surface support caused by soil erosion can be transformed into monitorable kinematic anomalies, which achieves lightweight and real-time environmental monitoring while ensuring the stable tracking of the mobile carrier. 2. This invention employs a state transition and data resampling mechanism based on spatial mileage triggering to transform time-domain data into a sequence with equal spatial intervals. This eliminates the comparison error caused by vehicle speed fluctuations in different inspection cycles, ensuring that historical and current data are precisely aligned at the same physical location, and significantly improving the comparison accuracy. 3. This invention extracts the Z-axis acceleration and pitch angle components from the kinematic information residuals to construct a feature vector and compares it with the benchmark difference; this method effectively eliminates the interference of single mechanical vibration or normal slope, and can sensitively and accurately reflect the real surface deterioration caused by local scour and collapse. 4. This invention constructs a state determination and dynamic response mechanism, which automatically maintains the route, generates early warnings or triggers emergency avoidance to replan the route based on the differential results, and generates a disturbance map to correct navigation parameters; this realizes a closed loop from anomaly identification to online avoidance, and actively avoids high-risk failure sections. 5. After the evaluation is completed on the vehicle, this invention only extracts the feature vectors of discrete points in the spatial domain and the generated disturbance intensity heat map for uploading; this avoids the direct back transmission of massive amounts of underlying raw data, greatly reduces bandwidth usage, and effectively meets the continuous monitoring needs of communication-constrained scenarios such as mountain orchards. Attached Figure Description

[0014] The invention will now be further described with reference to the accompanying drawings.

[0015] Figure 1 This is a schematic diagram of the modules of the soil erosion characteristic data processing system based on the filtering algorithm provided in the embodiments of this application. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Please see Figure 1A water and soil erosion characteristic data processing system based on filtering algorithm, the system is mounted on a mobile vehicle and includes: a data acquisition module configured to: acquire the underlying sensor data stream of the mobile vehicle in real time through integrated inertial sensing and positioning hardware; The main navigation module is configured to: receive data streams from the underlying sensors, perform time-domain updates using an error state Kalman filter configured with the first observation noise covariance to suppress terrain disturbances, generate a smooth navigation trajectory, and output the current geographic information system spatial mileage. The shadow monitoring module is configured to: receive the underlying sensor data stream and the current geographic information system spatial mileage, trigger the spatial domain state transition based on the current geographic information system spatial mileage, extract terrain disturbances using a filter configured with the second process noise covariance, and generate kinematic information residuals at discrete points in the spatial domain. The residual topology generation module is configured to: extract the feature components related to surface elevation and attitude from the kinematic innovation residuals to construct the current period kinematic perturbation feature vector; under the same geographic information system spatial coordinates, spatially align and differentially compare the current period kinematic perturbation feature vector with the historical baseline period kinematic perturbation feature vector of the same spatial segment obtained in advance to generate a navigation kinematic state anomaly assessment result; and based on the navigation kinematic state anomaly assessment result combined with the current geographic information system spatial mileage, map and generate a perturbation feature map for correcting the navigation parameters of the main navigation module.

[0018] This embodiment provides a soil erosion monitoring mechanism based on geographic information system. Specifically, the system is installed on an unmanned vehicle for patrolling mountain orchards. The unmanned vehicle periodically patrols along the terraced fields and drainage ditches according to a pre-set GIS operation route. The system does not directly map the three-dimensional topography of the ground surface, but transforms the phenomenon of changes in the ground support conditions after rainfall into a question of whether the navigation kinematics of the unmanned vehicle at the same spatial location is abnormal, which can be continuously monitored. Specifically, the data acquisition module continuously receives data streams from the underlying sensors. These data reflect the motion excitation experienced by the unmanned vehicle on the actual ground. For mountainous farmland, if the surface soil of a certain working road is washed away by rainwater to form shallow ditches, local collapses, or soft slip zones, the vertical vibration, pitch changes, short-term slippage, and trajectory correction actions of the vehicle will change accordingly when the unmanned vehicle passes through the area. The main navigation module's task is not to amplify these changes, but to ensure that the vehicle can travel stably along the predetermined route. Therefore, it adopts a robust time-domain filtering strategy to treat high-frequency disturbances as transient interferences that are detrimental to navigation control, thereby outputting a continuous and smooth navigation trajectory and providing the current GIS spatial mileage based on the trajectory. The spatial mileage can be understood as the distance the vehicle has traveled along the target operation route, and its role is to establish a unified spatial coordinate benchmark for subsequent historical comparisons. Based on this, the shadow detection module shares the same underlying sensor input with the main navigation module, but its purpose is not to control the vehicle, but to extract high-frequency ground disturbance information that will be actively suppressed by the main navigation link; this module does not observe the vehicle's state at this moment according to fixed time slices, but observes the vehicle's state when it reaches the same location according to spatial mileage. The reason for this is that the speed of daily orchard inspections can fluctuate due to road conditions, turns, obstacle avoidance, and mud levels. If only time is compared, the vibration peaks collected on different inspection days may correspond to different locations, which cannot truly reflect the changes on the same road section. After the shadow monitoring module uses spatial location as the trigger condition, it can extract the kinematic information residual at each discrete spatial point. This residual corresponds to the deviation that the vehicle should have passed smoothly according to the existing motion law, but was actually subjected to additional ground stimulation. The residual topology generation module further extracts the feature components most directly related to the surface elevation undulation and vehicle attitude change from the above residuals, and performs spatial alignment and differential comparison between the disturbance feature vector obtained in this inspection and the historical feature vector recorded in the benchmark inspection cycle for this road segment; if the difference between the two continues to widen, it indicates that the surface support conditions of this spatial road segment have changed. Although the system does not directly output the terrain profile, it can determine in a navigation sense that there are environmental changes in this road segment that are prone to cause trajectory deviation and attitude disturbance. The system writes back these abnormal assessment results as a disturbance feature map, which is used to correct the navigation parameters of the main navigation module in different spatial road segments, so that the vehicle can adopt a more cautious navigation strategy for high-risk road segments in subsequent inspections. As a fault-tolerance mechanism under abnormal operating conditions, if satellite signals are briefly blocked, tires slip and mileage information becomes unstable for a short time during the inspection, or sensor data for a certain period is missing, the system will prioritize maintaining the stable operation of the main navigation module and temporarily refrain from directly recording the abnormality of that period as a change in the ground surface. Only when similar disturbances continue to occur in subsequent adjacent spatial segments and are inconsistent with the characteristics of mechanical failures will they be included in the anomaly assessment. For road segments inspected for the first time and for which no historical benchmark has been established, they can be recorded as benchmark period data without outputting differential conclusions. For example, on the third day after continuous heavy rainfall, the orchard management sent an unmanned inspection vehicle to patrol along the No. 3 terrace on the south slope; when the vehicle passed through a section of road next to a drainage ditch that had been flat, the main navigation module kept the vehicle moving steadily along the route, treating the section as a local bump and smoothing it out. The shadow monitoring module detected that the vehicle's vertical impact and pitch changes were significantly enhanced compared to the previous baseline inspection at the same GIS mileage point, and this enhancement was concentrated on several consecutive spatial points. Based on this, the residual topology generation module marked the navigation kinematic state anomaly in this section of the road and generated a spatial map showing the enhanced disturbance area. When passing through this area again in the future, the main navigation parameters can be switched to a more conservative mode in advance to reduce the risk of vehicle deviation or getting stuck. The purpose of this step is to transform changes in surface support caused by soil erosion into kinematic anomalies that can be identified, aligned, and cumulatively analyzed within the navigation link, thereby achieving pure navigation-based monitoring that does not rely on laser mapping or visual modeling, while simultaneously ensuring the stable tracking capability of inspection vehicles and the ability to identify abnormal areas.

[0019] In this embodiment of the invention, the data acquisition module includes: an inertial measurement unit configured to acquire triaxial acceleration data and triaxial angular velocity data of a mobile carrier; a wheeled odometer configured to acquire relative motion displacement data of the mobile carrier; and a satellite navigation receiver configured to acquire absolute spatial coordinate data of the mobile carrier; wherein, the underlying sensor data stream is generated by aggregating triaxial acceleration data, triaxial angular velocity data, relative motion displacement data, and absolute spatial coordinate data.

[0020] This embodiment provides a low-level sensor data aggregation mechanism; specifically, the inspection unmanned vehicle uses an inertial measurement unit, a wheeled odometer, and a satellite navigation receiver to form a data acquisition link, so as to simultaneously perceive how the vehicle moves, how far the wheels have traveled, and where the vehicle is located; Specifically, the inertial measurement unit outputs triaxial acceleration and triaxial angular velocity, reflecting the transient excitation and attitude changes of the vehicle body. For vehicles operating in mountainous terrain, shallow ditches, collapsed edges, and soft zones formed after soil erosion will manifest as increased vertical impact on the vehicle body, intensified pitch changes, or short-term swaying when passing through them. Therefore, the inertial measurement unit is the most direct data source for identifying changes in ground support. Wheel odometers record relative displacement to help the system determine how far the vehicle has traveled along the route, especially when satellite signals fluctuate locally, so as to maintain continuous mileage reference on the work path; satellite navigation receivers provide absolute spatial coordinates to project the data obtained from each inspection onto a unified GIS coordinate system, ensuring that data collected on different dates, at different speeds, and under different weather conditions can still be located on the same road segment. In a simplified data example, the data continuously collected on a certain road segment can be understood as four types of synchronous sequences: inertial sequence A, angular velocity sequence B, wheel sequence C, and absolute coordinate sequence D; the system aggregates these four types of sequences within the same timestamp or adjacent valid sampling windows to form a low-level sensor data stream record; For example, in a certain record, A shows a short-term vertical peak, B shows a slight pitch change, C indicates that the vehicle is still moving forward, and D marks it as being located between 120 and 121 meters on the No. 3 terrace on the south slope; the system can then interpret the record as the vehicle encountering actual surface excitation in this spatial segment, rather than mistaking the isolated anomaly of a single sensor for soil erosion. In abnormal operating conditions, if the coordinate accuracy of the satellite navigation receiver decreases during a certain period of time, the system can temporarily rely on the wheel odometer to maintain relative mileage continuity, and then realign the spatial position after the satellite positioning is restored. If the wheel odometer reading fluctuates due to mud adhesion, its reliability is verified using inertial and satellite data. If the inertial measurement unit is temporarily saturated, the data in that window is used only for navigation stabilization and is not included in the highly sensitive surface anomaly analysis to avoid mistaking mechanical impact for long-term soil erosion characteristics. For example, during a post-rain inspection of an orchard on the south slope, when an unmanned vehicle passed the outer shoulder of a bend, the inertial measurement unit detected a simultaneous increase in vertical acceleration and pitch angular velocity. The wheeled odometer showed that the vehicle continued to move forward over a short distance, and the satellite navigation receiver provided fixed GIS coordinates indicating that the anomaly was located next to a drainage ditch. After aggregating these three types of information, the system does not simply regard the phenomenon as a vibration caused by driving actions, but further enters the subsequent spatial domain analysis process. The purpose of this step is to provide a homogeneous, complementary, and spatially aligned underlying data foundation for subsequent main navigation and shadow monitoring, thereby enabling reliable perception of changes in real ground support.

[0021] In this embodiment of the invention, the main navigation module includes: a time-domain filtering unit configured to update the state and measurement of the underlying sensor data stream using a preset time interval as a trigger condition; a smoothing control unit configured to filter and suppress the high-frequency transient inputs contained in the underlying sensor data stream using the first observation noise covariance to generate a smooth navigation trajectory; and a mileage calculation unit configured to calculate the current geographic information system spatial mileage of the mobile vehicle in the geographic information system coordinate system based on the smooth navigation trajectory.

[0022] This embodiment provides a main navigation stabilization control mechanism; specifically, after the aforementioned data aggregation link is established, the system first ensures that the inspection unmanned vehicle can continuously and stably travel along the GIS route in the complex road conditions of the mountain orchard through the main navigation module, and then provides the stable trajectory as a spatial reference to the monitoring link. Specifically, relying solely on raw inertia and displacement information, the vehicle's movement on bumpy roads will involve a large number of transient impacts. If these high-frequency signals are used directly to drive vehicle control, the vehicle is prone to frequent direction corrections, which can cause unnecessary swaying on slippery surfaces. Therefore, the time-domain filtering unit continuously integrates inertia, wheel travel, and satellite information according to a preset time rhythm to continuously update the vehicle's position, speed, and attitude. The smooth control unit treats ground impacts that last for less than a preset time threshold but have an amplitude greater than a preset acceleration threshold as transient inputs that are detrimental to tracking control and suppresses them in the main navigation link. In other words, the main navigation module focuses more on whether the vehicle maintains a stable trajectory along the road as a whole, rather than whether the vehicle is subjected to transient impacts from local terrain at a certain moment by a ditch. The mileage calculation unit further obtains the vehicle's current spatial mileage in the GIS coordinate system based on the smoothed navigation trajectory; the spatial mileage here is not the cumulative number of rotations of a simple wheel odometer, but a path location marker formed by combining the overall trajectory estimation. For example, on a 500-meter-long terraced field inspection route, the system can continuously provide the vehicle's current spatial position at approximately 80 meters, 80.5 meters, or 81 meters; all subsequent high-sensitivity monitoring is carried out around this stable spatial reference to avoid drift of monitoring points due to local vibrations. From an evolutionary perspective, if there is no stable main navigation link and instead the highly sensitive monitoring results are directly involved in vehicle control, the more obvious the changes in the ground, the more likely the system is to amplify anomalies into control disturbances, making it difficult for the vehicle to maintain the intended route. Therefore, this embodiment first separates navigation stability and anomaly perception in terms of function, forming a collaborative structure of seeking stability in control and seeking sensitivity in monitoring. As a fault-tolerant mechanism under abnormal operating conditions, if the satellite positioning suddenly changes within a certain time window, the time domain filtering unit can temporarily increase its reliance on inertia and wheel distance information to avoid unreasonable jumps in the trajectory; if the wheel distance and satellite trajectory are significantly inconsistent within multiple consecutive windows, the system marks this segment as a navigation reliability deterioration zone and only outputs a conservative spatial mileage estimate; if the vehicle is in a U-turn or moving at low speed, the spatial mileage update can be temporarily delayed or the resolution reduced to prevent the same spatial point from being repeatedly miscounted. For example, during the inspection of the No. 3 terrace on the south slope, the unmanned vehicle needs to travel along a narrow work path close to the irrigation canal. The road surface becomes soft due to rainfall, and the vehicle will experience continuous small bumps when passing through. The main navigation module does not treat each bump as an event that requires immediate route correction. Instead, it keeps the vehicle moving smoothly along the established GIS line and continuously outputs the spatial mileage information of the current location in the middle of the work path. In this way, the subsequent monitoring module can accurately know which section of the meter the anomaly occurred in. The purpose of this step is to ensure the continuous execution of inspection operations in the presence of complex surface disturbances, and to establish a stable and unified location benchmark for subsequent spatial domain anomaly comparison.

[0023] In this embodiment of the invention, the shadow monitoring module includes: a spatial domain triggering unit, configured to trigger a state update pipeline when the current geographic information system spatial mileage reaches a preset spatial interval; and a spatial mapping unit, configured to convert the underlying sensor data stream from a time domain integral model to a spatial domain integral model based on a preset spatial interval. The high-sensitivity residual extraction unit is configured to amplify the high-frequency transient input using the second process noise covariance under the spatial domain integral model, calculate the deviation between the predicted state derived from the spatial domain integral model and the actual observation value provided by the underlying sensor data stream, and generate kinematic innovation residuals.

[0024] This embodiment provides a spatial domain high-sensitivity residual extraction mechanism; specifically, after the main navigation module has given a stable spatial mileage, the shadow monitoring module no longer analyzes the data every certain period of time, but instead analyzes the data once every fixed spatial interval, thereby eliminating the speed difference, pause difference, and turning rhythm difference between different inspection cycles. Specifically, in the scenario of soil and water conservation monitoring, the same work path may be passed slowly today due to mud, and may be passed faster tomorrow due to drier ground. If the system still compares data by time slices, the 5th second today might correspond to the 20th meter position, while the 5th second tomorrow might correspond to the 28th meter position, resulting in misalignment between the two sets of data. The spatial domain triggering unit solves this problem by triggering a state update whenever the spatial mileage increases to a preset interval, such as every short distance. In this way, the system observes the kinematic response generated when the vehicle passes the same spatial point. The spatial mapping unit transforms the sensor inputs that originally accumulated over time into an analysis object that accumulates over the path length; its engineering significance lies in rewriting the vehicle's motion history into a spatial profile state sequence unfolding along the route. In the derivation of the integral model, the state transition matrix and the time differential variables in the system equations are considered. Replace with spatial displacement differential variables This makes the filter's state recursion no longer dependent on the speed of time, but strictly synchronized with the physical distance traveled by the mobile vehicle on the geographic information system's flight path. The high-sensitivity residual extraction unit differs from the main navigation module in that it no longer actively suppresses high-frequency inputs, but instead retains or even amplifies short-term shocks related to local ground mutations. The so-called kinematic information residual can be understood as the difference between the smooth state that a vehicle should exhibit under normal road conditions and the state that the vehicle actually exhibits after being stimulated by the ground at that point. If a roadway was originally compacted and stable, but after rain, fine ditches or shoulder collapses occur, the actual response of the vehicle at that point will deviate significantly from the historical smooth passage pattern, and the residual will increase accordingly. A simplified data example can be used to illustrate its data flow: Suppose the system divides a 30-centimeter stretch of road into three spatial points: S1, S2, and S3. During the first baseline inspection, the vehicles passing through these three points are relatively stable, and the residuals obtained by the shadow monitoring module are low, low, and low, respectively. During the second inspection after a rainstorm, a shallow ditch forms near S2, and the vehicle experiences a significant vertical impact at that point. The new residual sequence is low, high, and low. Because the anomaly occurs at the same spatial point rather than at the same time, the system can determine that its source is closer to changes in the ground surface than simply a difference in driving rhythm. Furthermore, to avoid the time-domain to spatial-domain representation being too abstract, the spatial mapping unit in this embodiment operates according to the following rules: first, the smooth trajectory output by the main navigation module is used as a path reference, the corresponding spatial increment is calculated between adjacent effective time samples, and the observations of each sensor are attached to the spatial increment; when the cumulative spatial increment reaches the preset spatial interval, a new spatial state node is formed. The system does not accumulate states based on a fixed number of seconds, but rather on a fixed distance traveled. For the same spatial node, the predicted state is derived from the spatial domain prediction value calculated from the previous spatial node along that spatial interval, while the actual observation value comes from the combined inertial, wheel distance, and satellite observations after resampling within that spatial interval. Thus, the kinematic innovation residual is limited to the difference between the predicted and actual passing states within the same spatial segment, rather than time-slice biases mixed with speed differences. Specifically, in the navigation computer's data processing pipeline, the spatial domain kinematic innovation residual is calculated using the following state-space digital solution equation: In the formula, Let be the kinematic innovation residual vector at the discrete node in space. The actual observation vectors provided for the aggregated underlying sensor data stream. For the system observation matrix, This is the spatial domain state transition matrix derived based on a preset spatial interval. This is the optimal state estimation vector of the previous spatial node. Through matrix multiplication and addition / subtraction operations, the digital mapping of the raw data from the underlying sensors to environmental state characteristics is completed. In terms of filter parameter configuration, the first observation noise covariance in the main navigation module is configured to a small value to force the system to trust historical smooth prediction, while the second process noise covariance in the shadow monitoring module is configured to be strictly larger than the corresponding process noise covariance in the main navigation module in order to improve the sensitivity to high-frequency ground excitation. Furthermore, the role of the second process noise covariance in the shadow monitoring module is not to amplify the noise itself without constraint, but to adopt a higher tolerance for state changes compared to the main navigation module, so that the filter does not rush to interpret local mutations as random errors and immediately filter them out; In other words, short-term terrain stimuli that are considered to be suppressed in the main navigation link are retained as candidate perturbations that can be included in the innovation calculation in the shadow monitoring link. Thus, although the main navigation module and the shadow monitoring module share the underlying sensor input, they correspond to two different tasks in terms of filtering objectives: ensuring smooth trajectory and revealing spatial anomalies. Furthermore, if the vehicle generates almost no effective spatial increment per unit time when it is at extremely low speed, turning in place, or slipping briefly, the spatial domain triggering unit will not forcibly generate a new spatial node, but will instead merge the time window into the adjacent effective spatial segments before and after it. Only when the cumulative spatial increment reaches the preset interval again will a spatial domain state update be performed. This avoids directly treating shaking in place, parking obstacle avoidance, or low-speed idling as ground anomalies at fixed spatial locations. Under abnormal operating conditions, if the vehicle temporarily deviates from the planned line due to obstacle avoidance within a certain spatial interval, or makes a sharp turn due to manual intervention, the residual at that spatial point may contain a sudden change in attitude caused by the control behavior; at this time, the system can filter out such errors by combining steering state, speed change and continuity of adjacent points. If an anomaly occurs only at a single isolated point and there is no continuity between the preceding and following spatial points, it is preferentially classified as an occasional impact and not immediately used as evidence of soil erosion; if multiple consecutive spatial points show highly sensitive residuals in the same direction, its credibility is increased and it is included in subsequent topological mapping. For example, on the road beside the drainage ditch in the orchard on the south slope, the system uses 5 centimeters as a spatial trigger interval; when the unmanned vehicle passes through three positions at 124.00 meters, 124.05 meters and 124.10 meters during a certain inspection, the responses at the two points before and after are close to the benchmark, but a significant vertical impact and nose-down tilt occur at the middle point; since this anomaly is fixed near 124.05 meters, rather than at a certain second, the system can interpret it as a local depression formed by erosion of the roadbed at that point; The purpose of this mechanism is to transform the originally difficult-to-align temporal navigation disturbances into spatial anomalies that can be compared point by point in the GIS route, thereby improving the stability of identifying the location of minor soil erosion.

[0025] In this embodiment of the invention, the shadow monitoring module further includes a data resampling unit, configured to resample the underlying sensor data stream acquired based on discrete time into a data sequence based on equal spatial intervals using a spline interpolation algorithm before the spatial domain triggering unit is triggered, and input the data sequence to the spatial mapping unit.

[0026] This embodiment provides a resampling mechanism with equal spatial intervals. Specifically, after using spatial mileage triggering, a practical problem still needs to be solved: the underlying sensors are naturally sampled by time, not by space. If the vehicle speed fluctuates drastically, the spatial length covered by the same number of time sampling points will not be consistent, and direct use for spatial analysis will result in uneven point density. In detail, although the previous layer solution solved the problem of where to compare, if resampling is not done first, the data around the 124-meter mark will sometimes have a denser number of inertial points and sometimes only a few points, which will cause the strength of local features to be affected by speed. Therefore, before spatial triggering, the data resampling unit first reorganizes the discrete-time data into a data sequence with equal spatial intervals based on the existing displacement information and absolute coordinate information; the engineering significance of spline interpolation here is not to do complex modeling, but to use a continuous and smooth way to complete the state changes of the vehicle in adjacent spatial positions, so that each preset spatial point can obtain the corresponding data representation. A simplified data processing model can be used to illustrate this: assume that the vehicle is sampled on a short stretch of road and four time points T1, T2, T3, and T4 are obtained, with corresponding spatial positions of approximately 0 cm, 3 cm, 9 cm, and 10 cm, respectively; If we directly use it for spatial analysis every 5 centimeters, there will be a lack of direct observations near the 5th centimeter. After resampling, the system can form new spatial sequences P0, P5, and P10, where P5 is obtained by smooth interpolation of adjacent original data. In this way, even if the driving rhythm is different in different inspection cycles, they can eventually be unified to the same spatial scale for comparison. From an evolutionary perspective, without this resampling process, spatial domain analysis may still produce errors in low-speed, sudden-stop, or accelerated-through scenarios. For example, when a vehicle slows down in muddy conditions, the sensor will collect many time points within a short distance. If these points are used directly, the system may mistakenly interpret the increased sampling density caused by speed changes as a stronger surface anomaly. Therefore, this embodiment preprocesses the sampling density difference by resampling at equal spatial intervals. As a fault-tolerance mechanism under abnormal operating conditions, if a segment of raw data is interrupted due to communication or caching issues, and the span of the interruption exceeds a preset upper limit, then this segment will not be interpolated and filled in, but will be marked as a spatial segment to be retested to avoid generating unreliable features based on excessively long gaps; if the vehicle slips in place, causing the correspondence between wheel travel and spatial displacement to become distorted, then the system will prioritize the absolute coordinates and attitude continuity, which have higher reliability, to constrain the resampling process; if adjacent raw points themselves undergo abnormal jumps, then deburring will be performed first, and then resampling will begin. For example, during a post-rain inspection of the orchard on the south slope, the unmanned vehicle slowed down significantly when passing through a section of soft mud, resulting in the collection of multiple sets of inertial data within 0.2 seconds, but it only actually moved a few centimeters. However, on a later compacted section, the vehicle accelerated and could move more than ten centimeters in 0.2 seconds. The data resampling unit unified these two types of data with different temporal densities into a sequence of spatial points every 5 centimeters, so that the features near the 124-meter mark and near the 130-meter mark could be compared at a consistent spatial scale. The purpose of this mechanism is to eliminate the sampling density differences caused by speed fluctuations, thereby achieving more stable spatial alignment and more realistic anomaly comparisons between road segments.

[0027] In this embodiment of the invention, the residual topology generation module includes: a feature extraction unit configured to separate the Z-axis acceleration component and pitch angle component in the moving vehicle's body coordinate system from the kinematic innovation residual; a vector splicing unit configured to combine the Z-axis acceleration component and pitch angle component to construct the current period kinematic perturbation feature vector; and a difference comparison unit configured to calculate the difference between the same components of the current period kinematic perturbation feature vector and the historical baseline period kinematic perturbation feature vector on a unified geographic information system spatial coordinate axis, and to perform dimensionless weighted fusion of the difference results to generate a one-dimensional risk intensity difference sequence.

[0028] This embodiment provides a mechanism for constructing and differentially comparing kinematic perturbation features. Specifically, after obtaining the kinematic innovation residuals at each spatial point, the system does not need to retain residual signals in all directions and frequency bands, but instead prioritizes extracting feature components that are most directly related to surface undulations and support changes. Specifically, for wheeled inspection unmanned vehicles traversing mountainous work areas, the Z-axis acceleration component mainly reflects the vertical impact intensity experienced by the vehicle body, and can correspond to whether the ground is flat, concave, or protruding. The Z-axis here specifically refers to the coordinate axis perpendicular to the chassis of the vehicle in the coordinate system of the mobile carrier itself, so as to ensure that the extracted acceleration can truly reflect the vertical vibration of the vehicle relative to the road surface, without being affected by the slope or attitude changes of the geographic coordinate system. The pitch component reflects the front-to-back tilt of the vehicle, and can correspond to the attitude changes when the front wheels sink first, the rear wheels compensate for the passage, or the whole vehicle crosses a shallow ditch. The reason these two types of features are used together is that when looking at vertical impact alone, it may be impossible to distinguish between mechanical resonance and actual road surface indentation; when looking at pitch change alone, it may be mixed with normal attitude changes when going uphill or downhill; combining the two can better characterize the actual passage characteristics caused by changes in road surface support conditions at the same spatial point. The vector concatenation unit combines the two features mentioned above at each spatial point into a kinematic perturbation feature vector for the current period; the difference comparison unit then differs this vector with the vector at the same spatial point of the historical reference period to obtain a difference sequence. The topology here refers to arranging the disturbance feature vectors at each isolated spatial point in a logical array on a one-dimensional spatial sequence according to the physical connection order of the geographic information system route, thereby reflecting the extension structure of the road surface support change in spatial continuity. This difference sequence does not represent the terrain geometry itself, but rather whether the kinematic environment has changed when a vehicle passes through that location; It can be represented by a minimal sandbox: Assuming a certain road segment has three consecutive spatial points P1, P2, and P3, the baseline periodic characteristics can be written as a sequence of three elements: the first spatial point is stationary and stationary, the second spatial point is slightly sloping and slightly impacted, and the third spatial point is stationary and stationary. The current cycle characteristics are as follows: the first spatial point is stable and steady, the second spatial point is obvious downward and obvious impact, and the third spatial point is slight aftershock and slight rebound. After differentiation, it can be seen that the main anomalies are concentrated in P2 and continue slightly towards P3. This is consistent with the typical passing behavior of a vehicle after passing through a newly formed shallow ditch, with the front wheels dipping down and the rear wheels straightening. In this way, the system can identify whether the surface condition has deteriorated without restoring the three-dimensional topography of the ground. From an evolutionary perspective, if a rough comparison is made directly on the complete residual sequence, it is easy to introduce irrelevant components related to steering, driving, and mechanical vibration, leading to divergent anomaly localization. Therefore, this embodiment converges the features to two core dimensions: vertical excitation and pitch attitude, which not only compresses the amount of data but also improves the directionality of changes in surface support. Furthermore, to avoid ambiguity caused by directly subtracting the Z-axis acceleration and pitch angle components due to their different dimensions, this embodiment uses a paired difference method with the same component and the same reference for the two components; specifically, the first component arranged along the unified GIS spatial coordinate axis... At each discrete point in space, the kinematic disturbance eigenvector of the current inspection cycle is denoted as:

[0029] in This indicates the Z-axis acceleration component corresponding to that point. This represents the pitch angle component corresponding to that point; The system first calculates the vertical impact difference and pitch change difference separately, and then inputs these two differences at the same spatial point as a set of joint anomaly descriptions into the subsequent state determination, instead of directly mixing different physical quantities into a single original value; where the superscript (ref) is used to indicate the component corresponding to the historical benchmark inspection cycle, rather than other filter states or control parameters. Furthermore, the kinematic disturbance feature vector of the historical reference cycle is preferably derived from the reference inspection data of the same road segment, similar load and normal traffic conditions; if the same road segment has uphill sections, downhill sections, curved sections or different working conditions, then corresponding sub-references are established respectively. In this way, the differential comparison unit does not perform a horizontal comparison at any time on the unified GIS spatial coordinate axis, but rather a comparison of the same spatial point, the same road segment attributes, and the same vehicle operating conditions, thereby reducing the contamination of the difference sequence by intrinsic slope changes, vehicle speed differences, and loading differences. Furthermore, when the current features of a spatial point differ significantly from the baseline features, but no similar patterns appear in its adjacent spatial points, the system first marks the point as a suspected isolated outlier. Only when the difference model exhibits a continuous or gradual relationship at nearby spatial points can its credibility as evidence of changes in surface support be further enhanced; this preserves sensitivity to small depressions while reducing the false recognition of random events such as instantaneous mechanical noises and accidental stone impacts. As a fault-tolerant mechanism under abnormal operating conditions, if the vehicle is continuously climbing or descending a slope, the pitch angle itself will have a slow trend of change. At this time, the system will use the historical benchmark of the same slope as a reference and only pay attention to the additional changes that deviate from the benchmark, without treating the slope itself as abnormal. If the vehicle suspension response changes due to changes in loading during a certain inspection, the benchmark cycle for the corresponding operating condition can be updated again. If a spatial point only shows an abnormality in the Z-axis but no significant change in pitch during a single inspection, or vice versa, the abnormality level of that point should be reduced first, and a comprehensive evaluation should be conducted after a continuous pattern is formed among the neighboring spatial points. For example, at the 124-meter section along the drainage ditch in the orchard on the south slope, the system extracted from the high-sensitivity residual that the Z-axis impact at this point was significantly increased, and was accompanied by a momentary downward tilt of the vehicle's front end; compared with the previous benchmark inspection on a sunny day, the difference in this combination of characteristics was concentrated and obvious; therefore, the system classified this spatial point and its neighboring points into the disturbance enhancement zone, rather than interpreting this anomaly as random vibration; The purpose of this mechanism is to characterize the impact of soil erosion on navigation passability using a small number of features with clear physical meaning, thereby improving the robustness of spatial comparison and reducing the burden of subsequent backhaul.

[0030] In this embodiment of the invention, the system further includes: a state determination module, configured to perform condition determination on the difference sequence based on a preset threshold: when the difference in the difference sequence is less than a preset safety threshold, it is determined that the navigation environment of the current spatial segment has not changed to affect the navigation accuracy, and the current inspection route of the mobile carrier is maintained; When the difference in the difference sequence is greater than or equal to the safety threshold and less than the preset danger threshold, it is determined that there is a navigation environment disturbance in the current spatial segment, and a local trajectory deviation warning information is generated and recorded in the navigation kinematic state anomaly assessment result; when the difference in the difference sequence is greater than or equal to the danger threshold, it is determined that there is a navigation failure risk in the current spatial segment, an emergency avoidance command is triggered and the subsequent geographic information system route of the mobile vehicle is replanned.

[0031] This embodiment provides a hierarchical status determination and route response mechanism; specifically, after obtaining the difference sequence of each spatial point, the system does not treat all anomalies equally, but makes hierarchical determinations according to the degree of impact on navigation safety and mission continuity; Specifically, the safety threshold corresponds to a state where the road surface has changed but has not significantly affected the accuracy of vehicle navigation, such as slight surface erosion, shallow loosening, or accumulation of small gravel; at this time, the vehicle can still pass through stably, and the system only needs to maintain the current inspection route. The danger threshold corresponds to the state where changes in the ground surface may cause the vehicle to lose its attitude, deviate significantly from its trajectory, or even get stuck in a local area, such as collapsed edges, deep ditches, soft sinkholes, or fractured edges formed after erosion; the state between the two indicates that the road surface has been disturbed by the navigation environment, but has not yet reached the point where it must be detoured immediately. It is suitable to generate a local trajectory deviation warning and retain the passage record for subsequent operation and maintenance. This tiered judgment essentially reflects the risk management approach in engineering: not all soil erosion should be immediately stopped during inspections, but if it has affected navigation controllability, the inspection task can no longer be prioritized over driving safety; the system makes tiered responses based on the strength and spatial continuity of the difference sequence, so that the monitoring results directly serve vehicle route management. A simplified data processing model can be used to represent this: the difference results of three consecutive spatial points on a certain road segment are low, medium, and high, respectively; the system maintains the original route for low, records an early warning for medium and prompts that the segment needs to be manually reviewed, and triggers risk avoidance for high; if all three points are medium and continuously distributed, the risk level of the road segment can also be improved as a whole, because the continuity indicates that the anomaly is not a single point of occasional disturbance, but rather an overall deterioration of the roadbed condition. From an evolutionary perspective, differential comparison alone is insufficient to guide vehicle behavior; if the system only generates offline reports without linking them to real-time inspection paths, the autonomous vehicle may still enter obviously unstable road sections; therefore, this embodiment further transforms the anomaly identification results into online navigation decisions. Furthermore, to ensure that the differences in the difference sequence have a unified and directly identifiable standard, the state determination module performs joint difference analysis on each spatial point obtained in the embodiment. Dimensionless fusion is performed to obtain the risk index for this spatial point. ;in, Indicates the first The difference in Z-axis acceleration at each spatial discrete point relative to the historical baseline inspection cycle. This represents the pitch angle difference at the same point relative to the historical baseline inspection cycle; the two together constitute the joint difference result for that point. ; The system then uses the upper bound of the normal fluctuation of the corresponding component in the benchmark period or the calibration tolerance to unify the scale of the two components, and combines them into a single-point risk intensity according to preset weights. Thus, the safety threshold and the danger threshold correspond to the same risk indicator. Different intervals, instead of directly using the original difference values ​​of different dimensions for mixed judgment; The dimensionless fusion calculation formula for the volume is expressed as follows:

[0032] in, This serves as a risk indicator for that spatial point. and For the extracted feature differences, and These are the pre-calibrated upper limits for Z-axis acceleration tolerance and pitch angle tolerance, respectively. and The preset weighting coefficients satisfy the following conditions: ; The single-point scoring, continuous segment confirmation, safety threshold comparison, and hazard threshold comparison mentioned below are all numbered sequentially along the inspection route. Risk indicators corresponding to spatial discrete points To determine the object; Furthermore, the state determination module preferably employs a single-point scoring plus continuous segment confirmation method; for a single spatial point, if If the value is below the safety threshold, the point is considered safe; if... If a point is located between the safety threshold and the danger threshold, it is recorded as a warning candidate; if (R_i) reaches or exceeds the danger threshold, it is recorded as a high-risk candidate. The system then combines the continuity of adjacent spatial points to perform segment-level corrections: when only a single point briefly exceeds the threshold and the points before and after it quickly recover, it is preferentially regarded as an occasional impact; when multiple adjacent points continuously exceed the safety threshold, a local trajectory deviation warning is output; when multiple adjacent points continuously exceed the danger threshold, or although a single point does not reach the danger threshold completely but the average risk of its continuous segment has significantly increased and is accompanied by obvious lateral deviation, sinking or repeated correction actions, it can also be judged as a navigation failure risk; thus, the threshold determination retains the sensitivity to sharp local damage and avoids excessive response triggered by isolated noise points. Furthermore, the safety threshold and danger threshold are preferably determined by the joint calibration of benchmark inspection samples, historical fault samples, and actual vehicle pass tests; specifically, threshold configuration tables can be established according to road surface type, slope range, load condition, and tire condition. For example, a more conservative threshold can be used for narrow sections of road next to drainage ditches, while a standard threshold can be used for compacted main work roads; in this way, the warning and risk avoidance conclusions output by the status determination module are matched with the specific working environment, rather than a set of thresholds that are applicable to all roads. Furthermore, once a certain spatial segment is determined to have a risk of navigation environment disturbance or navigation failure, the system will write the spatial number, risk level and dominant anomaly type of the segment into the navigation kinematic state anomaly assessment result for subsequent disturbance feature map generation and main navigation parameter correction. Among them, the dominant anomaly types can be described as vertical impact enhancement, front wheel downward pitch anomaly, continuous soft sinking disturbance, etc., so that the subsequent main navigation module can take strategies such as speed limit, increase trajectory tolerance zone or switch conservative control parameters in advance on the same road segment. Furthermore, when the difference in the difference sequence is greater than or equal to the danger threshold, the emergency avoidance instruction preferably includes at least one of deceleration, stopping, exiting the current high-risk space segment, or switching to a safe waiting point, and then the subsequent geographic information system route replanning is executed; Whether there is a rescue mission or the intention to continue passage manually does not change the main line of processing that triggers the emergency evacuation order and re-plans the subsequent geographic information system route; it only affects the specific degree of conservatism of the evacuation action and the waiting strategy. As a fault-tolerant mechanism under abnormal operating conditions, if the difference sequence fluctuates repeatedly near the safety threshold, the system can introduce a continuous spatial segment confirmation mechanism, that is, only when multiple adjacent spatial points meet the warning conditions will the warning be output, so as to avoid over-responding to occasional single-point vibrations. If an abnormal danger level occurs during a turn, U-turn, or manual remote control takeover area, the source of the danger should be identified by combining the vehicle control mode to avoid misjudging the operation as a ground failure. If the current task is of an emergency nature and must continue, the system should first perform emergency avoidance and complete the subsequent geographic information system route replanning. Then, based on the replanning results, a low-speed conservative passage or a safe waiting strategy should be selected instead of continuing to advance along the original high-risk route. For example, during an inspection of an orchard on the south slope, the system found that the difference in a small section of road near the 124-meter mark was significantly higher than usual, but still did not reach a high-risk level. Therefore, a local trajectory deviation warning was generated, indicating that the section may have been eroded by rainwater. As the vehicle continued forward to approximately 138 meters, the differences between several consecutive spatial points increased further, and the vehicle exhibited a significant tendency to veer sideways as it passed. Based on this, the system determined that there was a risk of navigation failure, immediately stopped proceeding along the original work path, and replanned to a nearby, more stable field passage. The purpose of this mechanism is to transform spatial anomalies from observable information into actionable navigation responses, thereby achieving a balance between the continuity of inspection tasks and vehicle safety.

[0033] In this embodiment of the invention, the system further includes: a cloud mapping module, configured to map the navigation kinematic state anomaly assessment result into a navigation environment disturbance intensity heatmap; and a feature backhaul module, configured to extract the current periodic kinematic disturbance feature vector and the navigation environment disturbance intensity heatmap at discrete points in the spatial domain, and upload them to a remote server.

[0034] This embodiment provides a cloud mapping and low-bandwidth backhaul mechanism; specifically, after completing anomaly identification and hierarchical response, the system does not need to upload all the high-frequency raw sensor waveforms, but only backhauls the key features and mapping results at spatial discrete points, thereby adapting to the application environment of limited outdoor network conditions in mountain orchards. Specifically, field inspections typically rely on low-power wireless links or intermittent cellular networks. Continuously uploading large amounts of raw inertial data not only consumes bandwidth but also increases latency and storage pressure. The cloud mapping module first organizes the abnormal results that have been evaluated locally into a heat map of the intensity of navigation environment disturbances. The heat map reflects which sections of the GIS route have a significant impact on navigation passability, making it suitable for orchard managers to visually view repair priorities; the feature feedback module extracts the disturbance feature vectors at each spatial discrete point and the corresponding heat map results for uploading; in this way, the cloud retains the core information required for subsequent review and cross-period comparison, while avoiding the transmission of raw big data streams. A simplified data processing model can be used to illustrate this: an inspection generates analysis results for 100 spatial points; the system does not need to upload the hundreds of raw time-sampled values ​​behind each point, but only needs to upload 100 sets of compressed feature vectors and corresponding risk color markers; The cloud platform can display on the map that the 120-126 meter range is a medium disturbance zone and the 136-140 meter range is a high disturbance zone; management personnel can then arrange on-site backfilling, drainage ditch repair, or temporary closure of passage based on this information. From an evolutionary perspective, without this feedback compression mechanism, although the system can complete the identification at the vehicle end, it would be difficult to form a unified GIS archive for long-term monitoring across days and regions; and if all the raw data is uploaded, the probability of failure in field deployment and communication would increase; therefore, in this embodiment, feature extraction is performed at the vehicle end first, and then macroscopic mapping is performed in the cloud. Furthermore, to ensure that the data flow between the cloud mapping module and the feature feedback module remains consistent, the navigation environment disturbance intensity heatmap in this embodiment preferably adopts a hierarchical representation: the vehicle first maps the navigation kinematic state anomaly assessment results into heatmap summary data according to a unified color grading rule or segment coding rule; the feature feedback module uploads the current period kinematic disturbance feature vector at the discrete points in the spatial domain along with the heatmap summary data to the remote server. The cloud mapping module on the remote server then restores and renders the complete navigation environment disturbance intensity heatmap according to the same mapping rules; thus, the heatmap extracted and uploaded corresponds to the heatmap encoding result or summary result, while the final generation and display of the visualized heatmap is completed on the remote server side. Furthermore, the heat map summary data can be organized by spatial point number, risk level, dominant anomaly type, and color label; for example, the GIS coordinate index, risk range, and corresponding color level can be recorded for a single spatial point or continuous spatial segment. After receiving the data, the server can directly stitch the data into a route heat map in spatial order without having to send back all the original waveforms. As a fault-tolerant mechanism under abnormal operating conditions, if the network is unavailable during the inspection process, the system can first cache the feature vector and heat map results in local storage, and then upload them in batches after returning to the base station or when the signal is restored. If a batch of data uploads is interrupted, the cloud can resume the upload based on the spatial point number and inspection cycle information. If the cloud only receives data for a portion of the road segment, it can still update the local heat map first, and then complete the full data summary for the entire route after the remaining data is completed. For example, after completing a day's inspection of the orchard on the south slope, the unmanned vehicle has identified two high-risk work lanes and three slightly disturbed road sections locally; Due to unstable communication in the orchard valley area, the system only uploads the spatial feature vectors and heat map summaries corresponding to these road sections to the remote server; the management platform then displays the road edge areas with obvious erosion on the GIS interface, and the maintenance personnel prioritize the slope protection and soil filling operations for high-risk sections based on this. The purpose of this mechanism is to retain the most critical navigation anomaly information under limited communication conditions, thereby achieving effective connection between real-time vehicle monitoring and long-term cloud-based operation and maintenance management.

[0035] In this embodiment of the invention, the inertial measurement unit adopts a six-axis microelectromechanical system inertial sensor; the wheel odometer adopts an optoelectronic encoder; the satellite navigation receiver adopts a real-time dynamic differential global navigation satellite system module; the data acquisition module, the main navigation module, and the shadow monitoring module are all integrated into the navigation computer of the mobile vehicle; wherein, the navigation computer, as the core data processing hardware device, contains at least one microprocessor and a non-volatile memory. The memory stores computer program code that implements Kalman filtering, spatial mapping integration, and topological feature differential comparison. The microprocessor periodically processes the digital signal input of the underlying sensor data stream by calling and executing the computer program code, and dynamically generates a disturbance feature map data structure in memory for correcting the main navigation parameters.

[0036] This embodiment provides a hardware integration mechanism suitable for field inspection deployment; specifically, the system can use a six-axis microelectromechanical system inertial sensor, photoelectric encoder and real-time dynamic differential global navigation satellite system module as the main sensing components, and deploy data acquisition, main navigation and shadow detection in a unified manner in the vehicle navigation computer; Specifically, the six-axis microelectromechanical system inertial sensor can continuously reflect the vertical impact and attitude changes of the vehicle body with a high update rate, making it suitable for installation on mobile platforms with limited space and power supply conditions, such as agricultural inspection unmanned vehicles; the photoelectric encoder can stably acquire the rotational displacement of the wheels, making it easy to form relative mileage information, especially suitable for repeated inspections along fixed routes. The real-time dynamic differential global navigation satellite system module provides higher precision absolute spatial coordinates, enabling the same road segment in different inspection cycles to overlap more accurately in GIS; the combination of the three can not only meet the stable tracking requirements of the main navigation, but also support the spatial positioning of shadow monitoring for minor surface disturbances. Furthermore, to maintain terminology consistency, in this embodiment, all descriptions of the hardware implementation of the satellite navigation receiver refer to the Real-Time Dynamic Differential Global Navigation Satellite System module; the aforementioned descriptions of positioning accuracy recovery, short-term failure compensation, or rooftop antenna installation location all refer to this Real-Time Dynamic Differential Global Navigation Satellite System module, and will no longer use other abbreviations to replace the same hardware object. Integrating multiple functional modules into the same navigation computer helps reduce cross-device communication latency and time synchronization errors. For this solution, time synchronization and spatial alignment are directly related to the reliability of residual extraction. If data acquisition is in one controller, main navigation is in another device, and shadow monitoring is in a third device, clock drift and interface buffer latency between modules may cause the same ground event to be misaligned in different links. After unified integration, raw data access, navigation calculation, spatial resampling and residual analysis can be completed directly on the same computing platform, making the engineering implementation more compact and deployment and maintenance more convenient; From the perspective of stable system operation, the aforementioned solution has provided a complete processing flow. However, if the hardware links are scattered and the equipment is highly heterogeneous, cable failures, interface interruptions, or synchronization mismatches are likely to occur in muddy, humid, and vibrating conditions in the field, affecting long-term stable inspection. Therefore, this embodiment further limits the hardware combination and integration method suitable for actual deployment. In abnormal operating conditions, if the real-time dynamic differential global navigation satellite system module fails briefly in forest edge or mountain valley areas, the vehicle navigation computer can still maintain continuous navigation for a short period of time based on inertial sensors and photoelectric encoders. If the encoder deviates due to the presence of mud and sand, the navigation computer can perform cross-correction by combining inertial information and satellite coordinates; if the load on the vehicle computer increases, the main navigation and status determination tasks can be prioritized, and the shadow monitoring results can be processed by delaying the writing to disk or uploading in batches to avoid affecting the real-time safety of the vehicle. For example, on the unmanned inspection vehicles deployed in batches in the Nanpo Orchard, a six-axis microelectromechanical system inertial sensor is installed near the chassis center of gravity, an optical encoder is arranged on the drive wheel side, and a real-time dynamic differential global navigation satellite system module is installed on the roof antenna position; All data is uniformly connected to the navigation computer, which synchronously outputs smooth trajectory, spatial mileage and high-sensitivity residual analysis results during vehicle movement; even if the vehicle passes through the edge of the tree canopy, causing a short-term decrease in the positioning accuracy of the real-time dynamic differential global navigation satellite system module, the system can still maintain continuous inspection and re-align the abnormal road section to the GIS coordinates after the signal is restored. The purpose of this mechanism is to build a long-term operational vehicle-mounted monitoring platform using existing mature hardware, thereby achieving easy-to-maintain and engineering-feasible navigation-based soil erosion monitoring.

[0037] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A water and soil erosion characteristic data processing system based on filtering algorithms, characterized in that, The system is mounted on a mobile carrier and includes: The data acquisition module is configured to: acquire the underlying sensor data stream of the mobile carrier in real time through integrated inertial sensing and positioning hardware; The main navigation module is configured to: receive data streams from the underlying sensors, perform time-domain updates using an error state Kalman filter configured with the first observation noise covariance to suppress terrain disturbances, generate a smooth navigation trajectory, and output the current geographic information system spatial mileage. The shadow monitoring module is configured to: receive the underlying sensor data stream and the current geographic information system spatial mileage, trigger the spatial domain state transition based on the current geographic information system spatial mileage, extract terrain disturbances using a filter configured with the second process noise covariance, and generate kinematic information residuals at discrete points in the spatial domain. The residual topology generation module is configured to: extract the feature components related to surface elevation and attitude from the kinematic innovation residuals to construct the current period kinematic perturbation feature vector; under the same geographic information system spatial coordinates, spatially align and differentially compare the current period kinematic perturbation feature vector with the historical baseline period kinematic perturbation feature vector of the same spatial segment obtained in advance to generate a navigation kinematic state anomaly assessment result; and based on the navigation kinematic state anomaly assessment result combined with the current geographic information system spatial mileage, map and generate a perturbation feature map for correcting the navigation parameters of the main navigation module.

2. The soil erosion characteristic data processing system based on filtering algorithm according to claim 1, characterized in that, The data acquisition module includes: An inertial measurement unit is configured to acquire triaxial acceleration and triaxial angular velocity data of a moving vehicle. Wheel-mounted odometer, configured to collect relative motion displacement data of a moving vehicle; A satellite navigation receiver, configured to collect absolute spatial coordinate data of a mobile vehicle; The underlying sensor data stream is generated by aggregating triaxial acceleration data, triaxial angular velocity data, relative motion displacement data, and absolute spatial coordinate data.

3. The soil erosion characteristic data processing system based on filtering algorithm according to claim 2, characterized in that, The main navigation module includes: The time-domain filtering unit is configured to update the state and measurement of the underlying sensor data stream by using a preset time interval as the trigger condition. The smoothing control unit is configured to use the first observation noise covariance to filter and suppress high-frequency transient inputs contained in the underlying sensor data stream to generate a smooth navigation trajectory; The mileage calculation unit is configured to calculate the current geographic information system spatial mileage of the mobile vehicle in the geographic information system coordinate system based on a smooth navigation trajectory.

4. The soil erosion characteristic data processing system based on filtering algorithm according to claim 3, characterized in that, The shadow detection module includes: The spatial domain triggering unit is configured to trigger the status update pipeline when the current geographic information system spatial mileage reaches a preset spatial interval. The spatial mapping unit is configured to convert the underlying sensor data stream from a time-domain integral model to a spatial-domain integral model based on a preset spatial interval. The high-sensitivity residual extraction unit is configured to amplify the high-frequency transient input using the second process noise covariance under the spatial domain integral model, calculate the deviation between the predicted state derived from the spatial domain integral model and the actual observation value provided by the underlying sensor data stream, and generate kinematic innovation residuals.

5. The soil erosion characteristic data processing system based on filtering algorithm according to claim 4, characterized in that, The shadow detection module also includes: The data resampling unit is configured to resample the underlying sensor data stream acquired based on discrete time into a data sequence based on equal spatial intervals using a spline interpolation algorithm before the spatial domain triggering unit is triggered, and then input the data sequence into the spatial mapping unit.

6. The soil erosion characteristic data processing system based on filtering algorithm according to claim 1, characterized in that, The residual topology generation module includes: The feature extraction unit is configured to separate the Z-axis acceleration component and pitch angle component of the moving carrier in the body coordinate system from the kinematic information residual. The vector splicing unit is configured to combine the Z-axis acceleration component with the pitch angle component to construct the kinematic perturbation feature vector for the current cycle. The differential comparison unit is configured to calculate the difference between the same components of the current periodic kinematic perturbation feature vector and the historical baseline periodic kinematic perturbation feature vector on a unified geographic information system spatial coordinate axis, and then perform dimensionless weighted fusion of the difference results to generate a one-dimensional risk intensity difference sequence.

7. The soil erosion characteristic data processing system based on filtering algorithm according to claim 6, characterized in that, The system also includes: The state determination module is configured to perform conditional determinations on the difference sequence based on a preset threshold. When the difference in the difference sequence is less than the preset safety threshold, it is determined that the navigation environment of the current spatial segment has not changed and has not affected the navigation accuracy, and the current inspection route of the mobile vehicle is maintained. When the difference in the difference sequence is greater than or equal to the safety threshold and less than the preset danger threshold, it is determined that there is a navigation environment disturbance in the current spatial road segment, and local trajectory deviation warning information in the navigation kinematic state abnormality assessment result is generated and recorded. When the difference in the difference sequence is greater than or equal to the danger threshold, it is determined that there is a risk of navigation failure in the current spatial segment, triggering an emergency avoidance command and replanning the subsequent geographic information system route of the mobile vehicle.

8. The soil erosion characteristic data processing system based on filtering algorithm according to claim 7, characterized in that, The system also includes: The cloud mapping module is configured to map the navigation kinematic state anomaly assessment results to a navigation environment disturbance intensity heatmap. The feature feedback module is configured to extract the current periodic kinematic perturbation feature vector and the navigation environment perturbation intensity heatmap at discrete points in the spatial domain, and upload them to a remote server.

9. The soil erosion characteristic data processing system based on filtering algorithm according to claim 2, characterized in that, The inertial measurement unit uses a six-axis microelectromechanical system (MEMS) inertial sensor; Wheel-type odometers use photoelectric encoders; The satellite navigation receiver uses a real-time dynamic differential global navigation satellite system module; The data acquisition module, main navigation module, and shadow detection module are all integrated into the navigation computer of the mobile vehicle.

Images