Road condition intelligent assessment and resource allocation system integrated with mobile platform

Abstract

This invention relates to the field of maintenance management decision-making technology and discloses an intelligent pavement condition assessment and resource allocation system integrated into a mobile platform. The system includes sequentially connected modules for mobile data acquisition, data preprocessing, pavement condition assessment, resource optimization allocation, and decision support and feedback. The mobile data acquisition module collects multi-source data through a sensor array and dynamic sampling strategy; the preprocessing module generates a high-quality dataset through cleaning, spatiotemporal calibration, and standardization; the assessment module calculates the Pavement Health Index (PHI) based on a three-level model to accurately classify pavement condition levels; the resource optimization allocation module aims to maximize overall maintenance benefits by solving for the optimal resource allocation using a multi-constraint linear programming model; and the decision support and feedback module provides visualized decision display and corrects model parameters based on post-maintenance re-inspection data. The system improves the accuracy of pavement assessment and the scientific nature of resource allocation, adapting to the needs of large-scale road network maintenance management.

Description

Technical Field

[0001] This invention relates to the field of maintenance management decision-making technology, and more specifically to a road surface condition intelligent assessment and resource allocation system integrated into a mobile platform. Background Technology

[0002] In the field of highway maintenance and management, pavement condition detection and maintenance resource allocation have long relied on traditional technical models. Data collection mainly depends on dedicated inspection vehicles (equipped with large equipment such as laser pavement leveling instruments) or manual inspections, with sampling frequencies mostly set to a fixed value. In the data preprocessing stage, outlier detection uses a simple threshold method, multi-source data is calibrated using a single timestamp, and image preprocessing relies on filtering and edge detection techniques with fixed parameters. Pavement assessment integrates relevant indicators through weighted summation, with weight settings mostly based on human experience. Maintenance resource allocation is based on the area of ​​pavement defects or the priority determined manually. The systems are mostly locally deployed client / server architectures, adapted to a single operating system, and data transmission is mainly done through offline import or scheduled uploads.

[0003] However, with the expansion of road networks and the upgrading of maintenance needs, existing technologies have gradually revealed the following shortcomings: high data collection costs and low efficiency; limited battery life of mobile terminals; insufficient accuracy in multi-source data synchronization; poor data preprocessing accuracy; poor defect extraction in complex environments; strong subjectivity in pavement assessment; limited consistency with actual conditions; lack of comprehensive consideration in resource allocation, easily leading to allocation imbalances; low efficiency in large-scale road network decision-making; poor system adaptability; lack of maintenance effect feedback and model optimization linkage mechanisms; and insufficient management convenience.

[0004] Therefore, there is an urgent need for an intelligent road condition assessment and resource allocation system integrated into a mobile platform. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, the present invention provides a road condition intelligent assessment and resource allocation system integrated into a mobile platform to solve the problems existing in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a road condition intelligent assessment and resource allocation system integrated into a mobile platform, characterized in that it includes a mobile data acquisition module, a data preprocessing module, a road condition assessment module, a resource optimization allocation module, and a decision support and feedback module that are sequentially connected in communication. The mobile data acquisition module is configured to use a sensor array mounted on a mobile terminal to collect multi-source data from the road surface and generate a multi-source fusion acquisition dataset. The data preprocessing module is configured to clean, spatiotemporally calibrate, and standardize the multi-source fusion acquisition dataset to generate a standardized preprocessed dataset. The road surface condition assessment module is configured to process the standardized preprocessed dataset based on the road surface health hierarchical assessment mathematical model to generate a comprehensive health assessment dataset containing the road surface health index (PHI); the road surface health hierarchical assessment mathematical model is a three-level calculation model without weight coupling. The resource optimization allocation module is configured to maximize the total maintenance benefits by establishing a multi-constraint linear programming model with the pavement health index (PHI) as the core input parameter and solving it to generate a maintenance resource allocation decision dataset containing the optimal resource allocation amount. The decision support and feedback module is configured to display the maintenance resource allocation decision dataset and generate a feedback correction dataset based on the pavement re-inspection data after maintenance is completed. The feedback correction dataset is used to correct the parameters in the pavement condition assessment model.

[0007] The technical effects and advantages of this invention are as follows: 1. This invention uses a mobile terminal equipped with a sensor array, supports cross-platform operation, and significantly reduces costs; it innovates a dynamic sampling strategy, combined with a low-power mode to ensure battery life, and a dual calibration mechanism to improve data synchronization accuracy and avoid data loss; 2. This invention handles outliers and missing data through an adaptive 3σ criterion and multiple interpolation methods, introduces a filtering algorithm to correct positioning drift, employs adaptive threshold image preprocessing to adapt to complex environments, and standardizes processing to ensure data consistency; 3. This invention constructs a weightless, coupled three-level model, eliminating the need for manual weight setting, dynamically correcting road surface decay rates, supplementing anomaly replacement mechanisms for indicators, improving assessment adaptability and reliability, and achieving computational efficiency that meets the needs of diverse road sections; 4. This invention integrates multiple factors to construct a maintenance priority index, establishes a multi-constraint linear programming model to avoid allocation imbalance, and optimizes the solution algorithm to adapt to rapid decision-making in large-scale road networks; 5. This invention employs a B / S architecture to support high concurrency and real-time communication, with "real-time upload + offline caching" ensuring stable data transmission; it also features an innovative closed-loop feedback mechanism that continuously optimizes adaptability by re-checking data to correct model parameters. Attached Figure Description

[0008] Figure 1 This is a block diagram of the overall structure of the present invention; Figure 2 This is a flowchart of the mathematical model for the stratified assessment of road surface health in this invention. Figure 3 This is a flowchart of the solution process for the mathematical model of multi-constraint dynamic resource optimization allocation in this invention; Figure 4 This is a schematic diagram of the closed-loop optimization of the present invention. Detailed Implementation

[0009] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. In addition, the forms of the various structures described in the following embodiments are merely illustrative. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0010] Reference Figure 1-4 This invention provides a road condition intelligent assessment and resource allocation system integrated into a mobile platform, including a mobile data acquisition module, a data preprocessing module, a road condition assessment module, a resource optimization allocation module, and a decision support and feedback module.

[0011] The mobile data acquisition module inputs the raw signals from the mobile terminal sensors to the dynamic sampling and data synchronization processing unit, and outputs a multi-source fusion acquisition dataset; at the same time, it ensures the mobile terminal's battery life through a low-power control strategy, making it suitable for full-domain road network monitoring scenarios.

[0012] The raw signals from the mobile terminal sensors include triaxial MEMS accelerometer signals, CMOS image sensor signals, GPS / BeiDou positioning signals, and temperature and humidity sensor signals. The multi-source fusion acquisition dataset includes vibration signal dataset, image dataset, positioning time series dataset, and environmental parameter dataset.

[0013] The mobile data acquisition module is equipped with a three-axis MEMS accelerometer (model: ADXL345, range ±16g, resolution 13-bit), a CMOS image sensor (resolution ≥1080P, frame rate 30fps), a GPS / BeiDou dual-mode positioning module (positioning accuracy ±1m), and a temperature and humidity sensor (model: SHT30, accuracy ±0.3℃ / ±2%RH). The mobile data acquisition module adopts a dynamic sampling strategy, with the sampling frequency adaptively adjusted according to the driving speed. The vibration signal sampling frequency f a The driving speed v satisfies f a =200+5v (v unit: km / h, f) a Unit: Hz), image sampling frequency f i=2+0.1v (enabled when v≥10km / h), the positioning data sampling frequency is fixed at 10Hz to ensure data spatiotemporal alignment, data acquisition adopts interrupt trigger mode, and enters low power mode (current ≤5mA) when idle to avoid excessive consumption of mobile terminal battery life. The vibration signal dataset includes z-axis vertical acceleration time series, sampling timestamp, and driving speed related data; the image dataset includes continuous road surface frame images, image acquisition time, and acquisition location coordinates; the positioning time series dataset includes continuous latitude and longitude coordinates, positioning timestamp, and instantaneous speed data; the environmental parameter dataset includes real-time temperature, relative humidity, and acquisition timestamp.

[0014] It should be further explained that sensor data synchronization adopts a dual calibration mechanism of hardware interrupt + software timestamp. The accelerometer and positioning module achieve interrupt synchronization through the GPIO interface, with a synchronization error of ≤2ms. Due to the large amount of image data, an asynchronous acquisition + timestamp backtracking alignment method is adopted. The GPS timestamp embedded in the image frame header is associated with the positioning data to ensure the spatiotemporal matching accuracy of the image and location information. A multi-level caching mechanism is set up during data acquisition. Vibration signals are stored in real time using a circular buffer (1024 bytes) to avoid data loss. Image data adopts a segmented caching + local temporary storage strategy on the SD card. Verification and storage are performed every 100 frames of images. After verification, the cached data is deleted to improve storage stability. In addition to the idle low-power mode, low power control also dynamically adjusts the sampling frequency according to the battery level. When the battery level is below 20%, the sampling frequency is reduced by 30%, the image acquisition function is turned off, and only the core data acquisition of vibration, positioning and temperature and humidity is retained to ensure the acquisition of critical data and extend battery life.

[0015] The data preprocessing module inputs the multi-source fusion dataset into the data cleaning, spatiotemporal calibration and feature standardization processing unit, and outputs a standardized preprocessed dataset to provide high-quality data support for subsequent road condition assessment.

[0016] The standardized preprocessing dataset includes a denoised vibration signal dataset, a preprocessed image dataset, a calibrated positioning dataset, and a standardized feature dataset.

[0017] The data preprocessing module performs outlier detection based on the 3σ criterion, removing data that deviates from the mean by three times the standard deviation. It uses linear interpolation to fill in missing data (effective when the missing rate is ≤5%), and achieves spatiotemporal calibration of multi-source data through timestamp alignment, with time synchronization error controlled within 10 milliseconds. It uses Z-Score standardization to process feature data to ensure data dimension uniformity. For image data, it uses Gaussian filtering (kernel size 3×3, standard deviation σ=1.2) for noise reduction and Canny edge detection (threshold T1=50, T2=150) to extract disease contours, providing a basis for subsequent disease area calculation.

[0018] The denoised vibration signal dataset includes the filtered z-axis acceleration sequence, valid data segment identifiers, and data quality scores; the preprocessed image dataset includes denoised images, disease contour masks, and image size normalized data; the calibrated positioning dataset includes spatiotemporally aligned latitude and longitude coordinates and road segment length calculation results; and the standardized feature dataset includes the standardized values ​​of each sensor signal and feature dimension identifiers.

[0019] It should be further explained that the outlier detection adopts the adaptive 3σ criterion, which dynamically calculates the mean and standard deviation of the current window through a sliding window (window size of 50 data points) to avoid misjudging normal data due to abnormal vibration of the entire road segment; for local data segments with a missing rate of more than 5%, polynomial interpolation (3rd degree polynomial) is used instead of linear interpolation to improve the accuracy of missing data filling. During the spatiotemporal calibration process, a Kalman filter algorithm is introduced to smooth the drift problem in the positioning data. The state equation is X(k)=AX(k-1)+BU(k)+W(k), and the observation equation is Z(k)=HX(k)+V(k), where the state vector X includes position and velocity information. This filtering process reduces the positioning drift error to within ±0.3m. Z-Score standardization sets an adaptive offset for different sensor data characteristics. For data with large fluctuations, such as vibration signals, a logarithmic transformation is performed before standardization to avoid the influence of extreme values ​​on the standardization results. In image preprocessing, the thresholds T1 and T2 for Canny edge detection adopt an adaptive threshold calculation method, which is dynamically adjusted according to the mean and variance of the image grayscale histogram, i.e., T1=μ-0.5σ, T2=μ+1.5σ (μ is the grayscale mean, σ is the grayscale variance), improving the accuracy of lesion contour extraction under different lighting conditions.

[0020] The road surface condition assessment module inputs the standardized preprocessed dataset into the road surface health hierarchical assessment mathematical calculation unit and outputs a comprehensive health assessment dataset, which is then used as a core input parameter to the resource optimization and allocation module.

[0021] The comprehensive health assessment dataset includes a set of basic indicator calculation results, a set of functional coupling indices, a set of pavement health indices (PHI), and a set of pavement condition levels.

[0022] The road surface condition assessment module calculates the Road Surface Health Index (PHI) based on a hierarchical assessment mathematical model for road surface health. The assessment time is ≤100ms / road segment, and it supports flexible configuration of road segment lengths from 0.1 to 5km.

[0023] The set of basic index calculation results includes the specific values ​​and calculation accuracy of the smoothness index (IRI), the density of defects index (DDI), and the coefficient of friction index (FC); the set of functional coupling indexes includes the pavement functional coupling index (PCI) of each road segment and intermediate parameters in the calculation process; the set of pavement health indexes (PHI) includes the PHI value of each road segment, the calculation timestamp, and the associated road segment identifier; the set of pavement condition grades includes the grade judgment results (excellent / good / average / poor) of each road segment and the judgment basis.

[0024] It should be further explained that the mathematical model for the hierarchical assessment of road surface health adopts a pipelined computing architecture. The five steps of basic index calculation, normalization, functional coupling, decay rate correction, and health derivation process data from different road segments in parallel. The computational efficiency is optimized through a thread pool (number of threads = number of CPU cores × 2), ensuring that the assessment time for a single road segment is controlled within 100ms. The specific calculation process is as follows: Basic index calculation: Input the denoised vibration signal dataset, preprocessed image dataset, calibrated positioning dataset, and environmental parameter dataset from the standardized preprocessed dataset into the basic index calculation unit, and output the basic index calculation result set; The basic index calculation result set includes three sub-datasets: smoothness index (IRI), disease density index (DDI), and friction coefficient index (FC). Each sub-dataset contains the specific index value, calculation accuracy, associated road segment identifier, and calculation timestamp. The specific calculation process, parameter calibration method, and verification standards are as follows: The roughness index (IRI) subset dataset is generated based on the z-axis vertical acceleration time series from the denoised vibration signal dataset in the standardized preprocessed dataset. It reflects the degree of road surface roughness. The calculation process is "signal filtering - integration calculation - calibration correction," and the specific formula is as follows: ; Among them, a z (t) represents the time series of vertical acceleration along the z-axis, in m / s². 2 The data originated from a denoised vibration signal dataset, which underwent further high-frequency noise removal via a 5Hz low-pass Butterworth filter (order 4); a z0 This represents the reference acceleration, with units of m / s². 2 The mean z-axis acceleration of the mobile terminal is obtained through on-site calibration when the mobile terminal is traveling at a constant speed (v=40km / h) on a smooth asphalt road.

[0025] The calibration process is as follows: Select a standard, bump-free road section with a length of ≥1km, drive at a constant speed of 40km / h 3 times, collect z-axis acceleration data for each drive, and take the average of the 3 data as a. z0Its value range is [0.04, 0.06]; T represents the acquisition time, in seconds, which is calculated from the road segment length L (unit: km) and driving speed v (unit: km / h) in the calibrated positioning dataset, T=3600L / v; C1 represents the calibration coefficient, which is used to correct sensor errors and is obtained by synchronous comparison calibration with a professional laser flatness meter (model: LRZ-III).

[0026] The calibration process is as follows: Select 10 road sections with different smoothness (IRI range 0-10m / km), collect data using this system and a laser smoothness meter respectively, and establish a linear regression model IRI. 标准 =C1×IRI 系统 +b, let the intercept b=0, and solve for C1, whose value range is [0.95, 1.05]. The error between the generated IRI subset values ​​and the standard values ​​should be controlled within 3%.

[0027] The Depth of Pavement (DDI) sub-dataset is generated based on the preprocessed image dataset, calibrated location dataset, and environmental parameter dataset from the standardized preprocessed dataset. It reflects the degree of pavement distress distribution. The calculation process is "disease outline extraction - area conversion - weighted summation - environmental correction," and the formula is as follows: ; Where i represents the disease type identifier (1=pothole, 2=crack, 3=sinkhole); A i This represents the actual area of ​​the i-th type of disease, in meters. 2 It is calculated from the lesion contour mask in the preprocessed image dataset, and the conversion formula is: S pix,i The pixel area of ​​the i-th type of disease is represented by the number of pixels within the contour counted using OpenCV's findContours function; d represents the focal length of the image sensor in mm, obtained from the mobile terminal hardware parameters (e.g., iPhone 13 focal length d=4.1mm); f represents the horizontal pixel count of the image (f=1920 for 1080P images); k i The severity coefficient represents the i-th type of disease, calibrated by the depth / width of the disease, where pit k1=1.2 (depth ≥ 5cm), crack k2=0.8 (width ≥ 5mm), and subsidence k3=1.0 (depth ≥ 3cm). The coefficient is calibrated by measuring the disease dimensions with vernier calipers; A total This indicates the actual total area of ​​the assessed road segment, in m². 2 The length L (in km) and width W (in m, obtained via a map API, such as the Gaode Maps road width interface) in the calibrated positioning dataset are calculated. C2 represents the environmental correction coefficient, which is determined by the temperature and humidity data in the environmental parameter dataset. The correction logic is as follows: sunny day (relative humidity RH≤60%) C2=1.0, rainy day (RH>80% and precipitation) C2=1.1, snowy day (temperature T≤0℃ and RH>70%) C2=1.2.

[0028] The environmental correction coefficient was determined through an "experiment on image recognition error calibration under different environments." The experimental design referenced Section 5.3 of the "Technical Specification for Image Recognition of Road Surface Defects" (CJJ / T306-2023). The specific experimental steps were as follows: Five typical road surface defects (potholes, transverse cracks, longitudinal cracks, network cracks, and subsidence) were selected, and three size gradients were set for each defect to create standard defect samples. Sample images were collected using the CMOS image sensor of this system under three different environments: sunny, rainy, and snowy. Thirty sets of data were collected for each environment. Manual annotations were used as standard values ​​to calculate the defect area recognition error under different environments. A correlation model between the error and environmental parameters was established to determine the correction coefficient C2. The corrected recognition error was ensured to be ≤5%, and the error between the generated DDI subset data and the manually measured value was controlled within 5%.

[0029] Friction coefficient (FC) index subset generation: Calculated based on the denoised vibration signal dataset and calibrated positioning dataset from the standardized preprocessed dataset, reflecting the road surface's anti-skid performance. The calculation process is as follows: "Braking event detection - initial and final velocity extraction - braking distance calculation - friction coefficient derivation - road surface type correction," with the formula as follows: ; Where v1 represents the initial braking velocity in m / s, determined by the instantaneous velocity data in the calibrated positioning dataset. The braking trigger condition is that the absolute value of the deceleration in the denoised vibration signal dataset is ≥2 m / s². 2 The velocity at the moment of triggering is v1; v2 represents the final braking velocity, in m / s, where the absolute value of the deceleration is ≤0.5 m / s². 2 The instantaneous velocity at time t; d represents the braking distance in meters, calculated from the position coordinates in the calibration dataset during braking, using the Haversine formula to calculate the distance between the two points: , where R is the Earth's radius (6371000m). for and The difference, , Latitude of the braking start and end points for and The difference, , Longitude; g is the acceleration due to gravity, in m / s². 2The value of g is 9.80665 m / s. 2 C3 is the road surface type correction coefficient, which is determined by the road surface texture features in the preprocessed image dataset. For asphalt pavement (texture mean ≥ 80), C3 = 1.0; for cement pavement (texture mean 40-80), C3 = 0.95; and for gravel pavement (texture mean < 40), C3 = 1.05. The texture mean is calculated by the standard deviation of the image grayscale value. The error between the generated FC subset values ​​and the measured values ​​of the pendulum friction instrument (model: BM-III) should be controlled within 4%.

[0030] Normalization and Functional Coupling Calculation: The calculation result set of the basic indicators is input into the normalization and functional coupling unit, and the functional coupling index set is output. The functional coupling index set includes the pavement functional coupling index (PCI) of each road segment, normalized intermediate parameters, and coupling calculation process data. The specific calculation details are as follows: Min-max normalization is performed on the IRI, DDI, and FC subsets of the basic indicator calculation result set to eliminate the influence of dimensions. The normalization formula is as follows: Where x represents the original value of the basic indicator; x min x max These represent the minimum and maximum thresholds of the indicator, respectively, calibrated through industry standards and extensive measured data: IRI's x min =0、x max =10; DDI's x min =0、x max =0.2; x of FC min =0.2、x max =0.8.

[0031] After normalization, the IRI is obtained. n (Normalized Index of Smoothness), DDI n (Normalized index of disease density), FC n (Normalized exponent of friction coefficient), all satisfy 0 < x n <1, when x <x min time x n =0, when x>x max time x n =1.

[0032] Based on the normalization result, unweighted coupled derivation operations are performed, and the co-representation between indicators is achieved by using the square root of the product. The coupling formula is as follows: In this context, PCI is the pavement function coupling index, which satisfies 0 < PCI < 1. The larger the PCI value, the better the pavement function.

[0033] The actual implementation of unweighted coupling in code is as follows: It is implemented using the Python programming language, and the IRI is calculated using the math.sqrt() function. n With DDI n The square root of the product, then FC n The core code snippet for performing multiplication is "PCI=math.sqrt(IRI)". n DDI n ) FC n ".

[0034] The calculated PCI values ​​are stored in 64-bit floating-point format and transmitted to the next module through a standardized data interface. CRC32 checksums are used during data transmission to ensure data integrity. This system has been verified through simulation in MATLAB R2023a with a simulated road network of 100 kilometers. Simulation results show that the computation time for unweighted coupled operations is ≤5ms / road segment, and the coupling results match the actual road surface functional state by ≥92%.

[0035] The rationale for the unweighted coupling is demonstrated as follows: Pearson correlation analysis shows that the correlation coefficient between smoothness (IRI) and distress density (DDI) is r=0.78 (significance level p<0.01), indicating a strong positive correlation between the two. That is, the more pavement distress, the worse the smoothness, which naturally provides a basis for synergistic characterization. The friction coefficient (FC) is weakly negatively correlated with the former two (r=-0.32, p<0.05), reflecting the independent correlation between anti-skid performance and pavement damage state. Based on this correlation characteristic, the unweighted operation of "product square root + product" can achieve synergistic amplification of strongly correlated indicators and supplementary characterization of weakly correlated indicators, and objective functional fusion results can be obtained without subjectively setting weights.

[0036] Decay Rate Correction and Health Derivation Calculation: The Functional Coupling Index (PCI) set and the road network historical inspection dataset are input into the decay rate correction and health derivation calculation unit, which outputs the Pavement Health Index (PHI) set and the pavement condition level set. The road network historical inspection dataset includes Pavement Structure Strength Index (SSI) data, pavement construction completion date data, and pavement material type data. The Pavement Health Index (PHI) set and the pavement condition level set together constitute the core content of the comprehensive health assessment dataset. Specific calculation details are as follows: The pavement decay rate is corrected using the PCI value of the Functional Coupling Index set, with the correction formula being... ,in, The basic decay coefficient, in units of 1 / year, is derived from the pavement material type data in the historical road network detection dataset. The specific value is determined according to Article 6.3.2 of the "Specifications for Design of Highway Asphalt Pavement" (JTGD50-2017): γ0=0.05 for cement concrete pavement (C30 strength grade) and γ0=0.08 for asphalt concrete pavement (AC-13 gradation). γ' is the corrected decay coefficient, in units of 1 / year. Its physical meaning is that the worse the pavement functional condition (the smaller the PCI), the faster the decay rate, enabling the evaluation model to dynamically adapt to changes in pavement condition.

[0037] The road health index (PHI) is derived based on the modified decay coefficient, and the formula is as follows: SSI stands for Structural Strength Index, which is dimensionless and derived from historical road network monitoring datasets. It is calculated according to the method specified in Section 4.2 of the "Highway Technical Condition Assessment Standard" (JTG5210-2018), specifically by collecting pavement deflection values ​​using a falling weight deflectometer (FWD) and applying the formula... Calculation yields ( SSI represents the pavement deflection value under 0.75 times the standard axle load (in mm), with a range of 0 < SSI < 1. SSI = 0 represents complete structural failure, and SSI = 1 represents intact structural integrity. t represents the pavement service life in years, calculated from the road construction completion date and assessment date in the historical road network monitoring dataset. 评估 -t 竣工 e is the natural constant (e = 2.71828).

[0038] The Road Health Index (PHI) value must satisfy 0 < PHI < 1. Road surface condition levels are classified based on PHI values, and the classification criteria strictly adhere to Article 5.1.2 of the "Highway Technical Condition Assessment Standard" (JTG5210-2018), satisfying the following: The comprehensive health assessment dataset is output to the resource optimization and allocation module via a JSON interface, serving as a core parameter for priority calculation and achieving positive data connection between assessment and allocation.

[0039] In the calculation of basic indicators, in order to improve the calculation accuracy, the Simpson integral method is used for the integral operation of vibration signals. Compared with the traditional rectangular integral method, the integration error is reduced by more than 40%. In the calculation of disease density, for the overlapping areas generated by image stitching, the IOU (Intersection over Union) threshold method (IOU ≥ 0.5 is judged as overlapping) is used to remove duplicates and avoid repeated statistics of disease area.

[0040] In functionally coupled calculations, to prevent PCI value distortion due to anomalies in any basic indicator, a mechanism for judging indicator validity is implemented. When the calculation accuracy of any basic indicator falls below 90%, the average value of the indicator from historical road sections of the same type is used as a substitute, ensuring the reliability of PCI calculations. The decay rate correction introduces a seasonal correction factor, adjusting the correction coefficient γ' based on temperature and precipitation data from different seasons (spring, summer, autumn, and winter). season =γ'×K season (K) season (This is a seasonal factor, with a value range of 0.9-1.1), used to improve the accuracy of PHI value assessment under different climatic conditions.

[0041] The resource optimization and allocation module inputs the comprehensive health assessment dataset and the road network basic attribute dataset into the multi-constraint linear programming solution unit, and outputs the maintenance resource allocation decision dataset.

[0042] The road network basic attribute dataset includes data on road segment damage area, average daily traffic flow, road segment length, and safety level; the maintenance resource allocation decision dataset includes the maintenance priority index (PI), optimal resource allocation, maintenance sequence, and completion deadline for each road segment.

[0043] The resource optimization and allocation module solves the resource allocation problem based on a linear programming algorithm, using the simplex method (iteration accuracy ε=10). -4 The optimization model is solved with a time ≤500ms / 100 road segments. The maintenance priority index (PI) set includes the PI value of each road segment and intermediate parameters during the calculation process; the optimal resource allocation data includes the funding allocation amount and allocation basis for each road segment; the maintenance sequence data includes a list of road segments arranged in descending order of PI value; the completion time limit data includes the latest repair time of each road segment and the calculation basis. The specific calculation logic and technical details are as follows: Maintenance Priority Index (PI) Calculation: The Road Surface Health Index (PHI) set, traffic flow data, and safety level data from the road network basic attribute dataset are input into the priority derivation calculation unit, which outputs the Maintenance Priority Index (PI) set. The Maintenance Priority Index (PI) set is a core subset of the maintenance resource allocation decision dataset, containing PI values ​​for each road segment, basic priority factors, and multi-factor coupling intermediate parameters. Specific calculation details are as follows: The basic priority factor is derived from the PHI value based on the Road Health Index (PHI) set, and the formula is as follows: , where P base As a basic priority factor, PHI is the pavement health index; Minimum value ( =10 -6 ), used to avoid infinity when PHI=0, ensuring computational stability; Pbase The physical meaning of PHI is that the worse the road surface condition (the smaller the PHI), the higher the foundation priority, with a value range of 1.000001 < P. base <2.5.

[0044] PI is obtained by performing an unweighted coupling operation between the basic priority factor and three influencing factors—traffic flow, safety hazards, and repair time sensitivity—from the road network basic attribute dataset. The formula is as follows: , where V n This is the normalized value of the daily average traffic flow for the road segment. The original traffic flow V (unit: vehicles / day) comes from traffic flow data in the road network basic attribute dataset, obtained through traffic flow monitoring equipment or map big data interface. The normalization formula is min-max normalization, where x is the normalization factor of V. min =500 vehicles / day, x max =50,000 vehicles / day, satisfying 0 < V n <1; S_n is the normalized value of the safety hazard coefficient. The original safety hazard level S comes from the safety level data in the road network basic attribute dataset, which is divided into 3 levels (Level 1: Roads around schools / hospitals, Level 2: Main roads, Level 3: Secondary roads), and is quantified using the level quantification formula. (S) q After quantizing the values, perform min-max normalization to satisfy 0 < S. q <1;T r To determine the time sensitivity (unitless), it is set according to the pavement condition level corresponding to the Road Health Index (PHI): when PHI < 0.4 (poor), T r =1.0 (emergency repair), 0.4≤PHI<0.6 (normal) T r =0.6 (routine repair), T when PHI≥0.6 (good / excellent) r =0.3 (delayed repair) to ensure priority treatment of urgent diseases.

[0045] The coupling logic is as follows: traffic flow and safety hazards have a synergistic impact (road sections with high traffic flow and high safety level need to be maintained first). The synergistic amplification of the two is achieved by taking the square root of the product, and then multiplying it with the basic priority factor and the time sensitivity factor to comprehensively represent the maintenance priority; the PI value satisfies 0.3000003 < PI < 2.5, and the larger the PI value, the higher the maintenance priority.

[0046] To maximize the overall maintenance benefits, a linear programming model is established: Objective function: Resource utilization efficiency data from the Maintenance Priority Index (PI) set, Pavement Health Index (PHI) set, and road network basic attribute dataset are input into the objective function construction unit. The output is an objective function that maximizes the total maintenance benefit and a set of related parameters. The set of related parameters includes the resource utilization efficiency, benefit coefficient, and variable identifiers of the resource allocation amount to be solved for each road segment. The specific formula is as follows: Where m is the number of road segments to be maintained (m is greater than 1, supporting a maximum of 1000 road segments in parallel calculation); j is the road segment identifier (1,2,...,m); PI j R is the maintenance priority index for the j-th road segment (derived from the maintenance priority index (PI) set); j The amount of maintenance resources allocated to the j-th road segment is expressed in ten thousand yuan, and is the variable to be solved in the model. The resource utilization efficiency (unitless) of the j-th road segment is derived from resource utilization efficiency data in the road network basic attribute dataset, and is obtained from historical maintenance data statistics. =ΔPHI j,hist / R j,hist ΔPHI j,hist R represents the PHI increase value after historical maintenance of road segment j. j,his The amount of resources invested in historical maintenance should meet the requirement of 0.8 < <1.0, the higher the resource utilization efficiency, the better the benefit of the same resource input; K j The maintenance benefit coefficient (unitless) for the j-th road segment is given by the following formula: PHI j Let K be the pavement health index of road segment j (derived from the pavement health index (PHI) set). j The physical meaning is that the worse the road surface condition (PHI), the better. j The smaller the size, the greater the potential for improved benefits after maintenance, satisfying 0 < K. j <0.6; Z represents the total maintenance benefit (unit: 10,000 yuan) -1 The Z-value is a quantifiable indicator of maintenance efficiency; a larger Z-value indicates a higher overall efficiency in the allocation of maintenance resources. The function logic is as follows: Through PI... j Ensure that high-priority road segments receive resources first, through K j Ensure that resources are allocated to road sections with greater potential for efficiency improvement, through To ensure that resources are concentrated on road sections with high utilization efficiency and to achieve optimal allocation of maintenance resources; all parameters in the objective function are obtained through quantitative mathematical formulas, without subjective parameter setting, to ensure the objectivity and scientific nature of the optimization results.

[0047] Constraints: This step inputs the Maintenance Priority Index (PI) set, Pavement Health Index (PHI) set, road network basic attribute dataset, and maintenance resource budget dataset into the constraint construction unit, outputting a multi-constraint set. The maintenance resource budget dataset includes data on the total available maintenance resources for the current period and data on the unit cost of repairing road defects. The multi-constraint set is used to limit the solution range of the optimization model, ensuring the rationality, feasibility, and balance of resource allocation, as detailed below: Total resource constraints: , where R total The total available maintenance resources for the current period, expressed in ten thousand yuan, are derived from the maintenance resource budget dataset and determined by the maintenance management department's annual budget. R total >0; This constraint ensures that resource allocation does not exceed the total budget limit, thus avoiding resource overspending.

[0048] Minimum resource requirement constraints: , where R min,j The minimum resource requirement for repairing road segment j, expressed in ten thousand yuan, is calculated by coupling the pavement health index (PHI) value with the defect area data from the road network basic attribute dataset and the unit repair cost data from the maintenance resource budget dataset. The formula is as follows: A disease,j The total defect area of ​​road segment j is expressed in m². 2 The data is derived from the defect area data in the road network basic attribute dataset and is obtained during the DDI calculation process; P unit,j The unit cost of repairing road defects for section j is given in ten thousand yuan / m. 2 Derived from the maintenance resource budget dataset, determined based on the type of damage: Pothole P unit,j =0.02, crack P unit,j =0.005, subsidence P unit,j =0.015; =10 -6 It is the minimum value; introduce This ensures that road sections in poorer condition receive more adequate minimum resource guarantees, guaranteeing effective repair of damage.

[0049] Repair time limit constraints: , where t repair,j The estimated repair time for road segment j, in days, is calculated from the damaged area data, repair efficiency data, and resource allocation data in the road network basic attribute dataset. , t unit,j The unit of time for repairing defects is days / m. 2 The data originates from basic road network attribute data, including potholes (t). unit,j =0.01, crack t unit,j =0.002, sinking tunit,j =0.008, E j The maintenance team efficiency coefficient (unitless) is derived from road network basic attribute data and is calculated from historical data. j The range is [0.8, 1.2]; t max,j The latest repair deadline for road segment j, in days, is the same as the PI of the maintenance priority index (PI) set. j Positive correlation, the formula is PI j The larger (higher priority), t max,j The smaller the size, the more likely it is to ensure that emergency road sections are repaired first.

[0050] Resource balance constraint: , where L j is the length of the j-th road segment, in km, derived from the road segment length data in the road network basic attribute dataset and calculated from GPS positioning data; The total length of the road sections to be maintained in the entire region is expressed in km. This constraint ensures that the resource allocation per kilometer of road section does not exceed twice the average level of the entire region, thus avoiding excessive concentration of resources.

[0051] Model Solution: The above linear programming model is solved using the improved simplex method. The solution steps are as follows: S1 transforms the objective function into standard form (maximization is transformed into minimization, and constraints are unified into equality constraints). S2 constructs the initial feasible basis matrix and determines the initial basic and non-basic variables; S3 calculates the test number to determine whether the current solution is the optimal solution (the optimal solution is when all test numbers are ≤0). If S4 does not reach the optimal value, select the entering and leaving variables and perform a basis transformation. S5 repeats steps S3-S4 until the optimal solution is obtained, with iteration accuracy... =10 -4 .

[0052] After solving, output the optimal resource allocation R for each road segment. Maintenance sequence (according to PI) j (arranged in descending order) and completion time t max,j Together, they constitute the maintenance resource allocation decision dataset.

[0053] Closed-loop optimization implementation method: Feedback Data Collection: After maintenance is completed, the system automatically generates a "Re-inspection Task Sheet" and pushes it to the mobile terminal of maintenance management personnel. The task sheet includes the start and end coordinates of the road section to be re-inspected, the data collection speed requirement (v=40±5km / h), and the data collection time window (within 72 hours after maintenance is completed). Management personnel use the same mobile terminal's data collection module to complete data collection along the designated road section. The data collection process uses the same sampling parameters (dynamic sampling frequency, sensor configuration, etc.) as the initial assessment to ensure data comparability.

[0054] Data synchronization and processing: After the data collection is completed, the mobile terminal uploads the re-inspection data to the server via the 4G / 5G network. The preprocessing module on the server side automatically cleans, calibrates, and standardizes the re-inspection data to generate a "re-inspection standardized preprocessing dataset".

[0055] Parameter correction and update: The evaluation module calls the same three-level evaluation model to calculate the pavement health index (PHI) of the re-inspected road section. after,j ), and calculate the actual lift value ΔPHI actual,j =PHI after,j -PHI before,j (PHI) before,j (The PHI value before maintenance was retrieved from the historical database); the actual improvement value was compared with the estimated improvement value. By comparison, through formula correction The updated fundamental decay coefficient γ was calculated. 0,new The system automatically adds γ 0,new Write the parameter configuration library into the evaluation model, overwriting the original γ. 0,old It also records correction logs (including correction time, road segment identification, parameter values ​​before and after correction, and error values); subsequent evaluation calculations for newly added road segments automatically call the updated parameters, realizing a two-way interactive closed loop of evaluation and allocation.

[0056] It should be further explained that the multi-constraint linear programming solution employs an improved simplex method. Through a preprocessing stage, constraints are filtered to eliminate redundant constraints (such as temporary optimizations that ignore resource balancing constraints when the total resource constraint is much greater than the sum of the minimum resource requirements of each road segment), thus improving solution efficiency. To address the solution requirements of large-scale road networks (more than 1000 road segments), a distributed solution architecture is introduced. The road network is divided into multiple sub-problems by region, and each sub-problem is assigned to different computing nodes for parallel solving. After the solution is completed, the results are integrated through a global coordination mechanism to ensure optimal global resource allocation.

[0057] In the calculation of the Maintenance Priority Index (PI), traffic flow data is processed using a time-weighted average, with the traffic flow of the past 30 days accounting for 70% of the weight and the traffic flow of the remaining time of the past 90 days accounting for 30% of the weight, to avoid the impact of short-term traffic flow fluctuations on the PI value.

[0058] After calculating the optimal resource allocation, a resource adjustment buffer mechanism is set up. When the difference between the optimal resource allocation of a certain road segment and the single-kilometer resource allocation of adjacent road segments exceeds 50%, a secondary optimization is triggered. Under the premise that the total resource constraint remains unchanged, the allocation is finely adjusted to improve the balance of resource allocation.

[0059] The decision support and feedback module inputs the maintenance resource allocation decision dataset into the visualization and information push unit, outputs the maintenance decision visualization results and early warning information; at the same time, it collects the actual road surface measurement data after maintenance is completed, generates a feedback correction dataset, and inputs it into the road surface condition assessment module.

[0060] The feedback correction dataset includes the pavement health index, actual PHI improvement value, and model parameter correction suggestions after maintenance; the maintenance decision visualization results include a heat map of PHI value distribution in the road network, a resource allocation bar chart, and a maintenance progress tracking table; the early warning information includes maintenance reminders for high-priority road sections, repair time limit warnings, and resource shortage warnings.

[0061] The decision support and feedback module adopts a B / S architecture, with the server deployed on a cloud server. It supports high-concurrency data processing and storage, and enables real-time communication between the mobile terminal and the server via WebSocket (communication latency ≤200ms). The visualization interface uses the ECharts chart library to visualize PHI values ​​and resource allocation schemes. Warning information is pushed to the maintenance management personnel's terminal via JPush service.

[0062] The system collects road surface inspection data after maintenance (collected using the same mobile terminal at the same driving speed), calculates the actual PHI improvement value, and uses this data to correct the evaluation model parameters. The system deployment and operating environment are as follows: For mobile terminal deployment, it is suitable for Android 10.0 and above systems (RAM≥4GB, ROM≥64GB) and iOS 14.0 and above systems (iPhone 8 and above models), with an installation package size ≤50MB and runtime memory usage ≤200MB; the mobile data acquisition module runs in background service mode, supporting continuous background data collection. For server deployment, the recommended configuration can meet the needs of up to 1000 mobile terminals simultaneously uploading data online, adopting a highly available database architecture that supports long-term storage and backup of historical data.

[0063] The data synchronization mechanism adopts a "real-time incremental upload + offline caching" mode. When the network is unobstructed, data is uploaded in real time (upload frequency is consistent with sampling frequency). When the network is interrupted, the data is cached locally (maximum cache capacity 10GB). The data is automatically re-uploaded after the network is restored. The server uses a message queue to process concurrent upload requests to ensure data transmission stability.

[0064] It should be further explained that the visualization supports multi-dimensional interactive queries. Managers can filter data by road segment number, pavement grade, maintenance priority, etc. Clicking on a road segment in the heatmap allows viewing a detailed assessment report for that segment (including basic indicator values, PHI calculation process, resource allocation basis, etc.). Data export is supported, allowing the export of maintenance decision reports in Excel format. These reports include key information such as road segment information, maintenance priority, resource allocation, and completion deadlines, facilitating offline archiving and approval. Early warning information is pushed using a tiered mechanism. High-priority warnings (such as road segments with PHI < 0.4 and nearing their repair deadline) utilize a triple approach: app push notification + SMS notification + voice alert. Medium and low-priority warnings only use app push notifications, ensuring that no critical warning information is missed.

[0065] During the feedback correction process, parameter correction thresholds are set. When the absolute value of the deviation between the actual and estimated PHI improvement values ​​is ≤10%, the basic decay coefficient γ0 is not corrected; when the absolute value of the deviation is >10% and ≤20%, linear correction is used; when the absolute value of the deviation is >20%, the model parameter recalibration process is triggered, and the initial value of γ0 is re-determined by collecting more data from similar road segments to avoid parameter distortion caused by a single deviation. Data synchronization adopts a breakpoint resume mechanism. After each data block (1MB in size) is uploaded, a verification code is returned. After the server verifies the code, the upload progress is recorded. Once the network is restored, the upload continues from the breakpoint, improving the upload efficiency of large-volume data (such as continuous image data).

[0066] The foregoing describes exemplary embodiments of this application. It should be understood that the above exemplary embodiments are not restrictive but illustrative, and the scope of protection of this application is not limited thereto. It should be understood that those skilled in the art can make modifications and variations to the embodiments of this application without departing from the spirit and scope of this application, and such modifications and variations should be within the scope of protection of this application.

Claims

1. A road condition intelligent assessment and resource allocation system integrated into a mobile platform, characterized in that, It includes a mobile data acquisition module, a data preprocessing module, a road condition assessment module, a resource optimization and allocation module, and a decision support and feedback module that are connected in sequence via communication. The mobile data acquisition module is configured to use a sensor array mounted on a mobile terminal to collect multi-source data from the road surface and generate a multi-source fusion acquisition dataset. The mobile data acquisition module adopts a dynamic sampling strategy, wherein the sampling frequency of the vibration signal is adaptively adjusted according to the driving speed, image acquisition is enabled when the speed reaches a threshold, and data acquisition adopts an interrupt-triggered and idle low-power mode. The data preprocessing module is configured to clean, spatiotemporally calibrate, and standardize the multi-source fusion acquisition dataset to generate a standardized preprocessed dataset. The road surface condition assessment module is configured to process the standardized preprocessed dataset based on the road surface health hierarchical assessment mathematical model to generate a comprehensive health assessment dataset containing the road surface health index (PHI). The mathematical model for the hierarchical assessment of pavement health is a three-level calculation model without weight coupling. This three-level calculation model includes: calculating the smoothness index IRI, the distress density index DDI, and the friction coefficient index FC based on vibration, image, and positioning data; after normalizing the IRI, DDI, and FC, performing unweighted coupling through product and square root operations to obtain the pavement functional coupling index PCI; using the PCI to correct the pavement decay rate, and combining it with the pavement structural strength index SSI and service life, the pavement health index PHI is calculated. The resource optimization allocation module is configured to maximize the overall maintenance benefit by establishing a multi-constraint linear programming model with the pavement health index (PHI) as the core input parameter, and generating a maintenance resource allocation decision dataset containing the optimal resource allocation amount. The resource optimization allocation module is further configured to: calculate the maintenance priority index (PI) for each road segment based on the PHI, road segment traffic volume, and safety level; construct an objective function using the PI, resource utilization efficiency, and maintenance benefit coefficient derived from the PHI; and construct and solve the multi-constraint linear programming model with constraints including total resource budget, minimum resource demand, repair time limit, and resource balance to obtain the optimal resource allocation amount. The decision support and feedback module is configured to display the maintenance resource allocation decision dataset and generate a feedback correction dataset based on the pavement re-inspection data after maintenance is completed. The feedback correction dataset is used to correct the parameters in the pavement condition assessment model.

2. The intelligent road condition assessment and resource allocation system integrated into a mobile platform according to claim 1, characterized in that, The specific calculation method of the unweighted coupling is as follows: calculate the square root of the product of the smoothness normalization index and the defect density normalization index, and then multiply it by the friction coefficient normalization index to obtain the pavement function coupling index PCI.

3. The intelligent road condition assessment and resource allocation system integrated into a mobile platform according to claim 1, characterized in that, The maintenance priority index PI is calculated as follows: based on the PHI, a basic priority factor is calculated, the square root of the product of the normalized value of traffic flow and the normalized value of the safety hazard coefficient is calculated, and then the basic priority factor, the square root, and the repair time sensitivity are multiplied to obtain the PI.

4. The intelligent road condition assessment and resource allocation system integrated into a mobile platform according to claim 1, characterized in that, The mobile data acquisition module includes a three-axis MEMS accelerometer, a CMOS image sensor, a positioning module, and a temperature and humidity sensor.

5. The intelligent road condition assessment and resource allocation system integrated into a mobile platform according to claim 1, characterized in that, The execution logic of the data preprocessing module includes: outlier detection and imputation based on the 3σ criterion, spatiotemporal calibration of multi-source data through timestamp alignment, feature standardization using the Z-Score method, and noise reduction and edge detection of image data to extract lesion contours.

6. The intelligent road condition assessment and resource allocation system integrated into a mobile platform according to claim 1, characterized in that, The process of generating the feedback correction dataset by the decision support and feedback module includes: collecting re-inspection data of the road surface after maintenance and calculating the actual PHI improvement value, comparing it with the model-predicted improvement value, and correcting the basic decay coefficient in the road surface condition assessment mathematical model accordingly.

7. The intelligent road condition assessment and resource allocation system integrated into a mobile platform according to claim 1, characterized in that, The system adopts a B / S architecture, with the server supporting high-concurrency data processing and storage, and communicating with mobile terminals in real time via WebSocket; the mobile terminals support Android and iOS systems, and the data acquisition module operates in a background service mode.

Images