Intelligent evaluation system for highway marking quality based on image analysis

Abstract

The application provides a highway marking quality intelligent evaluation system based on image analysis, and relates to the technical field of road facility monitoring. The system comprises a detection unit, a marking quality evaluation unit, a marking wear prediction unit, a GPS positioning unit, a vehicle driving information unit, a control unit and a remote central control unit. The marking quality evaluation unit is constructed based on differential geometry theory and comprises a curvature flow edge detection module, a marking geometric characteristic manifold representation module and a multi-scale differential invariant evaluation module. The system regards the marking as a two-dimensional manifold, calculates differential geometric quantities such as Gaussian curvature, average curvature and shape index, constructs a geodesic line distance measurement on the characteristic manifold, and evaluates the integrity, visibility and retroreflective performance of the marking. The marking wear prediction unit predicts the service life of the marking based on a traffic flow information and environmental characteristic mapping relationship model, and generates a maintenance suggestion.

Description

Technical Field

[0001] This invention relates to the field of road infrastructure monitoring technology, specifically to an intelligent assessment system for highway marking quality based on image analysis, used to automatically assess and predict quality parameters such as the integrity, visibility, and reflectivity of highway markings. Background Technology

[0002] Road markings are a crucial infrastructure for road traffic safety, and their quality directly impacts driving safety and the normal operation of intelligent driving systems. Traditional road marking quality assessments primarily rely on manual inspections, which suffer from low efficiency, strong subjectivity, and an inability to provide quantitative evaluation. Existing automated assessment methods often employ simple image analysis based on color and reflectivity, which struggles to adapt to complex road conditions and suffers from insufficient accuracy and stability. Especially under complex conditions such as varying lighting, road surface contamination, and localized wear of markings, traditional assessment techniques often fail to accurately identify and evaluate the quality status of road markings.

[0003] With the development of technologies such as differential geometry and computer vision, it has become possible to use the geometric characteristics of paving marks for quality assessment. However, there is currently a lack of a technical solution that deeply integrates differential geometry theory with paving mark quality assessment, capable of revealing the quality status of paving marks from their geometric essence and achieving high-precision, highly adaptable paving mark quality assessment. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent evaluation system for highway marking quality based on image analysis, which integrates differential geometry theory and image analysis technology to achieve accurate evaluation and prediction of highway marking quality.

[0005] This invention proposes an intelligent evaluation system for highway marking quality based on image analysis, comprising:

[0006] The detection unit is used to acquire road marking images and perform image preprocessing;

[0007] A road marking quality assessment unit, communicatively connected to the detection unit, is used to receive preprocessed road marking images sent by the detection unit. The road marking quality assessment unit includes:

[0008] The curvature flow edge detection module is used to treat the preprocessed marking image as a two-dimensional manifold, construct the curvature flow equation, extract marking edge features through multi-scale representation, and output the edge feature point set;

[0009] The grading geometric property manifold representation module is communicatively connected to the curvature flow edge detection module. It is used to receive the edge feature point set, construct the parameterized representation of the grading manifold, calculate the first and second basic forms of the grading surface, extract Gaussian curvature and average curvature features, and generate a geometric property dataset.

[0010] The multi-scale differential invariant evaluation module is communicatively connected to the marking geometric characteristic manifold characterization module. It is used to receive the geometric characteristic dataset, calculate multi-order differential invariants, fuse multi-scale features, construct geodesic distance metrics on the feature manifold, evaluate the integrity, visibility, and reflectivity of the marking, and output quality evaluation results.

[0011] The road marking wear prediction unit is communicatively connected to the road marking quality assessment unit. It is used to receive the quality assessment results, predict the service life of the road markings, and generate maintenance suggestions.

[0012] Preferably, the curvature flow edge detection module includes:

[0013] The image normalization submodule is used to convert the input image into a standard resolution and brightness range, and establish a mapping relationship from pixels to physical size;

[0014] The curvature calculation submodule is used to construct the image surface and calculate the principal curvature and average curvature at each point on the surface;

[0015] The evolution control submodule is used to control the evolution speed of the surface according to the curvature value, so that the high curvature region evolves slowly and the low curvature region evolves quickly and smoothly.

[0016] The multi-scale edge representation submodule is used to generate image representations at different scales and preserve edges that are stable at multiple scales by comparing the positional offsets of edges at different scales.

[0017] The arc length optimization submodule is used to define an arc length functional that includes a curvature term. This functional is used to optimize edge curves while maintaining their topology.

[0018] Preferably, the caliper geometric characteristic manifold characterization module includes:

[0019] The curve parameterization submodule is used to parameterize the edge curve into an arc length parameter representation, establish a local coordinate system, and divide the marking area into a regular grid.

[0020] The differential geometry computation submodule is used to compute the metric tensor, the first fundamental form and the second fundamental form on a discrete grid, describing the intrinsic and extrinsic geometric properties of the gauging surface;

[0021] The geometric invariant extraction submodule is used to calculate geometric invariants that are independent of the observation viewpoint, including Gaussian curvature, mean curvature, and shape exponent;

[0022] The geometric feature mapping submodule is used to establish a multi-dimensional feature space, generate geometric feature vectors for each region of the datum, and identify regions with abnormal geometric characteristics.

[0023] The quality parameter space construction submodule is used to define basic quality dimensions such as integrity, flatness, adhesion, and uniformity, mapping geometric characteristics to the quality parameter space.

[0024] Preferably, the multi-scale differential invariant evaluation module includes:

[0025] The differential invariant calculation submodule is used to calculate first-order invariants, second-order invariants, and higher-order invariants.

[0026] The multi-scale feature fusion submodule is used to construct multiple scale levels covering from details to the whole, decompose the features at each scale into basic components, and reconstruct the feature vector by weighting them according to their importance.

[0027] The Feature Manifold Construction Submodule is used to define similarity measures between features, construct the Riemannian metric of the feature manifold, and establish geodesic connections between feature points.

[0028] The quality score generation submodule is used to define the measurement function of quality difference, establish the mapping from geodesic distance to quality score, and generate integrity score, visibility score and reflectivity score;

[0029] The anomaly pattern recognition submodule is used to extract major variation patterns based on principal component analysis on the manifold, and to identify and classify anomalous regions.

[0030] As a preferred option, it also includes:

[0031] The GPS positioning unit is used to obtain the vehicle's real-time location and direction of travel.

[0032] The vehicle driving information unit is used to obtain the vehicle's real-time speed, acceleration, and steering angle;

[0033] The control unit is communicatively connected to the GPS positioning unit and the vehicle driving information unit, and is used to determine whether to activate the lane marking detection based on the vehicle position and driving information, and to send a detection command to the detection unit.

[0034] The detection unit includes a camera acquisition unit, a light detection unit, a color compensation unit, and an image preprocessing unit. The camera acquisition unit acquires the marking image according to the detection command. The light detection unit acquires the real-time light value. The color compensation unit adjusts the color and intensity of the light source according to the real-time light value. The image preprocessing unit performs noise filtering and feature extraction on the image after light compensation.

[0035] Preferably, the marking quality assessment unit further includes:

[0036] The regional scoring generation module is used to generate quality scoring maps for different areas of the markings, visually displaying the problem areas;

[0037] The overall rating module is used to calculate the overall quality level based on the regional scores, and divides the quality of the markings into five levels: excellent, good, medium, poor, and inferior.

[0038] The assessment report generation module is used to summarize quality scores and anomaly information to generate a detailed assessment report including charts and location markers;

[0039] The historical trend analysis module is used to compare the evaluation results of the same location at different times and analyze the trend of changes in the quality of the markings over time.

[0040] Preferably, the system further includes:

[0041] The remote central control unit is communicatively connected to the detection unit and the road marking quality assessment unit. It is used to issue illumination compensation commands to the detection unit via the Internet of Things, receive preprocessed feature images, and transmit them to the road marking quality assessment unit.

[0042] The remote central control unit includes a mobile network communication module and a network control module. The mobile network communication module connects to the remote data storage and analysis module via a 4G or 5G mobile network. The network control module is used to control the illumination compensation scheme of the detection unit, adjust image colors, and optimize image signal transmission.

[0043] Preferably, the marking wear prediction unit includes:

[0044] The feature mapping subunit is used to construct a model of the mapping relationship between the time period of the road markings and the environmental features based on the road marking positioning parameters, the driving data of the detected vehicles, and the environmental data.

[0045] The traffic flow prediction subunit is used to predict traffic flow by establishing a traffic flow time variation model;

[0046] The road marking life prediction subunit is used to predict the lifespan of road markings based on a mapping model of traffic flow information, road marking time period and environmental characteristics, and recommend maintenance plans based on the lifespan.

[0047] The environmental data includes lighting conditions and weather conditions.

[0048] Preferably, the system further includes:

[0049] The parameter adaptive adjustment module is used to automatically fine-tune the algorithm parameters based on the consistency between the evaluation results and manual verification.

[0050] The model update module is used to periodically collect validation data and optimize the evaluation model.

[0051] The abnormal sample processing module is used to record and evaluate abnormal line samples. After manual confirmation, they are added to the training set to improve the system's recognition accuracy.

[0052] The parameter adaptive adjustment module maintains different parameter configuration schemes for different road surface types, different weather conditions, and different marking types.

[0053] Preferably, the system further includes:

[0054] The cloud management module is used to receive and store line marking images and evaluation results;

[0055] The data analysis platform communicates with the cloud management module to provide historical data analysis and trend prediction.

[0056] A management interface, which communicates with the data analysis platform, is used to support remote configuration and monitoring;

[0057] Third-party system interfaces are used to integrate with external systems such as intelligent transportation systems and road maintenance management systems through standard APIs to achieve data sharing and collaborative decision-making.

[0058] The present invention has the following beneficial effects:

[0059] 1. By introducing curvature flow theory to construct an edge detection technology, the accuracy of marking edge detection in complex road environments is significantly improved, and the problem of traditional edge detection being susceptible to interference from noise, changes in lighting, and road surface cracks is solved.

[0060] 2. By treating the grading lines as a two-dimensional manifold for geometric characterization, the quality status of the grading lines is revealed from a geometric perspective, breaking through the limitations of traditional pixel-value-based analysis and enabling the capture of minute but critical quality changes;

[0061] 3. An evaluation model is constructed using multi-scale differential invariants, which enables comprehensive quantification and accurate rating of road marking quality, and can simultaneously focus on the macroscopic integrity and microscopic details of the road markings;

[0062] 4. A road marking life prediction technology based on geometric characteristics has been established, which can accurately predict the deterioration trend of road markings and provide a scientific basis for road marking maintenance;

[0063] 5. The system has strong environmental adaptability and can maintain stable evaluation performance under different lighting conditions, different road surface types and different weather conditions;

[0064] 6. It adopts a modular design, with clear data flow between components, which facilitates system integration and functional expansion, and has good industrial applicability. Attached Figure Description

[0065] Figure 1This is a diagram showing the overall architecture of the system of the present invention;

[0066] Figure 2 This is a functional block diagram of the road marking quality assessment unit of the present invention;

[0067] Figure 3 This is a flowchart illustrating the operation of the curvature flow edge detection module of the present invention.

[0068] Figure 4 This is a flowchart illustrating the workflow of the manifold characterization module for the geometric characteristics of the marking lines in this invention.

[0069] Figure 5 This is a flowchart of the multi-scale differential invariant evaluation module of the present invention;

[0070] Figure 6 This is a flowchart illustrating the operation of the marking wear prediction unit of the present invention. Detailed Implementation

[0071] Please refer to Figure 1 - Figure 6 The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0072] like Figure 1 As shown, the intelligent road marking quality assessment system based on image analysis provided by the present invention includes a detection unit 1, a road marking quality assessment unit 2, a road marking wear prediction unit 3, a GPS positioning unit 4, a vehicle driving information unit 5, a control unit 6, and a remote central control unit 7.

[0073] In one embodiment of the present invention, the control unit 6 is connected to the GPS positioning unit 4 and the vehicle driving information unit 5. Based on the acquired vehicle location and driving information, it determines whether to initiate road marking detection and sends a detection command to the detection unit 1. The detection unit 1 acquires a road marking image and performs image preprocessing, then transmits the preprocessed image to the road marking quality assessment unit 2. The road marking quality assessment unit 2 assesses the quality of the road marking image and outputs the assessment result, which is then transmitted to the road marking wear prediction unit 3. The road marking wear prediction unit 3 predicts the service life of the road marking based on the assessment result and generates maintenance recommendations.

[0074] Preferably, the remote central control unit 7 is communicatively connected to the detection unit 1 and the road marking quality assessment unit 2. It sends illumination compensation commands to the detection unit 1 via the Internet of Things, receives pre-processed feature images, and transmits them to the road marking quality assessment unit 2. This architecture enables centralized management and sharing of data, improving the system's flexibility and scalability.

[0075] like Figure 1 As shown, the detection unit 1 includes a camera acquisition unit 11, an illumination detection unit 12, a color compensation unit 13, and an image preprocessing unit 14.

[0076] In one embodiment of the present invention, the camera acquisition unit 11 uses an industrial-grade camera with a resolution of 1920×1080, a frame rate of 30fps, and a field of view of 75°, capable of covering the entire width of the lane. The camera is mounted on the front bumper of the vehicle, at a height of approximately 60 centimeters above the ground, and the imaging angle is adjustable.

[0077] The illumination detection unit 12 includes a photosensor with a dynamic range of 120dB, capable of acquiring ambient illumination values ​​in real time. In practical applications, illumination values ​​typically vary within the range of 0-100,000 lux. The system is configured with processing strategies for different illumination conditions. For example, when the illumination value is below 100 lux (nighttime or tunnel environment), the system increases the exposure time and activates the auxiliary light source; when the illumination value is between 100 and 10,000 lux (cloudy days or evening), the system uses standard exposure parameters; and when the illumination value is above 10,000 lux (sunny midday), the system reduces the exposure time and activates strong light suppression.

[0078] The color compensation unit 13 adjusts the color and intensity of the light source based on the real-time illumination value obtained by the illumination detection unit 12 to achieve illumination compensation for the datum line image. In a specific embodiment, the color compensation unit 13 uses the following illumination compensation formula: .

[0079] in: The compensated light intensity is expressed in lux. The initial intensity of the light source is expressed in lux; k is the compensation ratio coefficient, dimensionless, typically ranging from 0.8 to 1.2. The RGB values ​​of the color under standard lighting conditions are a three-dimensional vector. This represents the RGB values ​​of the current ambient light, and is a three-dimensional vector. Each component of the RGB value ranges from 0 to 255, representing the intensity of the red, green, and blue color channels, respectively.

[0080] The RGB values ​​of a light source color are calculated using the following formula: .

[0081] in: The output is the RGB value of the light source color, which is a three-dimensional vector; The initial value of the light source color is a three-dimensional vector; The function is dimensionless and represents the adjustment function; W represents weather information (sunny, cloudy, rainy, snowy, etc.); T represents time information (morning, noon, afternoon, dusk, night, etc.). Adjustment function The base color values ​​are adjusted based on weather and time conditions to make the compensated lighting closer to natural lighting conditions.

[0082] Image preprocessing unit 14 performs noise filtering and feature extraction on the image after illumination compensation. Specifically, Gaussian filtering is first used for noise reduction, with a filter kernel size typically of 5×5 or 7×7, and the σ value is automatically adjusted between 1.0 and 2.5 based on the image noise level. Then, color space conversion is performed, converting the RGB image to HSV or Lab color space to facilitate subsequent mark extraction. Finally, histogram equalization is performed to enhance image contrast. The preprocessed image is then transmitted to mark quality evaluation unit 2 for further analysis.

[0083] like Figure 2 As shown, the road marking quality assessment unit 2 is the core innovative module of this invention, including a curvature flow edge detection module 21, a road marking geometric characteristic manifold characterization module 22, a multi-scale differential invariant assessment module 23, a region score generation module 24, an overall grade assessment module 25, an assessment report generation module 26, and a historical trend analysis module 27.

[0084] like Figure 3 As shown, the curvature flow edge detection module 21 includes an image normalization submodule 211, a curvature calculation submodule 212, an evolution control submodule 213, a multi-scale edge representation submodule 214, and an arc length optimization submodule 215.

[0085] The image normalization submodule 211 converts the input image into a standard resolution and brightness range, establishing a mapping relationship between pixels and physical dimensions. In a preferred embodiment of the present invention, the standard resolution is set to 1280×720 pixels, the brightness range is normalized to 0-1, and the conversion relationship between pixels and physical dimensions is calculated based on the camera mounting height and angle, with a typical value of 1 pixel corresponding to 0.5 to 2 millimeters.

[0086] The curvature calculation submodule 212 treats the grayscale image as a height field, constructs a continuous image surface, and calculates the principal curvature and mean curvature at each point on the surface. In the actual implementation, the gradient field of the image is calculated first: .

[0087] in: For point The gradient vector at point is a two-dimensional vector; For the image at points The gray value at that location is dimensionless and ranges from 0 to 1. and These are grayscale values. right and The partial derivative of the gray value represents the gray value in the range of x and y. and Rate of change in direction.

[0088] Then calculate the Hessian matrix of the image: .

[0089] in: For image The Hessian matrix is ​​a 2×2 matrix; , , and These are grayscale values. The second partial derivative of represents the second rate of change of the gray value in each direction. Since the mixed partial derivatives are equal, i.e. Therefore, the Hessian matrix is ​​a symmetric matrix.

[0090] Calculate the principal curvature of the surface based on the gradient field and the Hessian matrix. and :

[0091] .

[0092] in: and Let |∇I| be the principal curvature of the surface, in pixels ⁻¹; |∇I| be the magnitude of the gradient vector, calculated as |∇I|=√((∂I / ∂x)²+(∂I / ∂y)²); eigenvalues ​​represent the eigenvalues ​​of the matrix, resulting in two real numbers: the maximum principal curvature κ1 and the minimum principal curvature κ2. The principal curvatures describe the degree of bending of the surface in different directions; their signs indicate the direction of bending, with positive values ​​representing convexity and negative values ​​representing concavity.

[0093] The mean curvature H and the Gaussian curvature K are respectively:

[0094] .

[0095] .

[0096] Where: H is the mean curvature, in pixels⁻¹, representing the average degree of surface curvature; K is the Gaussian curvature, in pixels⁻², representing the intrinsic curvature of the surface, which is closely related to the surface's topological properties. Mean curvature and Gaussian curvature are important geometric properties of a surface, related to its shape and deformation characteristics.

[0097] The evolution control submodule 213 controls the surface evolution rate based on the curvature value, ensuring that high-curvature regions (potential edges) evolve slowly while low-curvature regions evolve rapidly and smoothly. Specifically, the following curvature flow equation is used to control surface evolution:

[0098] .

[0099] in: This represents the rate of change of the image over time t; is the edge stopping function, which is dimensionless; div represents the divergence operator, which calculates the divergence of the vector field; The normalized gradient vector represents the direction of the gradient. This equation describes the evolution of an image under the influence of curvature flow. High curvature regions (such as edges) evolve slowly, preserving edge information, while low curvature regions evolve quickly and smoothly, suppressing noise.

[0100] Edge stopping function It is usually defined as:

[0101] .

[0102] in: This is a dimensionless control parameter, typically taken as 1.5-2 times the average value of the image gradient magnitude. The function approaches 0 in regions with large gradients (such as edges), slowing down the evolution process; and approaches 1 in regions with small gradients (such as flat regions), speeding up the evolution process.

[0103] In practical applications, the evolution step size is set to 0.05-0.2, automatically adjusted according to image complexity. The number of iterations is typically 20-50, with the convergence condition being an edge change of less than 0.5 pixels. For straight lines, the curvature threshold is set to 0.01-0.05 pixels - 1; for curved lines, the curvature threshold is set to 0.05-0.15 pixels - 1.

[0104] The multi-scale edge representation submodule 214 generates image representations at 5–8 different scales, with scale factors set to 1.5–2.0. Small scales (1–2 pixels) are used to capture details, while large scales (10–15 pixels) are used to suppress noise. By comparing the positional offsets of edges at different scales, edges that are stable across multiple scales are preserved.

[0105] The arc length optimization submodule 215 defines an arc length functional that includes a curvature term. This functional is minimized to optimize the edge curve while preserving its topological structure. The arc length functional is defined as follows:

[0106] .

[0107] in: Let be the arc length functional, representing the curve Energy; Indicates the edge curve; The curvature of the curve is expressed in pixels - 1. This is a weighting coefficient, dimensionless, and typically ranges from 0.5 to 2.0. Let be a infinitesimal element of the curve, representing a tiny length along the curve; Indicates along the curve The line integral is used to calculate the sum of function values ​​at all points on the curve. This functional consists of two parts: the first part corresponds to the length of the curve, and the second part... The smoothness of the corresponding curve. By minimizing this functional, a short and smooth curve can be obtained, which is suitable for representing the edge of the marker.

[0108] The arc length functional is minimized by gradient descent, with each iteration moving the edge point by no more than 0.5 pixels. At the same time, topology preservation constraints are applied to prevent curves from self-intersecting or breaking.

[0109] The optimized edge curve is output as an edge feature point set to the calibrator geometric property manifold representation module 22.

[0110] like Figure 4 As shown, the geometric characteristic manifold representation module 22 includes a curve parameterization submodule 221, a differential geometric quantity calculation submodule 222, a geometric invariant extraction submodule 223, a geometric feature mapping submodule 224, and a mass parameter space construction submodule 225.

[0111] The curve parameterization submodule 221 parameterizes the edge curve into an arc length parameter representation, establishes a local coordinate system, and divides the marking region into a regular grid. In the specific implementation, a feature point is selected every 5-10 cm along the edge curve, and sampling is intensified at curvature changes. For each feature point, a Frenet frame is established, and a tangent vector is defined. Normal vector and binormal vector :

[0112] .

[0113] .

[0114] .

[0115] in: The vector represents the position of a point on the curve with parameter s. It is a three-dimensional vector with units of centimeters; s is the arc length parameter along the curve, with units of centimeters. The unit tangent vector at that point is dimensionless and represents the direction of the tangent to the curve at that point. The unit normal vector at that point is dimensionless, perpendicular to the tangent vector, and points in the direction of the curve's curvature. Let be the unit binormal vector at that point, dimensionless, and perpendicular to both the tangent vector and the normal vector; Represents position vector The derivative with respect to the arc length parameter s; Represents the tangent vector The derivative with respect to the arc length parameter s; Representing vectors The modulus length; This represents the vector cross product operation. The Frenet frame is an important tool for describing the local geometric properties of curves and can be used to analyze the bending and torsional characteristics of curves.

[0116] The marked area is divided into a regular grid, typically 1×1 cm in size, with the corresponding pixel count on the image calculated based on the camera's mounting height and angle. Special boundary conditions are applied to the edges and endpoints of the marked areas to ensure computational stability.

[0117] The differential geometry computation submodule 222 computes the metric tensor, the first fundamental form, and the second fundamental form on the discrete grid, describing the intrinsic and extrinsic geometric properties of the gauging surface.

[0118] The first basic form is calculated as follows:

[0119] .

[0120] in: As the first basic form, it represents the intrinsic measurement of the surface; , , The coefficient is a dimensionless metric. and These represent the parametric representation of the gradation surface. For parameters and The partial derivatives; This represents the vector dot product operation; and Differentiate for parameters. The first fundamental form describes the intrinsic geometric quantities on the surface, such as distance, angle, and area, independent of how the surface is embedded in three-dimensional space.

[0121] The second basic form is calculated as follows:

[0122] .

[0123] in: This is the second basic form, representing the external geometric properties of the surface; , , The coefficient of the second basic form, in centimeters. ; This is the surface normal vector, which is dimensionless. , and They represent The second-order partial derivative. The second fundamental form describes the degree and direction of curvature of the surface in three-dimensional space, reflecting the external geometric properties of the surface.

[0124] The geometric invariant extraction submodule 223 calculates geometric invariants independent of the observation viewpoint, including Gaussian curvature, mean curvature, and shape exponent. The Gaussian curvature $K$ and the mean curvature $H$ are respectively:

[0125] .

[0126] .

[0127] in: Gaussian curvature, in centimeters. This indicates the degree of inherent curvature of the surface; The mean curvature is expressed in centimeters. This represents the average value of the external curvature of the surface; , , The coefficients are those of the first basic form; , , The coefficients are for the second fundamental form. Gaussian curvature is an intrinsic invariant of a surface, which does not change with how the surface is curved. It only changes when the surface is torn or glued together, and is an important indicator for describing the topological properties of a surface.

[0128] Shape Index Defined as:

[0129] .

[0130] in: The shape index is dimensionless and its value ranges from 1 to 1. and Principal curvature, in centimeters⁻¹, and , The arctangent function. Shape index. Different values ​​of represent different local shapes: Indicates a pit, Indicates valley, Indicates a saddle point. Indicates spine, It represents a convex peak. The shape index provides a visual description of the local open lines on the surface and is unaffected by surface scaling.

[0131] The geometric feature mapping submodule 224 establishes a multi-dimensional feature space, generates geometric feature vectors for each region of the datum line, and identifies regions with abnormal geometric characteristics. In one embodiment of the present invention, the geometric feature vector contains 12 geometric descriptors, including Gaussian curvature, mean curvature, shape index, principal curvature ratio, curvature derivative, etc.

[0132] In the assessment of pavement marking quality, different geometric characteristics and pavement marking quality problems have the following mapping relationship:

[0133] Abnormal Gaussian curvature (exceeding ±20% of the normal range) corresponds to material peeling or bulging;

[0134] Abrupt changes in average curvature (a change of more than 30% between adjacent areas) indicate edge wear;

[0135] Shape index change (deviation from normal range >25%) indicates abnormal surface morphology of the indicator line.

[0136] The quality parameter space construction submodule 225 defines basic quality dimensions such as integrity, flatness, adhesion, and uniformity, mapping geometric characteristics to the quality parameter space. Each quality parameter ranges from 0 to 100, with the following thresholds: Excellent: 90–100 points; Good: 75–89 points; Average: 60–74 points; Poor: 40–59 points; Inferior: 0–39 points.

[0137] The geometric property dataset is output to the multi-scale differential invariant evaluation module 23 for further analysis.

[0138] like Figure 5 As shown, the multi-scale differential invariant evaluation module 23 includes a differential invariant calculation submodule 231, a multi-scale feature fusion submodule 232, a feature manifold construction submodule 233, a quality score generation submodule 234, and an anomaly pattern recognition submodule 235.

[0139] The differential invariant calculation submodule 231 calculates first-order, second-order, and higher-order invariants. First-order invariants include the trace of the metric tensor (local surface area change) and the determinant (surface scaling ratio), sampled at 4–6 points per square centimeter. Second-order invariants include Gaussian curvature, mean curvature, and the ratio of principal curvatures, sampled at 2–3 points per square centimeter. Higher-order invariants include the eigenvalue spectrum and curvature derivative of the shape operator, sampled at 1–2 points per square centimeter in key regions.

[0140] The multi-scale feature fusion submodule 232 constructs 3-5 scale levels, covering features from details (centimeter level) to the overall scale (meter level). Features at each scale are decomposed into basic components, the most discriminative components are selected and retained, and the feature vector is reconstructed by weighting them according to importance. The scale weights can be automatically adjusted in different application scenarios: wear detection: fine-scale weights increase to 60%-70%; overall evaluation: medium-scale weights dominate (50%-60%); trend analysis: large-scale weights increase to 40%-50%.

[0141] The feature manifold construction submodule 233 defines a similarity measure between features, constructs the Riemannian metric of the feature manifold, and establishes geodesic connections between feature points. The geodesic distance between feature points is also considered. Defined as:

[0142] .

[0143] in: For feature points and The geodesic distance between them, in units that depend on the metric of the eigenvector; This means retrieve all connections. and path The minimum value in; This represents the integral over the parameter interval [0,1]. Indicates the connection of feature points and path , For the Riemannian metric, the inner product of vectors in the tangent space is defined; Let be the tangent vector of the path, representing the path in the parameter The tangent vector at the point; Indicates at point At that point, the tangent vector In measurement The geodesic distance measures the shortest path length between two points on a characteristic manifold and is a natural measure of feature similarity.

[0144] The quality score generation submodule 234 defines the measurement function for quality difference, establishes the mapping from geodesic distance to quality score, and generates integrity score, visibility score, and reflectivity performance score. Quality Score Geodetic distance The mapping relationship is as follows:

[0145] .

[0146] in: This is a quality score, dimensionless, with a value range of 0-100; It is a natural exponential function; This is a proportionality coefficient, with the unit being the reciprocal of the geodesic distance. It is adjusted according to the evaluation criteria and is usually taken as 0.1-0.5. This represents the geodesic distance. This mapping function converts the geodesic distance into a quality score; the smaller the distance, the higher the score, and the larger the distance, the lower the score, which aligns with the intuition of quality assessment.

[0147] The anomaly pattern recognition submodule 235 extracts major variation patterns based on principal component analysis on the manifold, and identifies and classifies abnormal regions. By comparing the actual marking features with standard patterns, the system can identify different types of anomalies such as wear, cracking, peeling, and contamination. In the embodiments of the present invention, the criteria for judging anomalies are as follows: Wear: The curvature change of the marking edge exceeds a threshold (usually 30%), and the surface smoothness decreases; Cracking: Local extrema appear in the Gaussian curvature, and the shape index jumps; Peeling: The adhesion parameter is below 60 points, and the surface uniformity is poor; Contamination: The reflectivity score is more than 50% lower than the normal value, but the geometric characteristics are basically normal.

[0148] The quality assessment results are output to the regional score generation module 24, the overall grade assessment module 25, the assessment report generation module 26, and the historical trend analysis module 27, and then passed to the marking wear prediction unit 3 for further analysis.

[0149] The regional scoring generation module 24 generates quality score maps for different areas of the marking line, visually displaying problem areas. Specifically, the marking line is divided into several areas (usually 10×10 cm or 20×20 cm), a quality score is calculated for each area, and a pseudo-color mapping is used to visually display the quality status. In a preferred embodiment of the invention, a red-yellow-green color scheme is used to represent quality from poor to good, facilitating managers to quickly identify problem areas.

[0150] The overall rating module 25 calculates the overall quality level based on the regional scores, classifying the marking quality into five levels: Excellent, Good, Average, Poor, and Inferior. The overall score calculation uses a weighted average method.

[0151] .

[0152] in: The overall score is dimensionless and ranges from 0 to 100. The score for the i-th region is dimensionless and ranges from 0 to 100. The weighting coefficients are dimensionless and satisfy the following conditions: ; Total number of regions; This represents a weighted summation over all areas. The weighting coefficients can be set according to the importance of the areas; for example, the weight of markings near turns and intersections can be appropriately increased.

[0153] The assessment report generation module 26 summarizes the quality scores and anomaly information, generating a detailed assessment report including charts and location markers. The report includes: pavement marking location information (GPS coordinates, road name, marker number, etc.), overall quality level, area rating map, list of anomaly areas and their types, severity, and locations, as well as maintenance recommendations.

[0154] The historical trend analysis module 27 compares the assessment results of the same location at different times to analyze the trend of pavement marking quality changes over time. By fitting a degradation curve, the trend of pavement marking quality changes can be predicted, providing a reference for maintenance decisions. The degradation curve typically uses an exponential model.

[0155] .

[0156] in: Indicates time The quality score at that time is dimensionless and ranges from 0 to 100. The initial quality score is dimensionless and ranges from 0 to 100. It is usually the quality score of the new caliber, close to 100. It is a natural exponential function; The degradation factor is expressed in units of time. It is related to factors such as traffic flow, climate conditions, and road marking materials; The time unit is months or years. This model describes the exponential decay of road marking quality over time, consistent with observed road marking degradation patterns.

[0157] like Figure 6 As shown, the road marking wear prediction unit 3 includes a feature mapping subunit 31, a traffic flow prediction subunit 32, and a road marking life prediction subunit 33.

[0158] Feature mapping subunit 31 constructs a mapping relationship model between road marking time periods and environmental features based on road marking positioning parameters, detected vehicle driving data, and environmental data. Environmental data includes lighting conditions (sunny, cloudy, rainy, snowy, etc.) and weather conditions (temperature, humidity, precipitation, etc.). The mapping relationship model employs a weighted fusion method.

[0159] .

[0160] in: The output of the mapping relationship model is a feature vector; For the first A feature mapping function maps input parameters to feature values; The parameters for marking location include location coordinates, road type, etc. The parameters for the detection period include date, time, etc. These are environmental parameters, including temperature, humidity, and light intensity. The weighting coefficients are dimensionless and satisfy the following conditions: The number of feature mapping functions; This represents a weighted summation over all feature mapping functions. This model fuses multidimensional heterogeneous data into a consistent feature representation, facilitating subsequent analysis.

[0161] Traffic flow prediction subunit 32 predicts traffic flow by establishing a traffic flow temporal variation model. The temporal variation model considers intraday, intraweekly, and seasonal variations:

[0162] .

[0163] in: Indicates time Traffic flow, in vehicles per hour; The baseline traffic flow is expressed in vehicles per hour, representing traffic flow under standard conditions. , and These represent the functions of intraday, intraweek, and seasonal variation, respectively, and are dimensionless. , , and These represent time variables for hours, days, months, and years, respectively. The annual growth rate is dimensionless and typically ranges from 2% to 5%. This model comprehensively considers the multi-scale temporal variations of traffic flow and can accurately predict traffic conditions at different times.

[0164] The road marking life prediction subunit 33 predicts the lifespan of road markings based on a mapping model of traffic flow information, road marking time periods, and environmental characteristics, and recommends maintenance plans based on the lifespan. The road marking lifespan prediction formula is as follows:

[0165] .

[0166] in: Indicates the predicted lifespan of the markings, in months; The baseline lifespan under standard conditions is measured in months and is determined based on the marking materials and construction quality. This is a correction factor, dimensionless, and related to environmental factors such as temperature and pollution index, typically ranging from 0.6 to 1.2. This is a dimensionless environmental attribute parameter for the road markings, which is related to the environmental characteristics of the location where the markings are situated, and is typically between 0.8 and 1.5. This is a traffic flow correction factor, dimensionless, and typically proportional to the -0.6 power of the traffic flow. ,in This represents the average daily traffic flow. The model considers various factors affecting the lifespan of road markings and can predict their lifespan relatively accurately.

[0167] In a preferred embodiment of the present invention, the system provides the following maintenance recommendations based on the predicted service life: service life < 3 months: priority replacement; service life 3-6 months: planned replacement; service life 6-12 months: periodic inspection; service life > 12 months: normal monitoring.

[0168] The remote central control unit 7 includes a mobile network communication module and a network control module.

[0169] The mobile network communication module connects to the remote data storage and analysis module via a 4G or 5G mobile network, and connects to the road marking quality assessment unit 2 and the road marking wear prediction unit 3. In one embodiment of the invention, the communication module adopts 4G / 5G dual-mode communication, supporting a maximum upload speed of 100Mbps to ensure real-time transmission of image data.

[0170] The network control module controls the illumination compensation scheme of detection unit 1, controls the image noise reduction scheme through camera data processing, adjusts image colors, and optimizes image signal transmission. In practical applications, the network control module automatically adjusts the data transmission strategy according to the current network conditions, such as reducing image resolution or using a higher compression rate when the network is congested, to ensure the real-time performance of the system.

[0171] The system of the present invention also includes a parameter adaptive adjustment module, a model update module, an abnormal sample processing module, a cloud management module, a data analysis platform, a management interface, and a third-party system interface.

[0172] The parameter adaptive adjustment module automatically fine-tunes the algorithm parameters based on the consistency between the evaluation results and manual verification. In practical applications, the system periodically collects the manually verified line marking quality evaluation results, compares them with the system evaluation results, and calculates a consistency index. When the consistency falls below a threshold (usually 85%), the system automatically adjusts the relevant algorithm parameters to improve the evaluation accuracy.

[0173] The model update module periodically collects validation data and optimizes the evaluation model. In a preferred embodiment of the invention, the system updates the model quarterly, using an incremental learning method to gradually improve the evaluation model.

[0174] The anomaly sample handling module records the marking samples with evaluation errors. After manual verification, these samples are added to the training set to improve the system's recognition accuracy. In particular, for samples where the system evaluation results differ significantly from the manual evaluation, the system will automatically mark them and submit them for manual review. The verified samples will be used as training data for model optimization.

[0175] The cloud management module is used to receive and store line marking images and evaluation results. In one embodiment of the invention, a distributed storage architecture is adopted to support petabyte-level data storage and provide efficient data retrieval and backup functions.

[0176] The data analytics platform provides historical data analysis and trend prediction capabilities, supporting multi-dimensional data queries and visualization. Through this platform, managers can understand the overall status and trends of road marking quality within the region, providing a basis for macro-level decision-making.

[0177] The management interface supports remote configuration and monitoring, and provides access via web and mobile devices, making it easy for administrators to understand the system's operating status and evaluation results at any time.

[0178] Third-party system interfaces integrate with external systems such as intelligent transportation systems and road maintenance management systems through standard APIs to achieve data sharing and collaborative decision-making.

[0179] The image analysis-based intelligent assessment system for highway marking quality of this invention is applicable to the assessment and maintenance management of marking quality on various types of roads, including highways and urban roads. The system can be installed on dedicated inspection vehicles or regular road patrol vehicles to achieve routine monitoring of marking quality.

[0180] In highway scenarios, the system focuses on the reflectivity and integrity of road markings, increasing the data acquisition frequency and reducing the processing area per instance to ensure accuracy under high-speed driving conditions. In urban road scenarios, the system enhances its anti-interference capabilities, focusing on the visibility and wear condition of road markings to adapt to road marking recognition in complex backgrounds.

[0181] For special road sections, such as tunnel entrances and exits, bridges and overpasses, and construction areas, the system adopts targeted processing strategies to ensure the accuracy and reliability of the assessment.

[0182] Through the system of this invention, road management departments can promptly identify pavement marking quality issues, scientifically formulate maintenance plans, improve road safety, and reduce maintenance costs. Simultaneously, the pavement marking quality information provided by the system also offers crucial support for autonomous driving and advanced driver assistance systems.

[0183] The above description is merely a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. An intelligent assessment system for highway marking quality based on image analysis, characterized in that, include: The detection unit is used to acquire road marking images and perform image preprocessing; A road marking quality assessment unit, communicatively connected to the detection unit, is used to receive preprocessed road marking images sent by the detection unit. The road marking quality assessment unit includes: The curvature flow edge detection module is used to treat the preprocessed marking image as a two-dimensional manifold, construct the curvature flow equation, extract marking edge features through multi-scale representation, and output the edge feature point set; The grading geometric property manifold representation module is communicatively connected to the curvature flow edge detection module. It is used to receive the edge feature point set, construct the parameterized representation of the grading manifold, calculate the first and second basic forms of the grading surface, extract Gaussian curvature and average curvature features, and generate a geometric property dataset. The multi-scale differential invariant evaluation module is communicatively connected to the marking geometric characteristic manifold characterization module. It is used to receive the geometric characteristic dataset, calculate multi-order differential invariants, fuse multi-scale features, construct geodesic distance metrics on the feature manifold, evaluate the integrity, visibility, and reflectivity of the marking, and output quality evaluation results. The road marking wear prediction unit is communicatively connected to the road marking quality assessment unit, and is used to receive the quality assessment results, predict the service life of the road markings, and generate maintenance suggestions. The multi-scale differential invariant evaluation module includes: The differential invariant calculation submodule is used to calculate first-order invariants, second-order invariants, and higher-order invariants. The multi-scale feature fusion submodule is used to construct multiple scale levels covering from details to the whole, decompose the features at each scale into basic components, and reconstruct the feature vector by weighting them according to their importance. The Feature Manifold Construction Submodule is used to define similarity measures between features, construct the Riemannian metric of the feature manifold, and establish geodesic connections between feature points. The quality score generation submodule is used to define the measurement function of quality difference, establish the mapping from geodesic distance to quality score, and generate integrity score, visibility score and reflectivity score; The anomaly pattern recognition submodule is used to extract major variation patterns based on principal component analysis on the manifold, and to identify and classify anomalous regions.

2. The system according to claim 1, characterized in that, The curvature flow edge detection module includes: The image normalization submodule is used to convert the input image into a standard resolution and brightness range, and establish a mapping relationship from pixels to physical size; The curvature calculation submodule is used to construct the image surface and calculate the principal curvature and average curvature at each point on the surface; The evolution control submodule is used to control the evolution speed of the surface according to the curvature value, so that the high curvature region evolves slowly and the low curvature region evolves quickly and smoothly. The multi-scale edge representation submodule is used to generate image representations at different scales and preserve edges that are stable at multiple scales by comparing the positional offsets of edges at different scales. The arc length optimization submodule is used to define an arc length functional that includes a curvature term. This functional is used to optimize edge curves while maintaining their topology.

3. The system according to claim 1, characterized in that, The caliper geometric characteristic manifold characterization module includes: The curve parameterization submodule is used to parameterize the edge curve into an arc length parameter representation, establish a local coordinate system, and divide the marking area into a regular grid. The differential geometry computation submodule is used to compute the metric tensor, the first fundamental form and the second fundamental form on a discrete grid, describing the intrinsic and extrinsic geometric properties of the gauging surface; The geometric invariant extraction submodule is used to calculate geometric invariants that are independent of the observation viewpoint, including Gaussian curvature, mean curvature, and shape exponent; The geometric feature mapping submodule is used to establish a multi-dimensional feature space, generate geometric feature vectors for each region of the datum, and identify regions with abnormal geometric characteristics. The quality parameter space construction submodule is used to define the basic quality dimensions of integrity, flatness, adhesion, and uniformity, mapping geometric properties to the quality parameter space.

4. The system according to claim 1, characterized in that, Also includes: The GPS positioning unit is used to obtain the vehicle's real-time location and direction of travel. The vehicle driving information unit is used to obtain the vehicle's real-time speed, acceleration, and steering angle; The control unit is communicatively connected to the GPS positioning unit and the vehicle driving information unit, and is used to determine whether to activate the lane marking detection based on the vehicle position and driving information, and to send a detection command to the detection unit. The detection unit includes a camera acquisition unit, a light detection unit, a color compensation unit, and an image preprocessing unit. The camera acquisition unit acquires the marking image according to the detection command. The light detection unit acquires the real-time light value. The color compensation unit adjusts the color and intensity of the light source according to the real-time light value. The image preprocessing unit performs noise filtering and feature extraction on the image after light compensation.

5. The system according to claim 1, characterized in that, The road marking quality assessment unit also includes: The regional scoring generation module is used to generate quality scoring maps for different areas of the markings, visually displaying the problem areas; The overall rating module is used to calculate the overall quality level based on the regional scores, and divides the quality of the markings into five levels: excellent, good, medium, poor, and inferior. The assessment report generation module is used to summarize quality scores and anomaly information to generate a detailed assessment report including charts and location markers; The historical trend analysis module is used to compare the evaluation results of the same location at different times and analyze the trend of changes in the quality of the markings over time.

6. The system according to claim 1, characterized in that, The system also includes: The remote central control unit is communicatively connected to the detection unit and the road marking quality assessment unit. It is used to issue illumination compensation commands to the detection unit via the Internet of Things, receive preprocessed feature images, and transmit them to the road marking quality assessment unit. The remote central control unit includes a mobile network communication module and a network control module. The mobile network communication module connects to the remote data storage and analysis module via a 4G or 5G mobile network. The network control module is used to control the illumination compensation scheme of the detection unit, adjust image colors, and optimize image signal transmission.

7. The system according to claim 1, characterized in that, The marking wear prediction unit includes: The feature mapping subunit is used to construct a model of the mapping relationship between the time period of the road markings and the environmental features based on the road marking positioning parameters, the driving data of the detected vehicles, and the environmental data. The traffic flow prediction subunit is used to predict traffic flow by establishing a traffic flow time variation model; The road marking life prediction subunit is used to predict the lifespan of road markings based on a mapping model of traffic flow information, road marking time period and environmental characteristics, and recommend maintenance plans based on the lifespan. The environmental data includes lighting conditions and weather conditions.

8. The system according to claim 1, characterized in that, The system also includes: The parameter adaptive adjustment module is used to automatically fine-tune the algorithm parameters based on the consistency between the evaluation results and manual verification. The model update module is used to periodically collect validation data and optimize the evaluation model. The abnormal sample processing module is used to record and evaluate abnormal line samples. After manual confirmation, they are added to the training set to improve the system's recognition accuracy. The parameter adaptive adjustment module maintains different parameter configuration schemes for different road surface types, different weather conditions, and different marking types.

9. The system according to claim 1, characterized in that, The system also includes: The cloud management module is used to receive and store line marking images and evaluation results; The data analysis platform communicates with the cloud management module to provide historical data analysis and trend prediction. A management interface, which communicates with the data analysis platform, is used to support remote configuration and monitoring; Third-party system interfaces are used to integrate with external systems such as intelligent transportation systems and road maintenance management systems through standard APIs to achieve data sharing and collaborative decision-making.

Images