PFWD and visual feature-based subgrade compaction quality evaluation method and system
By combining the PFWD dynamic modulus with visual features to establish a mapping relationship, the problems of insufficient coverage of point detection and lack of mechanical calibration of visual detection in the subgrade compaction quality detection are solved, realizing continuous spatial distribution evaluation and efficient detection of subgrade compaction quality.
Patent Information
- Application Number
- CN202611123092.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-28
- Publication Date
- 2026-08-25
AI Technical Summary
Existing technologies are insufficient to evaluate the continuous spatial distribution of roadbed compaction quality. Point-based detection coverage is inadequate, visual inspection lacks mechanical calibration, and re-measurement of low-compaction areas is inaccurate.
By combining the dynamic modulus of PFWD with the visual features obtained by UAVs and vehicle-mounted cameras, a mapping relationship is established. Through the equivalent PFWD dynamic modulus prediction model and compaction degree prediction model, the gridded and rapid evaluation of the compaction quality of the subgrade is realized.
It enables continuous spatial distribution evaluation of roadbed compaction quality, improves detection efficiency and the reliability of prediction results, and ensures accurate retesting and rework positioning of low-compaction areas.
Smart Images
Figure CN122636033A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of road construction quality inspection technology, and in particular to a method and system for assessing roadbed compaction quality based on PFWD and visual features. Background Technology
[0002] Subgrade compaction quality is a crucial factor affecting the long-term service performance of roads. Insufficient or uneven compaction can lead to reduced subgrade bearing capacity, increased differential settlement, pavement cracking, rutting, and frost heave. Existing PFWD (Power Plant Dynamic Modulus and Surface Settlement) testing can quickly obtain dynamic modulus and surface settlement, but it still mainly reflects the local quality status near the testing point and cannot cover the entire construction surface.
[0003] With the development of drones, vehicle-mounted cameras, intelligent sensing equipment for road rollers, and image recognition technology, large-scale images of the roadbed surface can be quickly acquired at construction sites. However, relying solely on surface images is insufficient to directly reflect the internal compaction quality and dynamic bearing capacity of the roadbed, as they are easily affected by factors such as sunlight, moisture content, filler type, and surface disturbance. On the other hand, while PFWD test results have relatively clear engineering physical significance, the limited number of test points makes it difficult to form a continuous spatial distribution of compaction quality. Summary of the Invention
[0004] To address the shortcomings of existing technologies, the purpose of this invention is to provide a method and system for evaluating the compaction quality of roadbeds based on PFWD (Power Factor Dynamic Modulus), compaction degree, moisture content, dry density, and construction compaction parameters. This method uses PFWD dynamic modulus, compaction degree, moisture content, dry density, and construction compaction parameters as calibration information to establish a mapping relationship between visual features and compaction quality. This enables gridded, rapid, and traceable evaluation of roadbed compaction quality, solving problems such as insufficient coverage of point detection, lack of mechanical calibration for visual detection, difficulty in continuously evaluating the spatial distribution of construction quality, and inaccurate retesting and rework location in low-compaction areas in existing roadbed compaction quality testing.
[0005] To achieve the above objectives, the present invention is implemented through the following technical solution: The first aspect of this invention provides a method for assessing the compaction quality of roadbed based on PFWD and visual features, comprising the following steps: Based on construction requirements, the roadbed construction surface to be inspected in the construction area is divided into several construction grids; Simultaneously acquire PFWD detection data and on-site quality data within the construction grid, and obtain corresponding subgrade compaction surface images, extracting the subgrade surface visual features from the subgrade compaction surface images; The equivalent PFWD dynamic modulus prediction model is used to process the visual features of the subgrade surface, PFWD detection data, construction parameter vectors and field quality data to obtain the equivalent PFWD dynamic modulus prediction value. The compaction degree prediction model is used to predict the quality of the subgrade surface visual features, field quality data, construction parameter vectors and the equivalent PFWD dynamic modulus prediction value to obtain the compaction degree prediction value and the corresponding prediction confidence. The subgrade compaction quality is assessed using the equivalent PFWD dynamic modulus prediction value, compaction degree prediction value, and corresponding prediction confidence.
[0006] Furthermore, PFWD test data includes dynamic modulus and surface settlement, while field quality data includes field compaction, dry density, and moisture content.
[0007] Furthermore, the specific steps for extracting the visual features of the roadbed surface from the compacted surface image are as follows: Preprocessing and construction mesh matching operations are performed on the compacted surface image of the roadbed. Visual features of the compacted roadbed surface are extracted from the preprocessed image to form a visual feature vector. The visual feature vector includes surface texture uniformity features, wheel track depth features, particle distribution features, wet spot features, local loose features, and surface subsidence or uneven compaction features.
[0008] Furthermore, the compaction parameters corresponding to each construction grid are collected to form a construction parameter vector. The compaction parameters include the number of compaction passes, compaction speed, vibration frequency, amplitude, compaction energy, and filler type. Among them, the compaction energy is characterized based on the roller mass, vibration parameters, number of compaction passes, and effective area.
[0009] Furthermore, the specific steps for processing the visual features of the roadbed surface, PFWD detection data, and field quality data using the equivalent PFWD dynamic modulus prediction model are as follows: An equivalent PFWD dynamic modulus prediction model was constructed and trained. The equivalent PFWD dynamic modulus prediction model includes a visual initial prediction sub-model and a construction parameter correction sub-model. The equivalent PFWD dynamic modulus prediction model uses the PFWD dynamic modulus as a supervision label, uses the visual initial prediction sub-model to predict the visual features of the subgrade surface, and the construction parameter correction sub-model corrects the results of the visual initial prediction sub-model according to the construction parameters, moisture content and filler type.
[0010] Furthermore, the specific steps for using the compaction degree prediction model to predict the quality of the roadbed surface based on visual characteristics, on-site quality data, construction parameter vectors, and the equivalent PFWD dynamic modulus are as follows: Construct and train a compaction degree prediction model; The trained compaction prediction model takes visual feature vector, construction parameter vector, moisture content, filler type and equivalent PFWD dynamic modulus prediction value as input, and uses the on-site compaction degree as the supervision label to predict quality.
[0011] Furthermore, the specific steps for assessing the compaction quality of the subgrade using the equivalent PFWD dynamic modulus prediction value, compaction degree prediction value, and corresponding prediction confidence level are as follows: Based on the predicted value of the equivalent PFWD dynamic modulus, the predicted value of compaction degree and the corresponding prediction confidence, the compaction quality level of the construction grid is determined and a compaction quality distribution map is generated. The recommended score is calculated based on the retesting of the compaction quality grade of the construction grid.
[0012] A second aspect of the present invention provides a roadbed compaction quality assessment system based on PFWD and visual features, comprising: The construction grid division module is used to divide the roadbed construction surface to be inspected in the construction area into several construction grids according to construction needs. The data acquisition module is used to simultaneously acquire PFWD detection data and field quality data within the construction grid, and to acquire corresponding subgrade compaction surface images and extract the visual features of the subgrade surface from the subgrade compaction surface images. The model prediction module is used to process the visual features of the subgrade surface, PFWD detection data, construction parameter vectors and field quality data using the equivalent PFWD dynamic modulus prediction model to obtain the equivalent PFWD dynamic modulus prediction value. The compaction degree prediction model is used to perform quality prediction on the visual features of the subgrade surface, field quality data, construction parameter vectors and the equivalent PFWD dynamic modulus prediction value to obtain the compaction degree prediction value and the corresponding prediction confidence level. The quality evaluation module is used to assess the compaction quality of the subgrade using the equivalent PFWD dynamic modulus prediction value, compaction degree prediction value, and corresponding prediction confidence.
[0013] A third aspect of the present invention provides a computer-readable storage medium storing a computer program adapted to be loaded by a processor and to execute steps in the method for assessing subgrade compaction quality based on PFWD and visual features as described in the first aspect of the present invention.
[0014] A fourth aspect of the present invention provides a computer device comprising: A processor, adapted to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by the processor, implements the subgrade compaction quality assessment method based on PFWD and visual features as described in the first aspect of the present invention.
[0015] The above one or more technical solutions have the following beneficial effects: This invention discloses a method and system for assessing roadbed compaction quality based on PFWD (Powered Falling Weight Deflectometer) and visual features. It combines point-based detection results from a portable falling weight deflectometer (PFWD) with area-based visual data acquired by drones, vehicle-mounted cameras, or cameras mounted on rollers. This expands the PFWD dynamic modulus and surface settlement of limited detection points into continuous compaction quality evaluation results for the construction area, overcoming the problems of insufficient spatial coverage and low detection efficiency in traditional compaction quality testing. This invention calibrates the visual features of the roadbed surface using PFWD dynamic modulus, on-site compaction degree, dry density, and moisture content, and introduces construction parameters such as the number of compaction passes, compaction speed, vibration frequency, amplitude, and compaction energy for correction. This ensures that the compaction quality prediction results have both visual characterization and mechanical testing evidence, improving the reliability of the equivalent PFWD dynamic modulus and compaction degree predictions.
[0016] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description
[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0018] Figure 1 This is a flowchart of the subgrade compaction quality assessment method based on PFWD and visual features in Embodiment 1 of the present invention; Figure 2 This is a schematic diagram of the grid division of the construction area and the layout of PFWD calibration points in Embodiment 1 of the present invention; Figure 3 This is a schematic diagram illustrating the simultaneous collection of PFWD detection, on-site quality inspection, and construction visual data in Embodiment 1 of the present invention. Detailed Implementation
[0019] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.
[0020] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof. The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.
[0021] Example 1: Embodiment 1 of the present invention provides a method for assessing the compaction quality of roadbed based on PFWD and visual features, such as... Figure 1 As shown, this method uses the point detection results of PFWD as the mechanical calibration benchmark and the visual features of the subgrade surface acquired by UAVs, vehicle-mounted cameras, inspection vehicle cameras, or cameras mounted on road rollers as the area perception information. Combined with compaction degree, dry density, moisture content, and rolling construction parameters, it establishes a mapping relationship between visual features, construction parameters, and PFWD dynamic modulus and compaction degree, so as to realize gridded prediction of subgrade compaction quality, identification of low compaction areas, recommendation of PFWD re-measurement points, and traceability of construction quality.
[0022] Specifically, the following steps are included: S1: Divide the roadbed construction surface to be inspected in the construction area into several construction grids according to the construction requirements.
[0023] In one specific implementation, such as Figure 2 As shown in this embodiment, during the roadbed filling construction process, the roadbed construction surface to be inspected is divided into several construction grids according to construction requirements such as construction station number, transverse width, number of filling layers, fill material type, loose paving thickness, and construction quality control accuracy. The construction grid corresponding to the i-th longitudinal station interval, the j-th transverse region, and the k-th filling layer is denoted as: .
[0024] In the formula, Indicates the construction grid; Indicates the first One longitudinal station interval; Indicates the first One horizontal region; Indicates the first Layered filling layer.
[0025] Based on the filler type, moisture content, loose thickness, number of compaction passes, construction location, and preliminary visual anomalies, determine whether PFWD (Potentially Positive Surface Drying) detection points need to be set up in the current construction grid. For the [number missing]... Each construction grid is used to determine the priority score. .
[0026] in: .
[0027] In the formula, Indicates the first Priority scoring for each construction grid; Indicates representative indicators of packing type; This indicates that the moisture content deviates from the target value; This indicates that the loose-lay thickness deviates from the specified value; This indicates that the number of compaction passes deviates from the target. Indicates preliminary visual abnormality indicators; Indicators representing the sensitivity of construction location; This represents the weighting coefficient of the corresponding indicator.
[0028] The PFWD calibration mesh set is represented as follows: .
[0029] In the formula, Indicates the PFWD calibration mesh set; This represents a representative set of coverage grids used to cover different filler types, different moisture content ranges, different loose thickness ranges, and different compaction pass ranges. This indicates the set of meshes with priority given to anomalies. It is used to select construction meshes with obvious wet spots, wheel tracks, local looseness, surface depression, or uneven texture. This represents a set of uniformly spaced check grids, used to select check grids according to preset station spacing, lateral spacing, or sampling ratio.
[0030] The anomaly priority grid set is represented as: .
[0031] In the formula, This indicates the preset calibration priority threshold.
[0032] When the construction grid belongs to When a construction grid is selected as the calibration grid, PFWD detection points are set at its center or representative locations. When there are obvious wet spots, wheel tracks, local loosening, surface subsidence, or weak edges within the grid, PFWD detection points are preferentially placed at the center of the abnormal area or at the boundary between the abnormal and normal areas. Construction grids without PFWD detection points are designated as grids to be predicted.
[0033] S2: Simultaneously acquire PFWD detection data and field quality data within the construction grid, and obtain corresponding subgrade compaction surface images, extracting the subgrade surface visual features from the subgrade compaction surface images.
[0034] S2.1: Simultaneously acquire PFWD detection data and on-site quality data within the construction grid.
[0035] In one specific implementation, such as Figure 3 As shown, a portable falling weight deflectometer was used for testing at the PFWD testing point. During the PFWD testing process, data on the falling weight impact load, load application time, bearing plate dimensions, surface settlement, and corresponding time response were collected. The PFWD dynamic modulus was calculated based on the load-settlement response. The PFWD dynamic modulus is a dynamic deformation modulus used to characterize the dynamic bearing capacity of the roadbed near the testing point under transient impact loads. For the (G_{ijk})th construction grid, its PFWD detection data is expressed as: .
[0036] In the formula, This represents the PFWD detection data vector; Indicates the dynamic modulus of PFWD; This indicates the amount of surface settlement under PFWD load.
[0037] Simultaneously collect on-site compaction degree, dry density, and moisture content at the same testing point to form an on-site quality data vector: .
[0038] In the formula, Represents a vector of on-site quality data; Indicates the degree of compaction; Indicates dry density; Indicates moisture content.
[0039] It should be noted that the PFWD dynamic modulus can be directly output by the PFWD device's built-in algorithm, or it can be calculated based on the collected impact load, bearing plate size, and surface settlement. Compaction degree, dry density, and moisture content can be obtained through the sand cone method, ring cutter method, nuclear density meter, rapid moisture content meter, or field sensors. All test data are linked to the construction grid number, station number, number of filling layers, fill material type, test time, and test equipment number.
[0040] S2.2: Obtain the corresponding image of the compacted surface of the roadbed.
[0041] In one specific implementation, such as Figure 3 As shown, this embodiment uses drones, vehicle-mounted cameras, inspection vehicle cameras, or cameras mounted on road rollers to collect images of the compacted roadbed surface. When using drones, the construction area is photographed in a grid pattern according to a preset flight path, and the flight altitude, heading angle, pitch angle, roll angle, GPS coordinates, or RTK coordinates are recorded. When using vehicle-mounted cameras, inspection vehicle cameras, or cameras mounted on road rollers, the vehicle position, speed, shooting angle, image timestamp, and corresponding construction grid number are recorded. Specifically, for the first... The original image collected from the construction grid is denoted as _____. .
[0042] S2.3: Extract visual features of the subgrade surface from the image of the compacted subgrade surface.
[0043] S2.3.1: Perform preprocessing and construction mesh matching operations on the compacted surface image of the roadbed.
[0044] In one specific implementation, the original roadbed surface image ( Image preprocessing is performed, including at least one of orthorectification, scale calibration, image registration, construction area segmentation, mesh matching, and low-quality image removal.
[0045] Obtain the preprocessed image The image preprocessing process is represented as follows: .
[0046] In the formula, This represents the image preprocessing function.
[0047] Low-quality image removal includes identifying image blur, strong light, shadows, rain / fog, occlusion, abnormal shooting angles, and scale distortion. Low-quality images are then assigned an image credibility index. Its value range is: .
[0048] when Less than the preset image quality threshold At that time, the image is marked as a low-confidence image or removed from the model training samples.
[0049] S2.3.2: Extract visual features of the compacted roadbed surface from the preprocessed image to form a visual feature vector.
[0050] In one specific implementation, the visual feature vector of this embodiment includes surface texture uniformity features, wheel track depth features, particle distribution features, wet spot features, local loose features, and surface depression or uneven compaction features.
[0051] Specifically, from the preprocessed image ( Visual features of the compacted roadbed surface are extracted from the data to form a visual feature vector. .
[0052] In the formula, Indicates the first Visual feature vectors of each construction grid; Indicates the uniformity of surface texture; Indicates wheel track depth characteristics; Indicates particle distribution characteristics; Indicates the characteristics of wet spots; Indicates a localized loose structure; This indicates surface subsidence or uneven compaction.
[0053] Among them, surface texture uniformity features It can be obtained from gray-level co-occurrence matrix, local binary pattern, gradient statistical features, or deep learning features; wheel trace depth features It can be obtained from scaled edge lines, shadow variations, 3D reconstruction, or depth estimation; particle distribution characteristics It can be obtained from particle boundaries, texture roughness, and local grayscale variations; wet spot characteristics It can be obtained from color space conversion, brightness difference, and semantic segmentation; local loose features It can be obtained from texture dispersion, particle aggregation degree, and surface fragmentation characteristics; surface subsidence or uneven compaction characteristics. It can be obtained from local elevation changes, shadow changes, or image region deformation features.
[0054] S3: The equivalent PFWD dynamic modulus prediction model is used to process the visual features of the subgrade surface, PFWD detection data, construction parameter vectors and field quality data to obtain the equivalent PFWD dynamic modulus prediction value. The compaction degree prediction model is used to predict the quality of the subgrade surface visual features, field quality data, construction parameter vectors and the equivalent PFWD dynamic modulus prediction value to obtain the compaction degree prediction value and the corresponding prediction confidence.
[0055] S3.1: Collect construction parameters and construct a construction quality calibration sample library.
[0056] In one specific implementation, compaction parameters corresponding to each construction grid are collected to form a construction parameter vector. These compaction parameters include the number of compaction passes, compaction speed, vibration frequency, amplitude, and compaction energy, expressed as: Specifically, the compaction parameters corresponding to each construction grid are collected to form a construction parameter vector: .
[0057] In the formula, Indicates the first Construction parameter vectors for each construction grid; Indicates the number of compaction passes; Indicates the compaction speed; Indicates the vibration frequency; Indicates amplitude; This indicates the compaction energy.
[0058] The type of packing is recorded separately as Moisture content is recorded separately as .in, Used to characterize the differences in material properties corresponding to different types of fillers. Used to characterize the water content of the current construction grid.
[0059] Among them, compaction energy It can be characterized by the roller's mass, vibration parameters, number of compaction passes, and effective area, and is expressed as follows:
[0060] In the formula, Indicates the mass of the road roller; Represents gravitational acceleration; Indicates the number of compaction passes; This represents the vibration energy correction factor determined by the vibration frequency and amplitude. This indicates the area of the construction grid.
[0061] When compaction energy cannot be directly obtained at the construction site, the number of compaction passes, compaction speed, vibration frequency, and amplitude can be used as alternative characterization parameters for compaction energy.
[0062] Next, the visual feature vectors within the calibration grid will be determined. Construction parameter vector Moisture content , packing type PFWD dynamic modulus PFWD surface settlement Compaction degree and dry density Spatial and temporal matching were performed to establish a sample library for construction quality calibration.
[0063] In the formula, This refers to the sample library for construction quality calibration.
[0064] Spatial matching refers to unifying the coordinates of PFWD detection points, image acquisition areas, and construction grids; temporal matching refers to matching PFWD detection time, visual acquisition time, and compaction construction time so that data under the same construction condition enters the same calibration sample.
[0065] S3.2: The equivalent PFWD dynamic modulus prediction model is used to process the visual features of the roadbed surface, PFWD detection data, construction parameter vectors and field quality data to obtain the equivalent PFWD dynamic modulus prediction value.
[0066] S3.2.1: Construct and train an equivalent PFWD dynamic modulus prediction model. The equivalent PFWD dynamic modulus prediction model includes a visual initial prediction sub-model and a construction parameter correction sub-model.
[0067] In one specific implementation, visual feature vectors are used. Construction parameter vector Moisture content and packing type As input, the PFWD dynamic modulus As a supervisory label, an equivalent PFWD dynamic modulus prediction model is established:
[0068] In the formula, Indicates the first Predicted equivalent PFWD dynamic modulus values for each construction grid; This represents the equivalent PFWD dynamic modulus prediction model.
[0069] This embodiment uses different types of packing. With construction parameter vector Separate inputs allow the model to characterize "construction process differences" and "material type differences" separately. The construction parameter vector reflects the impact of the compaction process on compaction quality, while the filler type reflects the differences in mechanical response of different fillers under the same compaction conditions and water content, thereby improving the applicability and stability of the equivalent PFWD dynamic modulus prediction results.
[0070] The equivalent PFWD dynamic modulus prediction model in this embodiment can be established using random forest, support vector machine, gradient boosting tree, convolutional neural network, fully connected neural network or multi-model fusion method, but is not limited to the above model forms.
[0071] The model training process can be represented as:
[0072] In the formula, Representation Model Parameters; This represents the calibration grid set for deploying PFWD detection points; This represents the regularization coefficient.
[0073] S3.2.2: The equivalent PFWD dynamic modulus prediction model uses the PFWD dynamic modulus as a supervision label, uses the visual initial prediction sub-model to predict the visual features of the subgrade surface, and the construction parameter correction sub-model corrects the results of the visual initial prediction sub-model according to the construction parameters, moisture content and filler type.
[0074] In one specific implementation, the equivalent PFWD dynamic modulus prediction model may include a visual initial prediction sub-model and a construction parameter correction sub-model, as follows: .
[0075] In the formula, This represents the initial equivalent PFWD dynamic modulus prediction value obtained from visual features; This indicates a correction term determined by construction parameters, moisture content, and filler type.
[0076] S3.3: Using the compaction degree prediction model, the quality of the roadbed surface is predicted based on visual features, field quality data, construction parameter vectors and equivalent PFWD dynamic modulus prediction values, to obtain the compaction degree prediction value and the corresponding prediction confidence.
[0077] S3.3.1: Construct and train a compaction degree prediction model.
[0078] In one specific implementation, the compaction degree prediction model uses a multi-input feature fusion fully connected neural network as its basic structure. The fully connected neural network includes an input layer, a feature fusion layer, several hidden layers, and an output layer.
[0079] The model input includes visual feature vectors Construction parameter vector Moisture content , packing type and equivalent PFWD dynamic modulus prediction value First, the above inputs are concatenated to form the compaction degree prediction input vector: .
[0080] In the formula, Indicates the first The input vector for predicting the compaction degree of each construction grid.
[0081] input vector The input is a fully connected neural network, which, after nonlinear mapping in the hidden layers, outputs a predicted compaction degree. , .
[0082] In the formula, This represents the output feature of the l-th hidden layer; and These represent the weight matrix and bias term of the l-th hidden layer, respectively. Represents a non-linear activation function; This indicates the output of the last hidden layer; and These represent the output layer weights and biases, respectively. Let G{ijk} represent the predicted compaction degree of the G{ijk}th construction grid.
[0083] The model training process uses on-site measured compaction. As a supervisory label, it is achieved by minimizing the error between the predicted compaction degree and the measured compaction degree. The training objective function is expressed as: .
[0084] In the formula, This represents the network parameters of the compaction degree prediction model. This represents a compaction degree prediction model; This represents the regularization coefficient.
[0085] By incorporating the equivalent PFWD dynamic modulus prediction value into the compaction degree prediction model, the compaction degree prediction results are simultaneously constrained by the surface visual state, the construction rolling process, the water content state, and the dynamic bearing response, thereby improving the engineering rationality of the prediction results.
[0086] It should be noted that the inputs to the compaction degree prediction model include the predicted value of the equivalent PFWD dynamic modulus. In addition, it also includes visual feature vectors. Construction parameter vector Moisture content and packing type .in, Used to characterize the equivalent dynamic bearing capacity of the construction grid under transient loads, serving as a mechanical constraint feature for compaction degree prediction; visual feature vector. Used to characterize the surface texture, wheel tracks, wet spots, looseness, and localized subsidence of the roadbed; construction parameter vector. Used to characterize construction process information such as number of compaction passes, compaction speed, vibration frequency, amplitude, and compaction energy; moisture content. and packing type Used to characterize differences in material state and material category.
[0087] Through the above input design, the compaction degree prediction model does not only predict compaction degree based on the equivalent PFWD dynamic modulus, but also comprehensively predicts compaction degree by combining surface visual condition, construction compaction process, water content, and filler type, based on mechanical bearing response constraints. This avoids misjudging construction grids with similar dynamic moduli under different filler types, different water contents, or different construction processes as having the same compaction state, thereby improving the engineering rationality and applicability of the compaction degree prediction results.
[0088] In this embodiment, the equivalent PFWD dynamic modulus prediction model and the compaction degree prediction model adopt a progressive prediction structure. First, the equivalent PFWD dynamic modulus prediction model uses the measured PFWD dynamic modulus as a monitoring label to establish a mapping relationship between visual features, construction parameters, moisture content, and filler type and dynamic bearing response, thereby extending the mechanical testing information obtained from a limited number of PFWD testing points to the grid to be predicted without PFWD testing points. Second, the compaction degree prediction model uses the field-measured compaction degree as a monitoring label. Based on the introduction of the equivalent PFWD dynamic modulus prediction value, it further integrates visual features, construction parameters, moisture content, and filler type to predict the compaction degree of each construction grid.
[0089] In addition, the PFWD dynamic modulus is used to provide mechanical calibration in terms of dynamic bearing response, so that the model prediction results have a mechanical testing basis; the field measured compaction degree is used to provide calibration in terms of construction quality acceptance indicators, so that the model prediction results can correspond to the roadbed compaction quality evaluation requirements. By first predicting the equivalent PFWD dynamic modulus and then predicting the compaction degree based on mechanical constraints, this embodiment realizes the transformation from "point-like PFWD mechanical testing" to "continuous compaction quality evaluation of the construction surface".
[0090] In summary, the equivalent PFWD dynamic modulus prediction model addresses the lack of mechanical interpretation in visual features, while the compaction degree prediction model addresses the difficulty of a single mechanical response fully reflecting compaction quality. Combining the two allows compaction quality prediction results to be simultaneously constrained by dynamic bearing response, surface visual condition, construction process, moisture content, and fill material type, thereby improving the accuracy, stability, and interpretability of the gridded rapid assessment of subgrade compaction quality.
[0091] S3.3.2: The trained compaction prediction model takes the visual feature vector, construction parameter vector, moisture content, filler type and equivalent PFWD dynamic modulus prediction value as input, and uses the on-site compaction degree as the supervision label to predict the quality.
[0092] In one specific implementation, the compaction degree prediction model of this embodiment uses visual feature vectors. Construction parameter vector Moisture content , packing type and equivalent PFWD dynamic modulus prediction value As input, on-site compaction degree As a supervisory label: .
[0093] In the formula, Indicates the first Predicted compaction degree of each construction grid; This represents a compaction degree prediction model.
[0094] To avoid misjudgments caused by low-quality images, abnormal construction parameters, or insufficient sample distribution, the prediction confidence level is calculated for each construction grid. The prediction confidence of a construction grid can be expressed as: .
[0095] In the formula, Indicates the reliability of the prediction; Indicates the credibility of image quality; This indicates the reliability of the completeness of the construction parameters; Indicates the reliability of the model's applicability; in: .
[0096] In the formula, , , These are the weighting coefficients.
[0097] Among them, image quality credibility Determined based on image sharpness, illumination uniformity, occlusion ratio, and geometric registration accuracy, and expressed as: .
[0098] in: .
[0099] In the formula, Indicates image sharpness score; Indicates the score for uniformity of illumination; Indicates the degree of occlusion; Indicates the image geometric registration score; , , , These are the weighting coefficients. All the above scores are normalized to between 0 and 1, with higher values indicating higher image quality.
[0100] Construction parameter integrity and reliability The parameters, such as the number of compaction passes, compaction speed, vibration frequency, amplitude, and compaction energy, corresponding to the current construction grid, are determined based on their completeness and within a reasonable range, and are expressed as follows: .
[0101] In the formula, n represents the number of construction parameters; This represents the weight of the r-th type of construction parameter; Indicates the first The validity flag of the r-th type of construction parameter in each construction grid. When this construction parameter exists and is within a preset reasonable range... When this construction parameter is missing or obviously abnormal, .
[0102] Model applicability and credibility The type of filler, moisture content, and visual characteristics used to characterize the current construction grid are determined based on the degree of similarity between the input characteristics of the current construction grid and existing samples in the calibration sample library.
[0103] First, construct the model input feature vector of the current construction mesh. : .
[0104] Will After normalization, the distance to the features of samples in the calibration sample library is calculated: .
[0105] In the formula, This represents the feature distance between the current construction grid and the closest sample in the calibration sample library; This represents the normalized input feature vector of the current construction mesh; ) represents the normalized input feature vector of the calibration sample; This indicates the PFWD calibration mesh set.
[0106] The model applicability credibility is expressed as: .
[0107] In the formula, This represents the preset distance scale parameter. The closer the input features of the current construction grid are to existing samples in the calibration sample library, the better. The smaller, The closer the value is to 1, the higher the model's applicability; when the input features of the current construction grid significantly deviate from the coverage range of the calibration sample library, Increase A decrease indicates that the prediction results should be used with caution or that a retest of PFWD is recommended.
[0108] S4: Use the equivalent PFWD dynamic modulus prediction value, compaction degree prediction value and corresponding prediction confidence to evaluate the subgrade compaction quality.
[0109] S4.1: Determine the compaction quality level of the construction grid and generate a compaction quality distribution map based on the predicted value of the equivalent PFWD dynamic modulus, the predicted value of compaction degree, and the corresponding prediction confidence level.
[0110] In one specific implementation, the trained equivalent PFWD dynamic modulus prediction model and compaction degree prediction model are applied to all construction grids to obtain the equivalent PFWD dynamic modulus prediction value, compaction degree prediction value, and prediction confidence for each construction grid.
[0111] Based on the predicted values of the equivalent PFWD dynamic modulus, compaction degree, and compaction quality grade for each construction grid, a compaction quality distribution map is generated. This compaction quality distribution map includes an equivalent PFWD dynamic modulus distribution map, a compaction degree prediction map, a compaction quality grade map, a map of suspicious areas, a map of recommended PFWD retest points, and a map of rework compaction areas.
[0112] The equivalent PFWD dynamic modulus distribution diagram is shown as follows: .
[0113] The compaction degree prediction diagram is shown as follows: .
[0114] The compaction quality grade diagram is represented as follows: .
[0115] In the formula, This represents the dynamic modulus distribution diagram of the equivalent PFWD. This represents a compaction degree prediction diagram; A diagram showing the compaction quality grades; Indicates the first The compaction quality grade of each construction grid.
[0116] The distribution map of suspicious areas is generated based on the results of the compaction quality grade determination of the construction grid. For the... Each construction grid is used to define suspicious areas as follows: .
[0117] In the formula, Indicates a suspicious area marker; when When this occurs, it indicates that the construction grid belongs to a suspicious area or an area requiring special attention; when When this time, it indicates that the construction grid has not been marked as a suspicious area. Based on all construction grids... Values are used to generate a distribution map of suspicious areas: .
[0118] The PFWD retest point recommendation map is generated based on the retest recommendation score. For construction grids within questionable and unqualified areas, the retest recommendation score is calculated. .when Greater than the preset retest threshold At that time, the construction grid is recommended as a PFWD retest point location, represented as follows: .
[0119] In the formula, Indicates the recommended marker for PFWD retest points; when When, it indicates the first One construction grid was recommended as a PFWD resurvey point. Based on each construction grid... Values are used to generate a PFWD retest point recommendation map: .
[0120] The rework compaction zone map is determined comprehensively based on unqualified grids, retested recommended grids, the degree of insufficient compaction, and the connectivity of adjacent grids. For the first... For each construction grid, the rework compaction marker is defined as: .
[0121] In the formula, Indicates rework compaction markings; Indicates a level of non-compliance; Indicates the degree of insufficient compaction; This indicates a preset threshold for insufficient compaction. For If multiple adjacent grids are continuously distributed in the longitudinal or transverse direction, they are merged into the same rework compaction area. This is based on the construction grid... Values and their spatial adjacency relationships are used to generate a map of the rework compaction area: .
[0122] Using the above methods, the equivalent PFWD dynamic modulus distribution map and compaction degree prediction map are used to reflect the continuous quality status of the construction area; the compaction quality grade map is used to distinguish between qualified areas, doubtful areas and unqualified areas; the doubtful area distribution map is used to locate the construction grid that needs to be focused on; the PFWD retesting point recommendation map is used to guide the layout of subsequent retesting points; and the rework compaction area map is used to guide the rework treatment of low compaction or suspected low compaction areas.
[0123] In one specific implementation, the trained equivalent PFWD dynamic modulus prediction model and compaction degree prediction model are applied to all construction grids to obtain the equivalent PFWD dynamic modulus prediction value for each construction grid. Predicted compaction value and prediction credibility .
[0124] Based on the preset PFWD dynamic modulus threshold Compaction threshold and credibility threshold The quality of the compaction of the construction grid is graded.
[0125] , , .
[0126] In the formula, This indicates the acceptable threshold for the PFWD dynamic modulus. This indicates the acceptable compaction threshold. This indicates the prediction confidence threshold; This indicates the threshold for PFWD dynamic modulus to be unqualified; This indicates the threshold for unacceptable compaction, where, , Based on the above assessment, the construction area is divided into qualified, questionable, and unqualified zones.
[0127] , , .
[0128] In the formula, This represents the dynamic modulus distribution diagram of the equivalent PFWD. This represents a compaction degree prediction diagram; This represents a compaction quality grade diagram.
[0129] In this embodiment, the PFWD dynamic modulus threshold, compaction threshold, confidence threshold, and retest threshold are all preset thresholds. These preset thresholds can be determined based on road grade, fill material type, fill layer, design documents, construction quality acceptance requirements, test section test results, historical engineering test data, or on-site quality control experience, and can be adjusted according to different construction sections, different fill material types, and different fill layers. Specifically, the compaction qualification threshold and compaction failure threshold can be determined based on design documents, construction quality acceptance standards, or engineering quality control requirements; the PFWD dynamic modulus qualification threshold and PFWD dynamic modulus failure threshold can be determined based on the calibration relationship between the PFWD dynamic modulus in the test section and on-site compaction, dry density, and surface settlement; the confidence threshold can be determined based on image quality, construction parameter integrity, model applicability, and verification sample prediction error; and the retest threshold can be determined based on the retest workload, retest ratio, number of suspicious areas, and construction quality control level.
[0130] By setting the thresholds as described above, the determination of the compaction quality grade of the construction grid and the recommendation of PFWD retest points can not only meet the requirements of specifications and design control, but also be adapted to the calibration results of the field test section, avoiding the reliance on manual experience for judgment.
[0131] S4.2: Calculate the recommended score based on the compaction quality grade of the construction grid.
[0132] In one specific implementation, a retest recommendation score is further calculated for both the doubtful and unqualified areas. The recommended score for the retest of each construction grid is expressed as follows: .
[0133] In the formula, This indicates the recommended score for the retest; This indicates an indicator function that takes the value 1 if the condition within the parentheses is true, and 0 otherwise. Indicates the degree of visual abnormality; Indicates the reliability of the prediction; Indicates the degree of insufficient compaction; , , , , These are the weighting coefficients.
[0134] Among them, the degree of visual abnormality The degree of compaction can be determined by a combination of factors, including the proportion of wet patch area, the proportion of locally loose areas, abnormal wheel track depth, and uneven texture. The number of compaction passes can be determined based on the difference between the actual number of compaction passes and the designed number of compaction passes: .
[0135] In the formula, Indicates the number of rolling passes designed; This indicates the actual number of compaction passes.
[0136] When the retest recommended score Greater than the preset retest threshold When this is the case, the corresponding construction grid will be recommended as a PFWD retest point. For multiple adjacent construction grids that meet the requirements... Areas that are deemed unqualified will be merged into rework compaction areas by the system.
[0137] This embodiment also includes a digital ledger for construction quality and a quality traceability function. The image data, PFWD test data, compaction test data, moisture content, dry density, construction parameters, equivalent PFWD dynamic modulus prediction value, compaction degree prediction value, prediction confidence, compaction quality grade, retest results, rework records, construction time, filler source and construction equipment number of each construction grid are stored in the digital ledger for construction quality.
[0138] The construction quality digital ledger can be represented as follows: .
[0139] In the formula, Indicates the first The quality traceability file corresponds to each construction grid.
[0140] The digital ledger of construction quality allows for the traceability of the compaction quality of each subgrade layer. It also enables correlation analysis between the compaction quality results during construction and the results of subsequent settlement monitoring or pavement distress, which can be used to optimize the compaction parameters, retesting strategies, and construction quality control methods for subsequent filling layers.
[0141] This invention divides the construction grid into qualified, suspicious, and unqualified areas based on the equivalent PFWD dynamic modulus, compaction prediction value, visual anomaly degree, insufficient compaction degree, and prediction reliability. It also automatically recommends PFWD retesting points and rework compaction areas, which can reduce the uncertainty of human experience judgment and improve the accuracy of low compaction area identification and construction treatment.
[0142] This invention establishes a digital ledger for construction quality, which links and stores image data, PFWD detection data, compaction test data, moisture content, dry density, rolling parameters, prediction results, retest results, and rework records for each construction grid. This enables full-process traceability of the compaction quality of each fill layer in the subgrade and provides data support for subsequent settlement analysis, quality responsibility determination, and construction parameter optimization.
[0143] Example 2: Embodiment 2 of the present invention provides a roadbed compaction quality assessment system based on PFWD and visual features, comprising: The construction grid division module is used to divide the roadbed construction surface to be inspected in the construction area into several construction grids according to construction needs. The data acquisition module is used to simultaneously acquire PFWD detection data and field quality data within the construction grid, and to acquire corresponding subgrade compaction surface images and extract the visual features of the subgrade surface from the subgrade compaction surface images. The model prediction module is used to process the visual features of the subgrade surface, PFWD detection data, construction parameter vectors and field quality data using the equivalent PFWD dynamic modulus prediction model to obtain the equivalent PFWD dynamic modulus prediction value. The compaction degree prediction model is used to perform quality prediction on the visual features of the subgrade surface, field quality data, construction parameter vectors and the equivalent PFWD dynamic modulus prediction value to obtain the compaction degree prediction value and the corresponding prediction confidence level. The quality evaluation module is used to assess the compaction quality of the subgrade using the equivalent PFWD dynamic modulus prediction value, compaction degree prediction value, and corresponding prediction confidence.
[0144] Example 3: Embodiment 3 of the present invention provides a computer-readable storage medium storing a computer program adapted for loading by a processor and executing the steps in the subgrade compaction quality assessment method based on PFWD and visual features as described in Embodiment 1 of the present invention.
[0145] Example 4: Embodiment 4 of the present invention provides a computer device, the device comprising: A processor, adapted to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by the processor, implements the steps in the subgrade compaction quality assessment method based on PFWD and visual features as described in Embodiment 1 of the present invention.
[0146] The steps and methods involved in Examples 2, 3 and 4 above correspond to those in Example 1. For specific implementation details, please refer to the relevant description section of Example 1.
[0147] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.
[0148] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this application is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in or transmitted through a computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless means. The computer-readable storage medium can be any available medium that a computer can access or a data processing device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium, an optical medium, or a semiconductor medium, etc.
[0149] The above description is only a specific implementation of this application, but the protection scope of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the protection scope of this application.
Claims
1. A method for assessing roadbed compaction quality based on PFWD and visual features, characterized in that, Includes the following steps: Based on construction requirements, the roadbed construction surface to be inspected in the construction area is divided into several construction grids; Simultaneously acquire PFWD detection data and on-site quality data within the construction grid, and obtain corresponding subgrade compaction surface images, extracting the subgrade surface visual features from the subgrade compaction surface images; The equivalent PFWD dynamic modulus prediction model is used to process the visual features of the subgrade surface, PFWD detection data, construction parameter vectors and field quality data to obtain the equivalent PFWD dynamic modulus prediction value. The compaction degree prediction model is used to predict the quality of the subgrade surface visual features, field quality data, construction parameter vectors and the equivalent PFWD dynamic modulus prediction value to obtain the compaction degree prediction value and the corresponding prediction confidence. The subgrade compaction quality is assessed using the equivalent PFWD dynamic modulus prediction value, compaction degree prediction value, and corresponding prediction confidence.
2. The method for assessing roadbed compaction quality based on PFWD and visual features as described in claim 1, characterized in that, PFWD test data includes dynamic modulus and surface settlement, while field quality data includes field compaction, dry density, and moisture content.
3. The method for assessing roadbed compaction quality based on PFWD and visual features as described in claim 1, characterized in that, The specific steps for extracting visual features of the subgrade surface from images of compacted subgrade surfaces are as follows: Preprocessing and construction mesh matching operations are performed on the compacted surface image of the roadbed. Visual features of the compacted roadbed surface are extracted from the preprocessed image to form a visual feature vector. The visual feature vector includes surface texture uniformity features, wheel track depth features, particle distribution features, wet spot features, local loose features, and surface subsidence or uneven compaction features.
4. The method for assessing roadbed compaction quality based on PFWD and visual features as described in claim 1, characterized in that, The compaction parameters corresponding to each construction grid are collected to form a construction parameter vector. The compaction parameters include the number of compaction passes, compaction speed, vibration frequency, amplitude, compaction energy, and filler type. Among them, the compaction energy is characterized by the roller mass, vibration parameters, number of compaction passes, and effective area.
5. The method for assessing roadbed compaction quality based on PFWD and visual features as described in claim 1, characterized in that, The specific steps for processing roadbed surface visual features, PFWD detection data, and field quality data using the equivalent PFWD dynamic modulus prediction model are as follows: An equivalent PFWD dynamic modulus prediction model was constructed and trained. The equivalent PFWD dynamic modulus prediction model includes a visual initial prediction sub-model and a construction parameter correction sub-model. The equivalent PFWD dynamic modulus prediction model uses the PFWD dynamic modulus as a supervision label, uses the visual initial prediction sub-model to predict the visual features of the subgrade surface, and the construction parameter correction sub-model corrects the results of the visual initial prediction sub-model according to the construction parameters, moisture content and filler type.
6. The method for assessing roadbed compaction quality based on PFWD and visual features as described in claim 1, characterized in that, The specific steps for using a compaction degree prediction model to predict the quality of roadbed surface based on visual features, on-site quality data, construction parameter vectors, and the equivalent PFWD dynamic modulus are as follows: Construct and train a compaction degree prediction model; The trained compaction prediction model takes visual feature vector, construction parameter vector, moisture content, filler type and equivalent PFWD dynamic modulus prediction value as input, and uses the on-site compaction degree as the supervision label to predict quality.
7. The method for assessing roadbed compaction quality based on PFWD and visual features as described in claim 1, characterized in that, The specific steps for assessing subgrade compaction quality using the equivalent PFWD dynamic modulus prediction value, compaction degree prediction value, and corresponding prediction confidence level are as follows: Based on the predicted value of the equivalent PFWD dynamic modulus, the predicted value of compaction degree and the corresponding prediction confidence, the compaction quality level of the construction grid is determined and a compaction quality distribution map is generated. The recommended score is calculated based on the retesting of the compaction quality grade of the construction grid.
8. A roadbed compaction quality assessment system based on PFWD and visual features, characterized in that, include: The construction grid division module is used to divide the roadbed construction surface to be inspected in the construction area into several construction grids according to construction needs. The data acquisition module is used to simultaneously acquire PFWD detection data and field quality data within the construction grid, and to acquire corresponding subgrade compaction surface images, extracting the visual features of the subgrade surface from the subgrade compaction surface images; The model prediction module is used to process the visual features of the subgrade surface, PFWD detection data, construction parameter vectors and field quality data using the equivalent PFWD dynamic modulus prediction model to obtain the equivalent PFWD dynamic modulus prediction value. The compaction degree prediction model is used to perform quality prediction on the visual features of the subgrade surface, field quality data, construction parameter vectors and the equivalent PFWD dynamic modulus prediction value to obtain the compaction degree prediction value and the corresponding prediction confidence level. The quality evaluation module is used to assess the compaction quality of the subgrade using the equivalent PFWD dynamic modulus prediction value, compaction degree prediction value, and corresponding prediction confidence.
9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program adapted to be loaded by a processor and executed as described in any one of claims 1-7: the method for assessing subgrade compaction quality based on PFWD and visual features.
10. A computer device, characterized in that, include: A processor, adapted to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by the processor, implements the subgrade compaction quality assessment method based on PFWD and visual features as described in any one of claims 1-7.