Cold regeneration pavement repair management system and method

Abstract

The present application provides a cold regeneration pavement repair management system and method, which relates to the field of intelligent data analysis. It uses deep learning-based image analysis and data processing technology to regularize and time-series the temperature and humidity environmental parameters of the construction site, and then performs fine-grained implicit associations on the temperature and humidity time series features. At the same time, the surface state image of the pavement repair material is semantically encoded for performance, so as to automatically judge whether the construction conditions are suitable for construction based on the multi-feature perception between the temperature and humidity time series implicit association features and the material performance image semantic encoding features. In this way, by intelligently analyzing the temperature time series data, humidity time series data and surface images of pavement repair materials at the construction site, a comprehensive, accurate and intelligent evaluation of the construction conditions can be achieved.

Description

Technical Field

[0001] The present application relates to the field of intelligent data analysis, and more specifically, to a cold regeneration pavement repair management system and method. Background Art

[0002] With the rapid development of highway construction, cities are facing huge challenges in road network maintenance and repair. Although the rapid expansion of the highway network has significantly improved transportation efficiency, it has also brought about an increasing demand for maintenance. The traditional asphalt concrete pavement maintenance method is to cold mill the old pavement and then re-pave it with new cement-stabilized gravel materials or asphalt mixtures. Although this method solves the problem of pavement damage to a certain extent, it will lead to a large amount of waste of high-quality asphalt resources, and then cause soil and water pollution. In order to meet these challenges, cold recycling technology came into being. It achieves the reuse of old materials and improves construction efficiency and pavement performance by cutting and crushing old asphalt pavement materials on site, adding new materials, remixing, paving and compacting.

[0003] Pavement repair is an important part of road maintenance and management, and directly affects the safety and service life of roads. However, traditional pavement repair mainly relies on manual experience and simple instrument measurements to determine whether the construction site is suitable for construction. This method has some limitations. Specifically, manual judgment is highly subjective and can easily lead to inconsistent judgment results. In addition, traditional methods usually only consider the temperature and humidity values ​​of the current environment, lack time series analysis of historical data, and cannot capture the changing trends of environmental parameters in real time. At the same time, traditional methods ignore the impact of factors such as material properties and surface conditions on construction quality, resulting in an inability to fully reflect the complexity of construction conditions.

[0004] Therefore, an optimized cold regenerated pavement rehabilitation management program is needed. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiments of the present application provide a cold regeneration pavement repair management system and method.

[0006] According to one aspect of the present application, a cold regeneration pavement repair management system is provided, which includes: a material surface state image acquisition module, which is used to acquire a surface state image of a pavement repair material collected by a camera; a construction site environmental parameter acquisition module, which is used to acquire a time queue of construction site environmental parameters collected by a sensor group, wherein the construction site environmental parameters include temperature values ​​and humidity values; a temperature and humidity time series association module, which is used to perform data regularization and full-time series fine-grained implicit association on the time queue of the construction site environmental parameters to obtain temperature and humidity time series implicit association features; a material performance feature extraction module, which is used to perform material performance feature extraction on the surface state image of the pavement repair material to obtain material performance image semantic coding features; a material performance-construction condition fusion module, which is used to perform multi-feature perception adaptive coding on the material performance image semantic coding features and the temperature and humidity time series implicit association features to obtain material performance-construction condition cross-modal significant fusion coding features; a repair management result generation module, which is used to obtain a repair management result based on the material performance-construction condition cross-modal significant fusion coding features, and the repair management result is used to indicate whether the construction conditions are suitable for construction.

[0007] In the above-mentioned cold regeneration pavement repair management system, the temperature and humidity time series association module includes: a construction site environmental parameter structured sorting unit, which is used to regularize the time queue of the construction site environmental parameters according to the parameter sample dimension and arrange the data according to the time dimension to obtain a temperature time series input vector and a humidity time series input vector; a temperature and humidity full time series fine-grained association encoding unit, which is used to input the temperature time series input vector and the humidity time series input vector into a full time series fine-grained association encoder to obtain a temperature-humidity full time series fine-grained association matrix; a temperature and humidity time series implicit association feature generation unit, which is used to input the temperature-humidity full time series fine-grained association matrix into a temperature and humidity time series feature extractor based on a converter structure to obtain a temperature and humidity time series implicit association feature vector as the temperature and humidity time series implicit association feature.

[0008] In the above-mentioned cold regeneration pavement repair management system, the material performance feature extraction module is used to: input the surface state image of the pavement repair material into a material performance feature extractor based on a deep separable convolutional neural network model to obtain a material performance image semantic coding feature map as the material performance image semantic coding feature.

[0009] In the above-mentioned cold regeneration pavement repair management system, the material performance-construction condition fusion module includes: a material performance image semantic distillation unit, which is used to perform feature distillation on the material performance image semantic coding feature map to obtain a material performance image semantic distillation feature vector; a material performance-construction condition gated screening unit, which is used to perform gated implicit feature screening on the material performance image semantic distillation feature vector and the temperature and humidity time series implicit correlation feature vector to obtain the material performance-construction condition cross-modal significant fusion coding feature.

[0010] In the above-mentioned cold regeneration pavement repair management system, the material performance image semantic distillation unit is used to: perform point convolution encoding on the material performance image semantic coding feature map to obtain a material performance image semantic convolution coding feature map; perform average pooling processing on each feature matrix along the channel dimension of the material performance image semantic convolution coding feature map to obtain a material performance image semantic feature vector; and use a sigmoid function to process the material performance image semantic feature vector to obtain the material performance image semantic distillation feature vector.

[0011] In the above-mentioned cold regeneration pavement repair management system, the material performance-construction condition gating screening unit includes: a material performance-construction condition implicit association subunit, which is used to input the temperature and humidity time series implicit association feature vector and the material performance image semantic distillation feature vector into a multimodal implicit association feature capture network to obtain a material performance-construction condition implicit commonality feature vector; a primary material performance-construction condition gating calculation subunit, which is used to input the material performance-construction condition implicit commonality feature vector into a feature preliminary screening module based on the gating unit to obtain a primary material performance-construction condition feature preliminary screening weight vector; a primary material performance-construction condition joint encoding subunit, which is used to classify the material performance-construction condition feature based on the primary material performance-construction condition feature preliminary screening weight vector. The semantic coding feature map of the material performance image is used for preliminary feature screening to obtain a primary screening material performance-construction condition cross-modal joint coding feature map; a secondary material performance-construction condition gating calculation subunit is used to input the material performance-construction condition implicit common feature vector into a feature secondary screening module based on a multi-level gating unit to obtain a secondary material performance-construction condition feature screening weight vector; a secondary material performance-construction condition joint coding subunit is used to perform secondary feature screening on the primary screening material performance-construction condition cross-modal joint coding feature map based on the secondary material performance-construction condition feature screening weight vector to obtain a material performance-construction condition cross-modal significant fusion coding feature map as the material performance-construction condition cross-modal significant fusion coding feature.

[0012] In the above-mentioned cold regeneration pavement repair management system, the first-level material performance-construction condition gating calculation subunit includes: a first material performance-construction condition weight calculation secondary subunit, which is used to multiply the material performance-construction condition implicit common feature vector by the first weight matrix, and add the obtained first implicit common weight vector and the first bias vector by position to obtain the first material performance-construction condition weight vector; a first material performance-construction condition activation weight generation secondary subunit, which is used to input the first material performance-construction condition weight vector into the sigmoid function to obtain the first material performance-construction condition activation weight vector; a feature initial screening mask secondary subunit, which is used to use the mask function to generate the first material performance-construction condition activation weight vector. The active weight vector is subjected to feature initial screening and masking to obtain the primary material property-construction condition feature initial screening weight vector, wherein the feature initial screening mask secondary subunit is used to: compare each eigenvalue in the first material property-construction condition activation weight vector with a preset threshold to obtain the primary material property-construction condition feature initial screening weight vector composed of multiple updated eigenvalues; wherein, in response to the eigenvalue being greater than the second preset threshold, the original value of the eigenvalue is used as the updated eigenvalue; in response to the eigenvalue being greater than the first preset threshold and less than or equal to the second preset threshold, the value obtained by dividing the eigenvalue by two is used as the updated eigenvalue; in response to the eigenvalue being less than or equal to the first preset threshold, zero is used as the updated eigenvalue.

[0013] In the above-mentioned cold regeneration pavement repair management system, the secondary material performance-construction condition gated calculation subunit is used to: multiply the first material performance-construction condition weight vector by the second weight matrix, and add the obtained second weight vector and the second bias vector by position to obtain the second material performance-construction condition weight vector; input the second material performance-construction condition weight vector into the sigmoid function to obtain the second material performance-construction condition activation weight vector; use the mask function to perform secondary feature screening on the second material performance-construction condition activation weight vector to obtain the secondary material performance-construction condition feature screening weight vector.

[0014] In the above-mentioned cold regenerated pavement repair management system, the repair management result generation module is used to: input the material performance-construction condition cross-modal significant fusion coding feature map into the classifier-based repair management module to obtain the repair management result, and the repair management result is used to indicate whether the construction conditions are suitable for construction.

[0015] According to another aspect of the present application, a cold regenerated pavement repair management method is provided, which includes: obtaining a surface state image of a pavement repair material collected by a camera; obtaining a time queue of construction site environmental parameters collected by a sensor group, the construction site environmental parameters including temperature values ​​and humidity values; performing data regularization and full-time series fine-grained implicit association on the time queue of the construction site environmental parameters to obtain temperature and humidity time series implicit association features; performing material performance feature extraction on the surface state image of the pavement repair material to obtain material performance image semantic coding features; performing multi-feature perception adaptive coding on the material performance image semantic coding features and the temperature and humidity time series implicit association features to obtain material performance-construction condition cross-modal significant fusion coding features; based on the material performance-construction condition cross-modal significant fusion coding features, obtaining a repair management result, the repair management result is used to indicate whether the construction conditions are suitable for construction.

[0016] Compared with the prior art, the cold regeneration pavement repair management system and method provided by the present application uses deep learning-based image analysis and data processing technology to regularize and time-series the temperature and humidity environmental parameters of the construction site, and then performs fine-grained implicit association of the temperature and humidity time series features, and at the same time, performs performance semantic encoding of the surface state image of the pavement repair material, thereby automatically judging whether the construction conditions are suitable for construction based on the multi-feature perception between the temperature and humidity time series implicit association features and the material performance image semantic encoding features. In this way, by intelligently analyzing the temperature time series data, humidity time series data and surface images of pavement repair materials at the construction site, a comprehensive, accurate and intelligent evaluation of the construction conditions can be achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0018] In the attached picture: Figure 1 This is a system block diagram of a cold regeneration pavement repair management system according to an embodiment of the present application.

[0019] Figure 2 This is a data flow diagram of a cold regeneration pavement repair management system according to an embodiment of the present application.

[0020] Figure 3 It is a block diagram of the temperature and humidity timing association module in the cold regeneration pavement repair management system according to an embodiment of the present application.

[0021] Figure 4It is a block diagram of the material performance-construction condition fusion module in the cold regeneration pavement repair management system according to an embodiment of the present application.

[0022] Figure 5 It is a block diagram of a material property-construction condition gating screening unit in a cold regeneration pavement repair management system according to an embodiment of the present application.

[0023] Figure 6 The present invention is a flowchart of a cold regeneration pavement repair management method according to an embodiment of the present application. DETAILED DESCRIPTION

[0024] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0025] With the rapid development of highway construction, cities are facing increasingly severe challenges in road network maintenance and repair. The rapid expansion of the highway network has significantly improved transportation efficiency, but it has also brought about a growing demand for maintenance. The traditional asphalt concrete pavement maintenance method involves cold milling of the old pavement and re-paving with new cement-stabilized crushed stone materials or asphalt mixtures. Although this method can solve the problem of pavement damage to a certain extent, it will lead to a large waste of high-quality asphalt resources and may cause soil and water pollution. Therefore, cold recycling technology came into being. This technology achieves the reuse of old materials by cutting and crushing old asphalt pavement materials on site, adding new materials, re-mixing, paving and compacting, thereby improving construction efficiency and pavement performance.

[0026] As a key link in road maintenance and management, pavement repair has a direct impact on the safety and service life of roads. However, traditional pavement repair methods mainly rely on manual experience and simple instrument measurements to determine whether the construction site is suitable for construction. This method has certain limitations. Specifically, manual judgment is highly subjective and easily leads to inconsistency in judgment results. In addition, traditional methods usually only consider the temperature and humidity values ​​of the current environment, lack time series analysis of historical data, and cannot capture the changing trends of environmental parameters in real time. At the same time, traditional methods ignore the impact of factors such as material properties and surface conditions on construction quality, resulting in the inability to fully reflect the complexity of construction conditions.

[0027] It should be understood that cold recycling technology can efficiently add and mix cement and foamed asphalt (or emulsified asphalt) while cutting and crushing the old asphalt pavement. The cold recycled mixture that meets the road performance requirements is then directly transported to the paver hopper, re-paved in the pavement structure by the paver, and finally compacted into shape by a roller.

[0028] Based on this, the technical concept of this application is to collect the surface state image of the road repair material by the camera, and the sensor group collects the time queue of the construction site environmental parameters (temperature value and humidity value), and uses image analysis and data processing technology based on deep learning to regularize and time-series the construction site environmental parameters, and then perform fine-grained implicit association of temperature and humidity time series features, and semantically encode the performance of the surface state image of the road repair material, so as to automatically judge whether the construction conditions are suitable for construction based on the multi-feature perception between the implicit association features of the temperature and humidity time series and the semantic encoding features of the material performance image. In this way, a comprehensive multi-dimensional analysis can be carried out by combining the time series features of temperature and humidity and the image features of the surface state of the material to ensure that the evaluation of the construction conditions is more comprehensive and accurate. At the same time, the changing trend of environmental parameters can be captured in real time, and a dynamic construction condition evaluation can be provided to eliminate the subjectivity of manual judgment and achieve a comprehensive, accurate and intelligent evaluation of construction conditions.

[0029] Figure 1 This is a system block diagram of a cold regeneration pavement repair management system according to an embodiment of the present application. Figure 2 FIG. 1 is a data flow diagram of a cold regeneration pavement repair management system according to an embodiment of the present application. Figure 1 and Figure 2 As shown, in the cold regeneration pavement repair management system 100, it includes: a material surface state image acquisition module 110, which is used to acquire the surface state image of the pavement repair material collected by the camera; a construction site environmental parameter acquisition module 120, which is used to acquire the time queue of the construction site environmental parameters collected by the sensor group, and the construction site environmental parameters include temperature values ​​and humidity values; a temperature and humidity time series association module 130, which is used to perform data regularization and full-time fine-grained implicit association on the time queue of the construction site environmental parameters to obtain temperature and humidity time series implicit association features; a material performance feature extraction module 140, which is used to extract material performance features from the surface state image of the pavement repair material to obtain material performance image semantic coding features; a material performance-construction condition fusion module 150, which is used to perform multi-feature perception adaptive coding on the material performance image semantic coding features and the temperature and humidity time series implicit association features to obtain material performance-construction condition cross-modal significant fusion coding features; a repair management result generation module 160, which is used to obtain a repair management result based on the material performance-construction condition cross-modal significant fusion coding features, and the repair management result is used to indicate whether the construction conditions are suitable for construction.

[0030] In the embodiment of the present application, the material surface state image acquisition module 110 and the construction site environmental parameter acquisition module 120 are respectively used to obtain the surface state image of the pavement repair material collected by the camera and the time queue for obtaining the construction site environmental parameters collected by the sensor group, and the construction site environmental parameters include temperature values ​​and humidity values. It should be understood that the surface state image of the pavement repair material provides direct visual information of the pavement repair material, which can reflect the physical properties and potential problems of the material; and the construction site environment has a direct impact on the performance of the pavement repair material. The time queue of obtaining the two construction site environmental parameters of temperature value and humidity value can reveal the changing trend of the construction site environment. In the technical solution of the present application, the construction site environmental parameters are combined with the state of the pavement repair material for analysis, which can more scientifically and accurately judge the suitability of the construction conditions, thereby improving the effect of pavement repair.

[0031] In the embodiment of the present application, the temperature and humidity time series association module 130 is used to perform data regularization and full time series fine-grained implicit association on the time queue of the construction site environmental parameters to obtain the temperature and humidity time series implicit association characteristics. Specifically, Figure 3 FIG. 1 is a block diagram of a temperature and humidity timing correlation module in a cold regeneration pavement repair management system according to an embodiment of the present application. Figure 3 As shown, the temperature and humidity time series association module 130 includes: a construction site environmental parameter structured sorting unit 131, which is used to regularize the time queue of the construction site environmental parameters according to the parameter sample dimension and arrange the data according to the time dimension to obtain a temperature time series input vector and a humidity time series input vector; a temperature and humidity full time series fine-grained association encoding unit 132, which is used to input the temperature time series input vector and the humidity time series input vector into a full time series fine-grained association encoder to obtain a temperature-humidity full time series fine-grained association matrix; a temperature and humidity time series implicit association feature generation unit 133, which is used to input the temperature-humidity full time series fine-grained association matrix into a temperature and humidity time series feature extractor based on a converter structure to obtain a temperature and humidity time series implicit association feature vector as the temperature and humidity time series implicit association feature.

[0032] In the embodiment of the present application, the construction site environmental parameter structured arrangement unit 131 is used to regularize the time queue of the construction site environmental parameter according to the parameter sample dimension and arrange the data according to the time dimension to obtain the temperature time series input vector and the humidity time series input vector. Accordingly, considering that the temperature and humidity in the construction site environmental parameter are different physical quantities, and the temperature and humidity change with the change of time, that is, the temperature and humidity have time series characteristics in the time dimension. Therefore, in order to analyze and process the construction site environmental parameter separately, to better understand and analyze the time series change information of each parameter, in the technical scheme of the present application, the time queue of the construction site environmental parameter is regularized according to the parameter sample dimension and arranged according to the time dimension to obtain the temperature time series input vector and the humidity time series input vector, so that the environmental changes of the construction process can be understood more clearly, and a basis is provided for the formulation of the construction plan.

[0033] In an embodiment of the present application, the temperature and humidity full temporal fine-grained association encoding unit 132 is used to input the temperature temporal input vector and the humidity temporal input vector into a full temporal fine-grained association encoder to obtain a temperature-humidity full temporal fine-grained association matrix. It should be understood that the temperature and humidity in the construction process have a mutual correlation in time series. For example, asphalt mixture is easy to soften under high temperature and high humidity conditions, but may become too hard and brittle under low temperature and low humidity conditions. Based on this, in order to deeply explore the complex relationship between temperature and humidity at different time points to reveal the deeper temporal interaction and dependency between the two, in the technical solution of the present application, the temperature temporal input vector and the humidity temporal input vector are input into a full temporal fine-grained association encoder to obtain a temperature-humidity full temporal fine-grained association matrix. In particular, in one embodiment of the present application, the matrix obtained by multiplying the transposed vectors of the temperature temporal input vector and the humidity temporal input vector can be used as the temperature-humidity full temporal fine-grained association matrix. In this way, each element (eigenvalue) in the obtained temperature-humidity full-time series fine-grained association matrix represents the temporal correlation between temperature and humidity at various time points, providing data support for the subsequent identification of hidden relationships between temperature and humidity.

[0034] In an embodiment of the present application, the temperature and humidity time series implicit correlation feature generation unit 133 is used to input the temperature-humidity full time series fine-grained correlation matrix into the temperature and humidity time series feature extractor based on the converter structure to obtain the temperature and humidity time series implicit correlation feature vector as the temperature and humidity time series implicit correlation feature. Accordingly, in order to more deeply capture and mine the implicit correlation relationship between temperature and humidity in the time dimension, in the technical solution of the present application, the temperature-humidity full time series fine-grained correlation matrix is ​​input into the temperature and humidity time series feature extractor based on the converter structure to obtain the temperature and humidity time series implicit correlation feature vector. In other words, the core of the converter structure is the self-attention mechanism, which can capture the dependencies between different positions in the sequence, which is particularly useful for processing time series data, because the mutual influence of temperature and humidity at different time points may span a long time interval, so as to better characterize the implicit correlation between temperature and humidity in different time spans.

[0035] In the embodiment of the present application, the material performance feature extraction module 140 is used to extract material performance features from the surface state image of the pavement repair material to obtain the semantic coding features of the material performance image. Specifically, in the embodiment of the present application, the material performance feature extraction module is used to: input the surface state image of the pavement repair material into a material performance feature extractor based on a deep separable convolutional neural network model to obtain a material performance image semantic coding feature map as the material performance image semantic coding feature. It should be understood that the surface state image of the pavement repair material contains a large number of rich features, such as low-level edge and texture feature information or high-level semantic information, such as cracks, potholes and flatness. Based on this, in the technical solution of the present application, the surface state image of the pavement repair material is input into a material performance feature extractor based on a deep separable convolutional neural network model to obtain a material performance image semantic coding feature map. Those of ordinary skill in the art should know that the deep separable convolutional neural network model decomposes the traditional convolution operation into two steps: first, channel-by-channel convolution and then point convolution. This decomposition method significantly reduces the number of parameters of the model, thereby reducing the computational complexity. At the same time, since each channel of the input image is processed independently, it can better identify subtle differences on the surface of pavement repair materials, such as cracks, potholes and other defects, and thus more effectively capture the fine-grained features of the original input image to better characterize the performance of pavement repair materials.

[0036] In the embodiment of the present application, the material performance-construction condition fusion module 150 is used to perform multi-feature perception adaptive coding on the material performance image semantic coding features and the temperature and humidity temporal implicit correlation features to obtain material performance-construction condition cross-modal significant fusion coding features. Specifically, Figure 4FIG. 1 is a block diagram of a material performance-construction condition fusion module in a cold regeneration pavement repair management system according to an embodiment of the present application. Figure 4 As shown, the material performance-construction condition fusion module 150 includes: a material performance image semantic distillation unit 151, which is used to perform feature distillation on the material performance image semantic coding feature map to obtain a material performance image semantic distillation feature vector; a material performance-construction condition gating screening unit 152, which is used to perform gated implicit feature screening on the material performance image semantic distillation feature vector and the temperature and humidity time series implicit correlation feature vector to obtain the material performance-construction condition cross-modal significant fusion coding feature.

[0037] It should be understood that the material performance image semantic coding feature map provides visual material status information, while the temperature and humidity time series implicit correlation feature vector provides information on environmental conditions. Data of different modalities are often complementary. For example, image features can capture subtle changes in the surface of the material, while environmental parameter features can reflect the impact of construction conditions on material properties. Therefore, in order to more accurately evaluate the mutual influence and correlation between construction conditions and material properties, so as to more accurately evaluate whether the construction conditions are suitable, in the technical solution of the present application, the material performance image semantic coding feature map and the temperature and humidity time series implicit correlation feature vector are subjected to multi-feature perception adaptive encoding to obtain a material performance-construction condition cross-modal significant fusion coding feature map as a material performance-construction condition cross-modal significant fusion coding feature.

[0038] In detail, first, the feature distillation of the semantic encoding feature map of the material performance image is performed to obtain the semantic distillation feature vector of the material performance image. In particular, by performing distillation processing on the feature map, the redundant information in the feature map can be removed, and the most core and representative features can be retained, so that the surface state of the material can be evaluated more efficiently. Then, in order to capture and extract the nonlinear and complex implicit association between the temperature and humidity time series implicit association feature vector and the material performance image semantic distillation feature vector, which is crucial to understanding the relationship between material performance and construction conditions, the temperature and humidity time series implicit association feature vector and the material performance image semantic distillation feature vector are input into the multimodal implicit association feature capture network to obtain the material performance-construction condition implicit commonality feature vector. That is, the multimodal implicit association feature capture network enhances the cross-modal consistency expression by introducing a collaborative attention mechanism to better reveal the intrinsic connection between data from different sources. Then, in order to assign a weight to each feature to reflect the importance of the feature, a feature preliminary screening module based on a gating unit is used to perform gate screening on the implicit commonality feature vector to obtain a primary material performance-construction condition feature preliminary screening weight vector. In particular, the gating unit can dynamically decide which features are important and which features can be ignored. In this way, the features that contribute most to the evaluation of material properties and construction conditions can be effectively screened out. Subsequently, the material performance image semantic encoding feature map is feature-weighted based on the initial screening weight vector to enhance the parts of the feature map that are highly related to material properties and construction conditions, so as to obtain a more compact and core initial screening material performance-construction condition cross-modal joint encoding feature map. Furthermore, in order to be able to adjust the weight of each feature more finely and ensure that the final feature weight is more in line with actual needs, the implicit common feature vector is processed again using a feature secondary screening module based on a multi-level gating unit. In particular, the architecture introduces a variety of control mechanisms that can manage the information transfer process more finely. Each layer focuses on a specific type of feature extraction or pattern recognition, gradually excludes irrelevant or redundant features, retains only the most critical information, and obtains a secondary material performance-construction condition feature screening weight vector. Finally, the initial screening material performance-construction condition cross-modal joint coding feature map is weighted and refined based on the secondary screening weight vector to obtain the material performance-construction condition cross-modal significant fusion coding feature map that finally contains more significant and compact high-correlation related parts.

[0039] Specifically, in an embodiment of the present application, the material performance image semantic distillation unit is used to: perform point convolution encoding on the material performance image semantic coding feature map to obtain a material performance image semantic convolution coding feature map; perform average pooling processing on each feature matrix along the channel dimension of the material performance image semantic convolution coding feature map to obtain a material performance image semantic feature vector; and use a sigmoid function to process the material performance image semantic feature vector to obtain the material performance image semantic distillation feature vector.

[0040] Specifically, Figure 5 FIG. 1 is a block diagram of a material performance-construction condition gated screening unit in a cold regeneration pavement repair management system according to an embodiment of the present application. Figure 5 As shown, the material performance-construction condition gated screening unit 152 includes: a material performance-construction condition implicit association subunit 1521, which is used to input the temperature and humidity time series implicit association feature vector and the material performance image semantic distillation feature vector into a multimodal implicit association feature capture network to obtain a material performance-construction condition implicit commonality feature vector; a primary material performance-construction condition gated calculation subunit 1522, which is used to input the material performance-construction condition implicit commonality feature vector into a feature preliminary screening module based on a gate control unit to obtain a primary material performance-construction condition feature preliminary screening weight vector; a primary material performance-construction condition joint encoding subunit 1523, which is used to encode the material performance based on the primary material performance-construction condition feature preliminary screening weight vector. The image semantic coding feature map is used for preliminary feature screening to obtain a primary screening material performance-construction condition cross-modal joint coding feature map; a secondary material performance-construction condition gating calculation subunit 1524 is used to input the material performance-construction condition implicit common feature vector into a feature secondary screening module based on a multi-level gating unit to obtain a secondary material performance-construction condition feature screening weight vector; a secondary material performance-construction condition joint coding subunit 1525 is used to perform secondary feature screening on the primary screening material performance-construction condition cross-modal joint coding feature map based on the secondary material performance-construction condition feature screening weight vector to obtain a material performance-construction condition cross-modal significant fusion coding feature map as the material performance-construction condition cross-modal significant fusion coding feature.

[0041] More specifically, in an embodiment of the present application, the material performance-construction condition implicit association subunit is used to: cascade the temperature and humidity time series implicit association feature vector and the material performance image semantic distillation feature vector to obtain a material performance-construction condition cascade feature vector; multiply the material performance-construction condition cascade feature vector by a cascade weight matrix, and positionally add the obtained cascade weight vector and the cascade bias vector to obtain a material performance-construction condition implicit feature vector; input the material performance-construction condition implicit feature vector into a tanh function to obtain the material performance-construction condition implicit common feature vector.

[0042] More specifically, in the embodiment of the present application, the primary material performance-construction condition gating calculation subunit includes: a first material performance-construction condition weight calculation secondary subunit, which is used to multiply the material performance-construction condition implicit commonality feature vector by the first weight matrix, and add the obtained first implicit commonality weight vector and the first bias vector by position to obtain the first material performance-construction condition weight vector; a first material performance-construction condition activation weight generation secondary subunit, which is used to input the first material performance-construction condition weight vector into the sigmoid function to obtain the first material performance-construction condition activation weight vector; a feature initial screening mask secondary subunit, which is used to use the mask function to activate the first material performance-construction condition. The weight vector is subjected to feature initial screening and masking to obtain the primary material property-construction condition feature initial screening weight vector, wherein the feature initial screening mask secondary subunit is used to: compare each eigenvalue in the first material property-construction condition activation weight vector with a preset threshold to obtain the primary material property-construction condition feature initial screening weight vector composed of multiple updated eigenvalues; wherein, in response to the eigenvalue being greater than the second preset threshold, the original value of the eigenvalue is used as the updated eigenvalue; in response to the eigenvalue being greater than the first preset threshold and less than or equal to the second preset threshold, the value obtained by dividing the eigenvalue by two is used as the updated eigenvalue; in response to the eigenvalue being less than or equal to the first preset threshold, zero is used as the updated eigenvalue.

[0043] More specifically, in an embodiment of the present application, the first-level material property-construction condition joint encoding subunit is used to: perform position point multiplication on the eigenvalues ​​of each position in the first-level material property-construction condition feature preliminary screening weight vector and each feature matrix along the channel dimension in the material property image semantic coding feature map to obtain a material property-construction condition cross-modal joint encoding weighted feature map; input the material property-construction condition cross-modal joint encoding weighted feature map into a tanh function to obtain the preliminary screening material property-construction condition cross-modal joint encoding feature map.

[0044] More specifically, in an embodiment of the present application, the secondary material performance-construction condition gated calculation subunit is used to: multiply the first material performance-construction condition weight vector by the second weight matrix, and add the obtained second weight vector and the second bias vector by position to obtain the second material performance-construction condition weight vector; input the second material performance-construction condition weight vector into the sigmoid function to obtain the second material performance-construction condition activation weight vector; use the mask function to perform secondary feature screening on the second material performance-construction condition activation weight vector to obtain the secondary material performance-construction condition feature screening weight vector.

[0045] More specifically, in an embodiment of the present application, the secondary material performance-construction condition joint encoding subunit is used to: perform position point multiplication on the eigenvalues ​​of each position in the secondary material performance-construction condition feature screening weight vector and each feature matrix along the channel dimension in the primary screening material performance-construction condition cross-modal joint encoding feature map to obtain a primary screening material performance-construction condition cross-modal joint encoding weighted feature map; input the primary screening material performance-construction condition cross-modal joint encoding weighted feature map into the sigmoid function to obtain the material performance-construction condition cross-modal significant fusion encoding feature map.

[0046] In the embodiment of the present application, specifically, the material performance-construction condition fusion module is used to: perform multi-feature perception adaptive encoding on the material performance image semantic coding feature map and the temperature and humidity time series implicit correlation feature vector according to the following formula; wherein, the formula is: ;in, A semantic encoding feature map for the material performance image, To perform point convolution encoding on the feature map, To perform average pooling on each feature matrix along the channel dimension in the feature map, for function, Semantic distillation feature vector for material performance image, is the implicit correlation feature vector of the temperature and humidity time series, For cascade operation, and are the cascade weight matrix and cascade bias vector respectively, for function, is the implicit common eigenvector of material properties and construction conditions, and are the first weight matrix and the first bias vector respectively, represents matrix multiplication, is the first material property-construction condition activation weight vector, For masking operation, is the primary material performance-construction condition feature initial screening weight vector, It is the point multiplication by position. It is the cross-modal joint coding feature map of the initial screening material properties and construction conditions. and are the second weight matrix and the second bias vector, respectively, is the second material property-construction condition activation weight vector, is the secondary material properties-construction condition feature screening weight vector, is the preset threshold, It is the material performance-construction condition cross-modal significant fusion encoding feature map.

[0047] In the embodiment of the present application, the repair management result generation module 160 is used to obtain the repair management result based on the material performance-construction condition cross-modal significant fusion coding feature, and the repair management result is used to indicate whether the construction conditions are suitable for construction. Specifically, in the embodiment of the present application, the repair management result generation module is used to: input the material performance-construction condition cross-modal significant fusion coding feature map into the repair management module based on the classifier to obtain the repair management result, and the repair management result is used to indicate whether the construction conditions are suitable for construction. That is, the material performance-construction condition cross-modal significant fusion coding feature obtained by perceptual adaptive coding using the material performance image semantic coding feature map and the temperature and humidity time series implicit association feature vector is used for classification processing to automatically determine whether the construction conditions are suitable for construction. In this way, a comprehensive multi-dimensional analysis can be performed in combination with the time series features of temperature and humidity and the image features of the material surface state to ensure that the evaluation of the construction conditions is more comprehensive and accurate. At the same time, the changing trend of environmental parameters can be captured in real time, and a dynamic construction condition evaluation can be provided to eliminate the subjectivity of manual judgment and achieve a comprehensive, accurate and intelligent evaluation of construction conditions. In particular, in an embodiment of the present application, the material performance-construction condition cross-modal significant fusion coding feature map is input into a repair management module based on a classifier to obtain the repair management result, and the repair management result is used to indicate whether the construction conditions are suitable for construction. One possible implementation method may be: each material performance-construction condition cross-modal significant fusion coding feature matrix in the material performance-construction condition cross-modal significant fusion coding feature map is expanded into a one-dimensional feature vector according to a row vector or a column vector, and then cascaded to obtain a material performance-construction condition cross-modal significant fusion classification feature vector; the material performance-construction condition cross-modal significant fusion classification feature vector is fully connected encoded using the fully connected layer of the classifier to obtain a material performance-construction condition cross-modal significant fusion fully connected encoding classification feature vector; the material performance-construction condition cross-modal significant fusion fully connected encoding classification feature vector is input into the Softmax classification function of the classifier to obtain the probability value of the material performance-construction condition cross-modal significant fusion coding feature map belonging to each classification label, and the classification label includes a label used to indicate that the construction conditions are suitable for construction and a label used to indicate that the construction conditions are not suitable for construction; the classification label corresponding to the largest of the probability values ​​is determined as the repair management result.

[0048] In particular, when the material performance image semantic coding feature map and the temperature-humidity temporal implicit association feature vector respectively represent the image semantic features of the surface state image of the pavement repair material and the full-time domain temporal context association features of temperature-humidity, when performing multi-feature perception adaptive coding, the cross-modal implicit association fusion representation of the material performance-construction condition cross-modal significant fusion coding feature map will also have a long-range class regression understanding representation imbalance relative to the feature distribution field selective state space of the image semantic space domain and the parameter temporal association feature domain, thereby deviating from the source image semantics and source temporal distribution representation class understanding, affecting the accuracy of the classification results.

[0049] Preferably, the material performance-construction condition cross-modal significant fusion coding feature map is input into a classifier-based repair management module to obtain a repair management result, which includes: expanding the material performance-construction condition cross-modal significant fusion coding feature map into a material performance-construction condition cross-modal significant fusion coding feature vector; determining the number of zero eigenvalues ​​in the material performance-construction condition cross-modal significant fusion coding feature vector, and respectively calculating the reciprocal of the logarithm of the number of zero eigenvalues ​​with base two and the exponential value of the reciprocal of the number of zero eigenvalues ​​with base e to obtain a first material performance-construction condition cross-modal significant fusion coding representation value and a second material performance-construction condition cross-modal significant fusion coding representation value; calculating a power function of each eigenvalue of the material performance-construction condition cross-modal significant fusion coding feature vector with the reciprocal of the number of zero eigenvalues ​​as an exponent to obtain a material performance-construction condition cross-modal significant fusion coding structure state vector; and The performance-construction condition cross-modal significant fusion coding structure state vector is point-multiplied with the first material performance-construction condition cross-modal significant fusion coding representation value to obtain the material performance-construction condition cross-modal significant fusion coding information representation vector; the autocorrelation matrix of the material performance-construction condition cross-modal significant fusion coding structure state vector is point-multiplied with the second material performance-construction condition cross-modal significant fusion coding representation value and the weight hyperparameter respectively to obtain the material performance-construction condition cross-modal significant fusion coding regression understanding matrix; the material performance-construction condition cross-modal significant fusion coding information representation vector is matrix-multiplied with the material performance-construction condition cross-modal significant fusion coding regression understanding matrix to obtain the optimized material performance-construction condition cross-modal significant fusion coding feature vector; the optimized material performance-construction condition cross-modal significant fusion coding feature vector is input into the classifier-based repair management module to obtain the repair management result.

[0050] Here, the first material performance-construction condition cross-modal significant fusion coding representation value and the second material performance-construction condition cross-modal significant fusion coding representation value are expressed as: ;in, is the number of zero eigenvalues ​​in the material property-construction condition cross-modal significant fusion encoding eigenvector, represents the logarithmic function with base 2, represents an exponential function with the natural constant e as the base, The first material property-construction condition cross-modal significant fusion encoding represents the value, The encoding represents the value for the significant cross-modal fusion of the second material property-construction condition.

[0051] And the optimized material performance-construction condition cross-modal significant fusion encoding feature vector is expressed as: ;in, The material properties-construction conditions cross-modal significant fusion encoding feature vector, and is a row vector, The first material property-construction condition cross-modal significant fusion encoding represents the value, It means point multiplication by position. is the number of zero eigenvalues ​​in the material property-construction condition cross-modal significant fusion encoding eigenvector, To calculate the power function of each eigenvalue of the material property-construction condition cross-modal significant fusion encoding eigenvector with the inverse of the number of zero eigenvalues ​​as the exponent, The material properties-construction conditions cross-modal significant fusion encoding information representation vector, represents the weight hyperparameter, The second material property-construction condition cross-modal significant fusion encoding represents the value, for The transposed vector of The material properties-construction conditions cross-modal significant fusion encoding regression understanding matrix, represents matrix multiplication, The material properties-construction conditions are significantly fused across modalities to encode feature vectors for the optimization.

[0052] That is, the present application realizes effective long-range modeling of the material performance-construction condition cross-modal significant fusion coding feature vector under field linear complexity by modeling the vector field zero dimension of the material performance-construction condition cross-modal significant fusion coding feature vector expanded by the material performance-construction condition cross-modal significant fusion coding feature map as a global field redundant dependency, so as to perform structured correlation self-distillation between information representation and class understanding under the selective state space based on feature distribution in the feature regression process of the material performance-construction condition cross-modal significant fusion coding feature vector, thereby maintaining the overall information complexity of the feature distribution of the material performance-construction condition cross-modal significant fusion coding feature vector, while promoting the long-range dynamic class understanding balance of its feature convergence, so as to improve the class understanding consistency between the classification regression and feature extraction process, and improve the accuracy of the repair management result obtained by inputting the material performance-construction condition cross-modal significant fusion coding feature map into the classifier-based repair management module. In this way, a comprehensive multi-dimensional analysis can be performed in combination with the temporal characteristics of temperature and humidity and the image characteristics of the material surface state to ensure that the evaluation of construction conditions is more comprehensive and accurate. At the same time, it can capture the changing trends of environmental parameters in real time and provide dynamic construction condition assessment, thereby eliminating the subjectivity of manual judgment and achieving a comprehensive, accurate and intelligent assessment of construction conditions.

[0053] In summary, the cold regeneration pavement repair management system 100 based on the embodiment of the present application is explained, which uses image analysis and data processing technology based on deep learning to perform data regularization and time series arrangement of temperature and humidity environmental parameters at the construction site, and then performs fine-grained implicit association of temperature and humidity time series features, and at the same time, the surface state image of the pavement repair material is semantically encoded for performance, so as to automatically judge whether the construction conditions are suitable for construction based on the multi-feature perception between the temperature and humidity time series implicit association features and the material performance image semantic encoding features. In this way, by intelligently analyzing the temperature time series data, humidity time series data and surface images of pavement repair materials at the construction site, a comprehensive, accurate and intelligent evaluation of the construction conditions can be achieved.

[0054] As described above, the cold regeneration pavement repair management system 100 according to the embodiment of the present application can be implemented in various terminal devices, such as a server for cold regeneration pavement repair management. In one example, the cold regeneration pavement repair management system 100 according to the embodiment of the present application can be integrated into the terminal device as a software module and / or a hardware module. For example, the cold regeneration pavement repair management system 100 can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the cold regeneration pavement repair management system 100 can also be one of the many hardware modules of the terminal device.

[0055] Alternatively, in another example, the cold regenerated pavement repair management system 100 and the terminal device may also be separate devices, and the cold regenerated pavement repair management system 100 may be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0056] Figure 6 FIG. 1 is a flow chart of a cold regeneration pavement repair management method according to an embodiment of the present application. Figure 6 As shown, in the cold regeneration pavement repair management method, it includes: S110, obtaining the surface state image of the pavement repair material collected by the camera; S120, obtaining the time queue of the construction site environmental parameters collected by the sensor group, and the construction site environmental parameters include temperature values ​​and humidity values; S130, performing data regularization and full-time series fine-grained implicit association on the time queue of the construction site environmental parameters to obtain temperature and humidity time series implicit association features; S140, extracting material performance features from the surface state image of the pavement repair material to obtain material performance image semantic coding features; S150, performing multi-feature perception adaptive coding on the material performance image semantic coding features and the temperature and humidity time series implicit association features to obtain material performance-construction condition cross-modal significant fusion coding features; S160, based on the material performance-construction condition cross-modal significant fusion coding features, obtaining the repair management results, and the repair management results are used to indicate whether the construction conditions are suitable for construction.

[0057] Here, those skilled in the art will appreciate that the specific operations of each step in the above-mentioned cold regeneration pavement repair management method have been described in detail above. Figures 1 to 5 The cold regenerated pavement repair management system has been described in detail, and therefore, its repeated description will be omitted.

[0058] In summary, the cold regeneration pavement repair management method based on the embodiment of the present application is explained, which uses deep learning-based image analysis and data processing technology to regularize and time-series the temperature and humidity environmental parameters of the construction site, and then performs fine-grained implicit associations on the temperature and humidity time series features, and at the same time, semantically encodes the performance of the surface state image of the pavement repair material, so as to automatically judge whether the construction conditions are suitable for construction based on the multi-feature perception between the temperature and humidity time series implicit association features and the material performance image semantic encoding features. In this way, by intelligently analyzing the temperature time series data, humidity time series data and surface images of pavement repair materials at the construction site, a comprehensive, accurate and intelligent evaluation of the construction conditions can be achieved.

[0059] In other embodiments of the present application, a new type of in-situ cold regeneration process based on synchronous paving is also provided, which includes a single-lane single-layer foamed asphalt in-situ cold regeneration construction in which a new type of in-situ regeneration machine cooperates with a paver and a single-lane single-layer cement in-situ cold regeneration construction in which a new type of in-situ regeneration machine cooperates with a paver. In the foamed asphalt in-situ cold regeneration construction, in detail, the regeneration machine performs foamed asphalt in-situ cold regeneration on the damaged asphalt surface layer (or the asphalt surface layer and part of the base layer), and the regeneration depth can reach 30cm. The regenerated foamed asphalt cold regeneration mixture is directly transported to the hopper of the paver for paving. In addition, the regeneration process also includes double-layer in-situ cold regeneration construction. In detail, the cold regeneration machine performs foamed asphalt in-situ cold regeneration on the damaged asphalt surface layer material, and the regenerated material is transported to the material transport vehicle. Then another cold regeneration machine performs cement in-situ cold regeneration on the damaged base material. The regenerated material is directly paved by the paver. Finally, the foamed asphalt mixture in the material transport truck is paved by another paver on the structural layer that has completed cement in-situ regeneration, and the two structural layers are cured synchronously. This new in-situ cold recycling process can achieve 100% in-situ reuse of pavement materials, saving resources, and completely eliminating deep-seated pavement diseases, extending the service life. It can also reduce the round-trip transportation of materials, reduce transportation costs, and help reduce emissions and energy consumption, achieving energy conservation and emission reduction.

[0060] The basic principles of the present application are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present application are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. are required by each embodiment of the present application. In addition, the specific details of the above application are only for the purpose of illustration and ease of understanding, rather than limitation, and the above details do not limit the present application to the use of the above specific details to be implemented.

Claims

1. A cold regeneration pavement repair management system, characterized in that: include: A material surface state image acquisition module is used to acquire a surface state image of a road repair material collected by a camera; A construction site environmental parameter acquisition module is used to acquire a time queue of construction site environmental parameters collected by a sensor group, wherein the construction site environmental parameters include temperature and humidity values; a temperature and humidity time series association module is used to perform data regularization and full-time fine-grained implicit association on the time queue of the construction site environmental parameters to obtain temperature and humidity time series implicit association features; a material performance feature extraction module is used to perform material performance feature extraction on the surface state image of the pavement repair material to obtain material performance image semantic coding features; a material performance-construction condition fusion module is used to perform multi-feature perception adaptive coding on the material performance image semantic coding features and the temperature and humidity time series implicit association features to obtain material performance-construction condition cross-modal significant fusion coding features; A repair management result generation module is used to obtain a repair management result based on the material performance-construction condition cross-modal significant fusion coding feature, and the repair management result is used to indicate whether the construction condition is suitable for construction; Among them, the temperature and humidity time series association module includes: a construction site environmental parameter structured sorting unit, which is used to regularize the time queue of the construction site environmental parameters according to the parameter sample dimension and arrange the data according to the time dimension to obtain a temperature time series input vector and a humidity time series input vector; a temperature and humidity full time series fine-grained association encoding unit, which is used to input the temperature time series input vector and the humidity time series input vector into a full time series fine-grained association encoder to obtain a temperature-humidity full time series fine-grained association matrix; a temperature and humidity time series implicit association feature generation unit, which is used to input the temperature-humidity full time series fine-grained association matrix into a temperature and humidity time series feature extractor based on a converter structure to obtain a temperature and humidity time series implicit association feature vector as the temperature and humidity time series implicit association feature.

2. The cold regeneration pavement repair management system according to claim 1, characterized in that: The material performance feature extraction module is used to: input the surface state image of the pavement repair material into a material performance feature extractor based on a deep separable convolutional neural network model to obtain a material performance image semantic coding feature map as the material performance image semantic coding feature.

3. The cold regeneration pavement repair management system according to claim 2, characterized in that: The material performance-construction condition fusion module includes: a material performance image semantic distillation unit, which is used to perform feature distillation on the material performance image semantic coding feature map to obtain a material performance image semantic distillation feature vector; a material performance-construction condition gating screening unit, which is used to perform gated implicit feature screening on the material performance image semantic distillation feature vector and the temperature and humidity time series implicit correlation feature vector to obtain the material performance-construction condition cross-modal significant fusion coding feature.

4. The cold regeneration pavement repair management system according to claim 3, characterized in that: The material performance image semantic distillation unit is used to: perform point convolution encoding on the material performance image semantic coding feature map to obtain a material performance image semantic convolution coding feature map; perform average pooling processing on each feature matrix along the channel dimension of the material performance image semantic convolution coding feature map to obtain a material performance image semantic feature vector; and use a sigmoid function to process the material performance image semantic feature vector to obtain the material performance image semantic distillation feature vector.

5. The cold regeneration pavement repair management system according to claim 4, characterized in that: The material performance-construction condition gated screening unit includes: a material performance-construction condition implicit association subunit, which is used to input the temperature and humidity time series implicit association feature vector and the material performance image semantic distillation feature vector into a multimodal implicit association feature capture network to obtain a material performance-construction condition implicit common feature vector; a primary material performance-construction condition gated calculation subunit, which is used to input the material performance-construction condition implicit common feature vector into a feature preliminary screening module based on a gated unit to obtain a primary material performance-construction condition feature preliminary screening weight vector; a primary material performance-construction condition joint encoding subunit, which is used to semantically encode the material performance image based on the primary material performance-construction condition feature preliminary screening weight vector. The code feature map is used for preliminary feature screening to obtain a primary screening material performance-construction condition cross-modal joint coding feature map; a secondary material performance-construction condition gating calculation subunit is used to input the material performance-construction condition implicit common feature vector into a feature secondary screening module based on a multi-level gating unit to obtain a secondary material performance-construction condition feature screening weight vector; a secondary material performance-construction condition joint coding subunit is used to perform secondary feature screening on the primary screening material performance-construction condition cross-modal joint coding feature map based on the secondary material performance-construction condition feature screening weight vector to obtain a material performance-construction condition cross-modal significant fusion coding feature map as the material performance-construction condition cross-modal significant fusion coding feature.

6. The cold regeneration pavement repair management system according to claim 5, characterized in that: The first-level material performance-construction condition gating calculation subunit includes: a first material performance-construction condition weight calculation secondary subunit, which is used to multiply the material performance-construction condition implicit common feature vector by the first weight matrix, and add the obtained first implicit common weight vector and the first bias vector by position to obtain the first material performance-construction condition weight vector; a first material performance-construction condition activation weight generation secondary subunit, which is used to input the first material performance-construction condition weight vector into the sigmoid function to obtain the first material performance-construction condition activation weight vector; a feature initial screening mask secondary subunit, which is used to use the mask function to perform feature screening on the first material performance-construction condition activation weight vector. The feature initial screening is masked to obtain the primary material property-construction condition feature initial screening weight vector, wherein the feature initial screening mask secondary subunit is used to: compare each eigenvalue in the first material property-construction condition activation weight vector with a preset threshold to obtain the primary material property-construction condition feature initial screening weight vector composed of multiple updated eigenvalues; wherein, in response to the eigenvalue being greater than the second preset threshold, the original value of the eigenvalue is used as the updated eigenvalue; in response to the eigenvalue being greater than the first preset threshold and less than or equal to the second preset threshold, the value obtained by dividing the eigenvalue by two is used as the updated eigenvalue; in response to the eigenvalue being less than or equal to the first preset threshold, zero is used as the updated eigenvalue.

7. The cold regeneration pavement repair management system according to claim 6, characterized in that: The secondary material performance-construction condition gated calculation subunit is used to: multiply the first material performance-construction condition weight vector by the second weight matrix, and add the obtained second weight vector and the second bias vector by position to obtain the second material performance-construction condition weight vector; input the second material performance-construction condition weight vector into the sigmoid function to obtain the second material performance-construction condition activation weight vector; use the mask function to perform secondary feature screening on the second material performance-construction condition activation weight vector to obtain the secondary material performance-construction condition feature screening weight vector.

8. The cold regeneration pavement repair management system according to claim 7, characterized in that: The repair management result generation module is used to: input the material performance-construction condition cross-modal significant fusion coding feature map into the classifier-based repair management module to obtain the repair management result, and the repair management result is used to indicate whether the construction conditions are suitable for construction.

9. A cold regenerated pavement repair management method, using the cold regenerated pavement repair management system according to claim 1, characterized in that: include: Acquire a surface state image of a road repair material collected by a camera; Acquire a time queue of construction site environmental parameters collected by a sensor group, wherein the construction site environmental parameters include temperature and humidity values; perform data regularization and full-time series fine-grained implicit association on the time queue of the construction site environmental parameters to obtain a temperature and humidity time series implicit association feature; The material performance feature is extracted from the surface state image of the pavement repair material to obtain the semantic coding feature of the material performance image; the semantic coding feature of the material performance image and the temperature and humidity temporal implicit correlation feature are adaptively coded with multi-feature perception to obtain the cross-modal significant fusion coding feature of material performance and construction conditions; Based on the material performance-construction condition cross-modal significant fusion coding features, a repair management result is obtained, and the repair management result is used to indicate whether the construction conditions are suitable for construction.

Images