Quality control system and method for foamed asphalt mixture

Abstract

The present application provides a quality control system and method for foamed asphalt mixture, which relates to the field of intelligent detection. It uses machine vision-based image recognition and detection technology to grid-cut the sample image after the foamed asphalt mixture is compacted and semantically encode the local image blocks of the sample performance, and then perform spatial semantic enhancement on the semantic encoding features of the local image blocks of each sample performance, so as to intelligently judge whether the material performance meets the preset standards based on the energy global aggregation of the semantic enhancement features of the local image blocks of each sample performance after spatial semantic enhancement. In this way, by analyzing the image of the entire sample area, the quality of the overall material can be fully reflected. At the same time, it can detect subtle defects that may be overlooked by traditional methods, thereby achieving comprehensive, accurate and intelligent quality control of foamed asphalt mixture.

Description

Technical Field

[0001] The present application relates to the field of intelligent detection, and more specifically, to a quality control system and method for foamed asphalt mixture. Background Art

[0002] Foamed asphalt mixture (with some cement added) is a material between cement-stabilized macadam and asphalt concrete. It produces foam by mixing hot water with asphalt, which can well balance rigidity and flexibility, and ensure that vertical rutting deformation does not occur under high axle loads, and it is not easy to produce reflective cracks under temperature differences. It is very suitable for the structural layer of heavy-load traffic pavement. This mixture not only reduces construction costs, but also improves the durability and environmental adaptability of the pavement. Therefore, quality control of foamed asphalt mixture is particularly important in road construction and maintenance.

[0003] However, traditional methods usually manually test a limited number of samples, which makes it difficult to fully reflect the quality of the overall material. For example, a batch of foamed asphalt mixture may have local areas with higher porosity or more cracks, and these areas may not be detected in a small number of samples, thus affecting the final quality assessment results. In addition, manual inspection relies on the experience and attention of operators, and it is easy to overlook some subtle defects, such as tiny cracks, pores or color changes. Although these subtle features may not be obvious in a single sample, they may have a significant impact on the performance of the overall material. For example, tiny cracks may gradually expand during long-term use, eventually leading to early damage to the pavement.

[0004] Therefore, an optimized quality control system for foamed asphalt mixture is desired. 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 quality control system and method for a foamed asphalt mixture.

[0006] According to one aspect of the present application, a quality control system for foamed asphalt mixture is provided, which includes: a mixture sample image acquisition module, used to acquire a sample image of the foamed asphalt mixture after compaction collected by a camera; a sample image segmentation feature extraction module, used to perform grid segmentation and sample state feature extraction on the sample image of the foamed asphalt mixture after compaction to obtain a set of semantic coding feature maps of local image blocks of sample performance; a sample performance spatial enhancement module, used to perform spatial semantic perception feature enhancement on the set of semantic coding feature maps of local image blocks of sample performance to obtain a set of semantic enhancement coding feature maps of local image blocks of sample performance; a sample performance global aggregation module, used to perform global feature aggregation based on energy distribution on the set of semantic enhancement coding feature maps of local image blocks of sample performance to obtain a global sample performance significant aggregation representation; a material performance judgment module, used to obtain a quality control result based on the global sample performance significant aggregation representation, and the quality control result is used to indicate whether the material performance meets the preset standard.

[0007] In the above-mentioned quality control system of foamed asphalt mixture, the sample image segmentation feature extraction module includes: a sample image gridding segmentation unit, which is used to grid the sample image after the foamed asphalt mixture is compacted to obtain a set of sample image blocks; a sample image semantic encoding unit, which is used to input each sample image block in the set of sample image blocks into a sample state feature extractor based on a deep separable convolutional neural network model to obtain a set of semantic encoding feature maps of the sample performance local image blocks.

[0008] In the quality control system of the foamed asphalt mixture, the sample performance space enhancement module includes: a sample feature dispersion unit, which is used to perform feature dispersion on the semantic coding feature map of the sample performance local image block along the channel dimension to obtain a set of semantic pixel granularity feature vectors of the sample performance local image block; a sample image block semantic association score calculation unit, which is used to calculate the semantic association score between any two sample performance local image block semantic pixel granularity feature vectors in the set of the sample performance local image block semantic pixel granularity feature vectors to obtain a sample performance local image block semantic association score topological matrix; a sample image block semantic space modulation unit, which is used to calculate the semantic association score between any two sample performance local image block semantic pixel granularity feature vectors in the set of the sample performance local image block semantic pixel granularity feature vectors to obtain a sample performance local image block semantic association score topological matrix; The spatial distance between any two sample performance local image block semantic pixel granularity feature vectors in the set of sample performance local image block semantic pixel granularity feature vectors is used to perform spatial attention attenuation modulation on the sample performance local image block semantic association score topology matrix to obtain the sample performance local image block semantic association space soft constraint score topology matrix; a sample image block semantic enhancement unit is used to perform convolution coding-based context enhancement on the sample performance local image block semantic association space soft constraint score topology matrix and the set of sample performance local image block semantic pixel granularity feature vectors to obtain the sample performance local image block semantic enhancement coding feature map.

[0009] In the above-mentioned quality control system of foamed asphalt mixture, the sample image block semantic association score calculation unit includes: a sample pixel feature mapping subunit, which is used to map each sample performance local image block semantic pixel granularity feature vector in the set of sample performance local image block semantic pixel granularity feature vectors to the Poincare space to obtain a set of mapped sample performance local image block semantic pixel granularity feature vectors; a sample pixel feature semantic association subunit, which is used to calculate the semantic association score between any two mapped sample performance local image block semantic pixel granularity feature vectors in the set of mapped sample performance local image block semantic pixel granularity feature vectors to obtain the sample performance local image block semantic association score topological matrix composed of multiple sample performance local image block semantic association score values.

[0010] In the above-mentioned quality control system of foamed asphalt mixture, the sample image block semantic enhancement unit is used to: perform hole convolution coding on the sample performance local image block semantic association space soft constraint score topology matrix to obtain the sample performance local image block semantic association space soft constraint score topology feature matrix; input the set of sample performance local image block semantic pixel granularity feature vectors and the sample performance local image block semantic association space soft constraint score topology feature matrix into the graph convolution coding module to obtain the set of sample performance local image block context pixel granularity feature vectors; perform feature shape reshaping on the set of sample performance local image block context pixel granularity feature vectors to obtain the sample performance local image block semantic enhancement coding feature map.

[0011] In the quality control system of the foamed asphalt mixture, the sample performance global aggregation module includes: a sample performance local image block semantic enhancement coding feature dimension reduction unit, which is used to perform pooling processing of each feature matrix along the channel dimension on each sample performance local image block semantic enhancement coding feature map in the set of the sample performance local image block semantic enhancement coding feature map to obtain a set of sample performance local image block semantic enhancement coding feature vectors; a sample performance semantic static energy factor calculation unit, which is used to calculate the static energy factor of each sample performance local image block semantic enhancement coding feature vector in the set of the sample performance local image block semantic enhancement coding feature vector to obtain a set of sample performance semantic static energy factors; a sample performance semantic clustering initial center vector selection unit, which is used to select the sample performance local image block semantic enhancement coding feature vector corresponding to the largest sample performance semantic static energy factor in the set of the sample performance semantic static energy factors as the sample performance semantic clustering initial center vector; a sample performance semantic dynamic energy factor calculation unit, which is used to calculate the sample performance local image block semantic enhancement coding feature vector corresponding to the largest sample performance semantic static energy factor in the set of the sample performance semantic static energy factors The spatial span between each sample performance local image block semantic enhancement coding feature vector in the set of block semantic enhancement coding feature vectors and the initial center vector of the sample performance semantic clustering, as well as the static energy factor of each sample performance local image block semantic enhancement coding feature vector and the static energy factor of the initial center vector of the sample performance semantic clustering, calculate the dynamic aggregation energy factor of each sample performance local image block semantic enhancement coding feature vector in the set of sample performance local image block semantic enhancement coding feature vectors to obtain a set of sample performance semantic dynamic aggregation energy factors; a sample performance semantic dynamic aggregation weight factor calculation unit, used to input the set of sample performance semantic dynamic aggregation energy factors into a gated mask unit to obtain a set of sample performance semantic dynamic aggregation weight factors; a global sample performance significant aggregation representation generation unit, used to calculate the weighted sum of the set of sample performance local image block semantic enhancement coding feature vectors with the set of sample performance semantic dynamic aggregation weight factors to obtain a global sample performance significant aggregation representation vector as the global sample performance significant aggregation representation.

[0012] In the above-mentioned quality control system of foamed asphalt mixture, the sample performance semantic static energy factor calculation unit is used to: calculate the mean and variance of the semantic enhancement coding feature vector of the local image block of the sample performance to obtain the sample performance semantic mean and the sample performance semantic variance; subtract the sample performance local image block semantic enhancement coding feature vector from the sample performance semantic mean by position, and calculate the fourth power of each position of the feature vector after subtraction to obtain the sample performance semantic difference modulation vector; calculate the expected value of the sample performance semantic difference modulation vector to obtain the sample performance semantic expected value; divide the sample performance semantic expected value by the square of the sample performance semantic variance, and input the divided value into the sigmoid function to obtain the sample performance semantic static energy factor.

[0013] In the above-mentioned quality control system of foamed asphalt mixture, the sample performance semantic dynamic energy factor calculation unit is used to: multiply the static energy factor of the sample performance local image block semantic enhancement coding feature vector and the static energy factor of the sample performance semantic clustering initial center vector by a first weighting parameter to obtain a first sample performance semantic dynamic aggregation energy factor; multiply the square of the spatial span between the sample performance local image block semantic enhancement coding feature vector and the sample performance semantic clustering initial center vector by a second weighting parameter to obtain a second sample performance semantic dynamic aggregation energy factor; divide the first sample performance semantic dynamic aggregation energy factor by the second sample performance semantic dynamic aggregation energy factor to obtain the sample performance semantic dynamic aggregation energy factor.

[0014] In the above-mentioned quality control system of foamed asphalt mixture, the material performance judgment module is used to: input the global sample performance significant aggregation representation vector into the classifier-based quality control module to obtain the quality control result, and the quality control result is used to indicate whether the material performance meets the preset standard.

[0015] According to another aspect of the present application, a quality control method for a foamed asphalt mixture is provided, which includes: acquiring a sample image of the foamed asphalt mixture after compaction, which is captured by a camera; performing grid segmentation and sample state feature extraction on the sample image of the foamed asphalt mixture after compaction to obtain a set of semantic coding feature maps of local image blocks of sample performance; performing spatial semantic perception feature enhancement on the set of semantic coding feature maps of local image blocks of sample performance to obtain a set of semantic enhancement coding feature maps of local image blocks of sample performance; performing global feature aggregation based on energy distribution on the set of semantic enhancement coding feature maps of local image blocks of sample performance to obtain a global sample performance significant aggregation representation; and obtaining a quality control result based on the global sample performance significant aggregation representation, wherein the quality control result is used to indicate whether the material performance meets the preset standard.

[0016] Compared with the prior art, the quality control system and method of foamed asphalt mixture provided by the present application adopts image recognition and detection technology based on machine vision to grid-cut the sample image after the foamed asphalt mixture is compacted and semantically encode the local image blocks of the sample performance, and then perform spatial semantic enhancement on the semantic encoding features of the local image blocks of each sample performance, so as to intelligently judge whether the material performance meets the preset standards based on the energy global aggregation of the semantic enhancement features of the local image blocks of each sample performance after spatial semantic enhancement. In this way, by analyzing the image of the entire sample area, the quality of the overall material can be fully reflected. At the same time, it can detect subtle defects that may be overlooked by traditional methods, so as to achieve comprehensive, accurate and intelligent quality control of foamed asphalt mixture. 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 4 is a system block diagram of a quality control system for foamed asphalt mixture according to an embodiment of the present application.

[0019] Figure 2 Schematic diagram of data flow of a quality control system for foamed asphalt mixture according to an embodiment of the present application.

[0020] Figure 3 It is a block diagram of a sample image segmentation feature extraction module in a quality control system of foamed asphalt mixture according to an embodiment of the present application.

[0021] Figure 4 It is a block diagram of a sample performance space enhancement module in a quality control system of foamed asphalt mixture according to an embodiment of the present application.

[0022] Figure 5 It is a block diagram of a sample performance global aggregation module in a quality control system of foamed asphalt mixture according to an embodiment of the present application.

[0023] Figure 6 The present invention is a flowchart of a method for controlling the quality of foamed asphalt mixture 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] Foamed asphalt mixture (with some cement added) is a new type of material between cement-stabilized macadam and asphalt concrete. This material achieves a balance of rigidity and flexibility by mixing hot water with asphalt to produce foam, and ensures that there is no vertical rutting deformation when bearing large axle loads. At the same time, it is not easy to produce reflective cracks under temperature changes. It is particularly suitable for pavement structures that bear heavy traffic. This mixture not only reduces construction costs, but also improves the durability of the pavement and its adaptability to environmental changes. Therefore, quality control of foamed asphalt mixture is crucial during road construction and maintenance.

[0026] However, traditional testing methods usually rely on manual testing of limited samples, which can hardly fully reflect the quality of the overall material. For example, in a batch of foamed asphalt mixture, there may be areas with high local porosity or more cracks, which may not be found in the testing of a small number of samples, thus affecting the final quality assessment results. In addition, manual testing is affected by the experience and attention of operators, and it is easy to overlook some subtle defects such as tiny cracks, pores or color changes. Although these subtle features may not be obvious in a single sample, they may have a significant impact on performance in the overall material. For example, tiny cracks may gradually expand during long-term use, eventually leading to early damage to the pavement.

[0027] In response to the above technical problems, the technical concept of the present application is to collect sample images of the compacted foamed asphalt mixture by a camera, and use image recognition and detection technology based on machine vision to grid the sample images and semantically encode the local image blocks of the sample performance, and then perform spatial semantic enhancement on the semantic coding features of the local image blocks of each sample performance, so as to intelligently judge whether the material performance meets the preset standards based on the energy global aggregation of the semantic enhancement features of the local image blocks of each sample performance after spatial semantic enhancement. In this way, by collecting images of the entire sample area through the camera, a larger area can be covered to ensure that each area can be detected, thereby fully reflecting the quality of the overall material. At the same time, subtle defects that may be overlooked by traditional methods can be detected, thereby achieving comprehensive, accurate and intelligent quality control of foamed asphalt mixtures.

[0028] Figure 1 4 is a system block diagram of a quality control system for foamed asphalt mixture according to an embodiment of the present application. Figure 2 Schematic diagram of data flow of the quality control system of the foamed asphalt mixture according to the embodiment of the present application. Figure 1 and Figure 2 As shown, in the quality control system 100 of the foamed asphalt mixture, it includes: a mixture sample image acquisition module 110, which is used to acquire a sample image of the foamed asphalt mixture after being compacted, which is collected by a camera; a sample image segmentation feature extraction module 120, which is used to perform grid segmentation and sample state feature extraction on the sample image of the foamed asphalt mixture after being compacted to obtain a set of semantic coding feature maps of local image blocks of sample performance; a sample performance spatial enhancement module 130, which is used to perform spatial semantic perception feature enhancement on the set of semantic coding feature maps of local image blocks of sample performance to obtain a set of semantic enhancement coding feature maps of local image blocks of sample performance; a sample performance global aggregation module 140, which is used to perform global feature aggregation based on energy distribution on the set of semantic enhancement coding feature maps of local image blocks of sample performance to obtain a global sample performance significant aggregation representation; a material performance judgment module 150, which is used to obtain a quality control result based on the global sample performance significant aggregation representation, and the quality control result is used to indicate whether the material performance meets the preset standard.

[0029] In the embodiment of the present application, the mixture sample image acquisition module 110 is used to acquire a sample image of the foamed asphalt mixture after being compacted, which is captured by a camera. It should be understood that the sample image of the foamed asphalt mixture after being compacted contains texture information such as pores and cracks on the sample surface, surface color information after a mixture of multiple materials, and defect information such as stratification and shedding on the sample surface. By analyzing this information in the sample image, it is possible to effectively determine whether the material meets the preset standards, and then identify potential problems in advance to ensure the quality and safety of road construction and maintenance.

[0030] In the embodiment of the present application, the sample image segmentation feature extraction module 120 is used to perform grid segmentation and sample state feature extraction on the sample image after the foamed asphalt mixture is compacted to obtain a set of semantic coding feature maps of local image blocks of sample performance. Specifically, Figure 3 FIG. 4 is a block diagram of a sample image segmentation feature extraction module in a quality control system for foamed asphalt mixture according to an embodiment of the present application. Figure 3 As shown, the sample image segmentation feature extraction module 120 includes: a sample image gridding segmentation unit 121, used for gridding the sample image after the foamed asphalt mixture is compacted to obtain a set of sample image blocks; a sample image semantic coding unit 122, used for inputting each sample image block in the set of sample image blocks into a sample state feature extractor based on a deep separable convolutional neural network model to obtain a set of semantic coding feature maps of the sample performance local image blocks.

[0031] In the embodiment of the present application, the sample image gridding and segmentation unit 121 is used to grid-divide the sample image after the foamed asphalt mixture is compacted to obtain a set of sample image blocks. Accordingly, considering that different local areas in the sample image after the foamed asphalt mixture is compacted contain different state characteristics, such as different areas may show different colors, textures or other physical properties, and these subtle differences are crucial for evaluating the uniformity and performance of the material. Based on this, in the technical solution of the present application, the sample image after the foamed asphalt mixture is compacted is grid-divided to obtain a set of sample image blocks, so that by dividing the entire sample image into multiple small image blocks, the characteristics of each local area can be analyzed in more detail.

[0032] In an embodiment of the present application, the sample image semantic encoding unit 122 is used to input each sample image block in the set of sample image blocks into a sample state feature extractor based on a deep separable convolutional neural network model to obtain a set of semantic encoding feature maps of local image blocks of sample performance. It should be understood that in order to capture and extract high-level semantic information in each sample image block in the set of sample image blocks, such as cracks, pores, density, etc. of the material, to help and identify the state of the material in different regions, in the technical solution of the present application, each sample image block in the set of sample image blocks is input into a sample state feature extractor based on a deep separable convolutional neural network model to obtain a set of semantic encoding feature maps of local image blocks of sample performance. It should be known to those skilled in the art that a deep separable convolutional neural network is a special convolutional neural network architecture that decomposes the standard convolution operation into two parts: deep convolution and point-by-point convolution. Specifically, deep convolution is to perform a convolution operation on each input channel separately to extract the features of each channel, while point-by-point convolution is to use a 1×1 convolution kernel to linearly combine the output of the deep convolution, thereby achieving feature fusion between channels. This decomposition method allows the deep separable convolutional network model to maintain its feature extraction capability while significantly reducing the number of model parameters and computational complexity. Processing each sample image block through the deep separable convolutional neural network model can effectively improve the efficiency and accuracy of sample image block feature extraction, thereby more accurately judging material properties and providing strong support for the quality control of foamed asphalt mixtures.

[0033] In an embodiment of the present application, the sample performance spatial enhancement module 130 is used to perform spatial semantic perception feature enhancement on the set of sample performance local image block semantic coding feature maps to obtain a set of sample performance local image block semantic enhancement coding feature maps. Accordingly, considering that each sample performance local image block semantic coding feature map contains local features and texture details of the material state, in order to further emphasize the key features of the material state while maintaining the integrity of the material state features and reduce background interference, in the technical solution of the present application, the set of sample performance local image block semantic coding feature maps is subjected to spatial semantic perception feature enhancement to obtain a set of sample performance local image block semantic enhancement coding feature maps.

[0034] Specifically, first, the semantic encoding feature map of the sample performance local image block is feature scattered along the channel dimension to convert its shape into pixel-sized features to obtain a set of semantic pixel-sized feature vectors of the sample performance local image block. Then, in order to better characterize the semantic relevance between any two pixel-sized vectors, the set of semantic pixel-sized feature vectors of the sample performance local image block is mapped to the Poincare space to obtain a set of semantic pixel-sized feature vectors of the sample performance local image block after mapping. In particular, the Poincare space can naturally represent the hierarchical structure of the data, which is very useful for capturing features of different scales in foamed asphalt mixtures (such as tiny cracks, pores, etc.). Then, based on the Poincare distance between any two vectors as the semantic association score, the relationship between distant points can be better represented, which helps to capture the long-distance dependencies in the material, and the semantic association score topological matrix of the sample performance local image block is obtained. Next, in order to give a higher weight to the interaction between neighboring regions, the semantic association score topological matrix is ​​spatially spatially attenuated modulated based on the spatial distance between any two vectors in the set of pixel-sized feature vectors to obtain the semantic association space soft constraint score topological matrix of the sample performance local image block. Afterwards, the spatial soft constraint score topology matrix is ​​subjected to dilated convolutional coding to capture and extract the complex semantic associations between the feature vectors of each pixel granularity to obtain the semantic association spatial soft constraint score topology feature matrix of the local image block of the sample performance. Next, in order to better capture the global contextual semantic information of the sample state features, the graph convolutional coding module is used to process the set of spatial soft constraint score topology feature matrix and pixel granularity feature vectors to obtain the set of contextual pixel granularity feature vectors of the local image block of the sample performance. Finally, these contextual pixel granularity features are reshaped to obtain the semantic enhancement coding feature map of the local image block of the sample performance.

[0035] Specifically, Figure 4 FIG. 4 is a block diagram of a sample performance space enhancement module in a quality control system for foamed asphalt mixture according to an embodiment of the present application. Figure 4As shown, the sample performance space enhancement module 130 includes: a sample feature dispersion unit 131, which is used to disperse the feature of the sample performance local image block semantic coding feature map along the channel dimension to obtain a set of sample performance local image block semantic pixel granularity feature vectors; a sample image block semantic association score calculation unit 132, which is used to calculate the semantic association score between any two sample performance local image block semantic pixel granularity feature vectors in the set of sample performance local image block semantic pixel granularity feature vectors to obtain a sample performance local image block semantic association score topological matrix; a sample image block semantic space modulation unit 13 ... in the set of sample performance local image block semantic pixel granularity feature vectors in the set of sample performance local image The spatial distance between any two sample performance local image block semantic pixel granularity feature vectors in the set of sample performance local image block semantic pixel granularity feature vectors performs spatial attention attenuation modulation on the sample performance local image block semantic association score topology matrix to obtain the sample performance local image block semantic association space soft constraint score topology matrix; the sample image block semantic enhancement unit 134 is used to perform convolution coding-based context enhancement on the sample performance local image block semantic association space soft constraint score topology matrix and the set of sample performance local image block semantic pixel granularity feature vectors to obtain the sample performance local image block semantic enhancement coding feature map.

[0036] More specifically, in an embodiment of the present application, the sample image block semantic association score calculation unit includes: a sample pixel feature mapping subunit, which is used to map each sample performance local image block semantic pixel granularity feature vector in the set of the sample performance local image block semantic pixel granularity feature vectors to the Poincare space to obtain a set of mapped sample performance local image block semantic pixel granularity feature vectors; a sample pixel feature semantic association subunit, which is used to calculate the semantic association score between any two mapped sample performance local image block semantic pixel granularity feature vectors in the set of mapped sample performance local image block semantic pixel granularity feature vectors to obtain the sample performance local image block semantic association score topological matrix composed of multiple sample performance local image block semantic association score values.

[0037] More specifically, in an embodiment of the present application, the sample pixel feature mapping subunit is used to: multiply the first weight matrix with the sample performance local image block semantic pixel granularity feature vector to obtain a sample performance weight vector; multiply the sample performance weight vector with the second weight matrix to obtain the mapped sample performance local image block semantic pixel granularity feature vector.

[0038] More specifically, in the embodiment of the present application, the sample pixel feature semantic association subunit is used to: extract the first pixel from the set of semantic pixel granularity feature vectors of the local image block of the mapped sample performance. After the mapping, the sample performance of the local image block semantic pixel granularity feature vector and the The mapped sample performance local image block semantic pixel granularity feature vector; calculate the first The sample performance after mapping is the local image block semantic pixel granularity feature vector and the The pixel granularity difference vector of the local image block of the sample performance after mapping is obtained by calculating the position difference between the semantic pixel granularity feature vectors of the local image block of the sample performance after mapping, and calculating the square of the norm of the pixel granularity difference vector of the local image block of the sample performance after mapping to obtain the difference norm value of the local image block of the sample performance after mapping; calculating the constant one and the first The difference between the square of the norm of the semantic pixel granularity feature vector of the local image block after the mapping is used to obtain the first The semantic difference of the local image block of the sample performance after mapping; calculating the constant one and the first The difference between the square of the norm of the semantic pixel granularity feature vector of the local image block after the mapping is used to obtain the first The semantic difference of the local image block of the sample performance after mapping; Calculate the The semantic difference of the local image block of the sample performance after mapping is consistent with the The product of the semantic difference values ​​of the local image blocks of the mapped sample performance is obtained by multiplying the difference norm value of the local image blocks of the mapped sample performance by the semantic product value of the local image blocks of the mapped sample performance. The obtained value is first multiplied by a constant two and then added to a constant one to obtain a semantic metric value of the local image blocks of the mapped sample performance. The semantic metric value of the local image blocks of the mapped sample performance is input into an arccosh function to obtain the first The sample performance after mapping is the local image block semantic pixel granularity feature vector and the The sample performance local image block semantic association score value between the mapped sample performance local image block semantic pixel granularity feature vectors.

[0039] More specifically, in an embodiment of the present application, the sample image block semantic space modulation unit is used to: calculate the spatial distance between the semantic pixel granularity feature vectors of any two sample performance local image blocks to obtain a set of sample performance local image block spatial distance values; calculate an exponential function with the natural constant e as the base and each sample performance local image block spatial distance value in the set of sample performance local image block spatial distance values ​​as the exponent to obtain a set of sample performance local image block spatial distance index values; multiply the set of sample performance local image block spatial distance index values ​​by the attenuation modulation value by position to obtain a set of sample performance local image block spatial distance weighted values; divide each sample performance local image block semantic association score value in the sample performance local image block semantic association score topology matrix by the corresponding sample performance local image block spatial distance weighted value to obtain the sample performance local image block semantic association space soft constraint score topology matrix composed of multiple sample performance local image block semantic association space soft constraint score values.

[0040] More specifically, in an embodiment of the present application, the sample image block semantic enhancement unit is used to: perform hole convolution coding on the sample performance local image block semantic association space soft constraint score topology matrix to obtain the sample performance local image block semantic association space soft constraint score topology feature matrix; input the set of sample performance local image block semantic pixel granularity feature vectors and the sample performance local image block semantic association space soft constraint score topology feature matrix into a graph convolution coding module to obtain a set of sample performance local image block context pixel granularity feature vectors; perform feature reshaping on the set of sample performance local image block context pixel granularity feature vectors to obtain the sample performance local image block semantic enhancement coding feature map.

[0041] In an embodiment of the present application, the sample performance space enhancement module is used to: perform spatial semantic perception feature enhancement processing on the sample performance local image block semantic coding feature map according to the following formula, so as to obtain the sample performance local image block semantic enhancement coding feature map; wherein the formula is expressed as: ;in, is the semantic encoding feature map of the local image block of the sample performance, and , is the set of real numbers, and Respectively The width and height of each feature matrix along the channel dimension, express The number of channels, To reshape the feature map, is the first in the set of semantic pixel granularity feature vectors of the local image block of the sample performance The sample performance local image block semantic pixel granularity feature vector, and , is the set of semantic pixel granularity feature vectors of the local image block of the sample performance, and are the first weight matrix and the second weight matrix respectively, is the set of semantic pixel granularity feature vectors of local image blocks of sample performance after mapping, are respectively the first, second, and third in the set of semantic pixel granularity feature vectors of the local image block of the sample performance after mapping. , and After mapping, the sample performance local image block semantic pixel granularity feature vector, is the square of the one-norm of the vector, is the inverse hyperbolic cosine function, For calculation and The sample performance between the local image patch semantic association score, is the eigenvalue of each position in the sample performance local image block semantic association score topology matrix, for and The spatial distance between is the attenuation modulation value, is the eigenvalue of each position in the topological matrix of the soft constraint score of the semantic association space of the local image block of the sample performance, is the sample performance local image patch semantic association space soft constraint score topology matrix, is the dilated convolutional coding operation, is the sample performance local image patch semantic association space soft constraint score topological feature matrix, is the graph convolutional encoding operation, is the set of pixel-granular feature vectors of the local image block context of the sample performance, It is the sample performance local image block semantic enhancement encoding feature map.

[0042] In an embodiment of the present application, the sample performance global aggregation module 140 is used to perform global feature aggregation based on energy distribution on the set of semantically enhanced coded feature maps of local image blocks of the sample performance to obtain a global sample performance significantly aggregated representation. Accordingly, considering that in order to aggregate the semantic feature information of all sample performance local image blocks to form a global feature representation that can comprehensively reflect the performance of the entire sample, rather than just the characteristics of the local area, so as to evaluate the overall performance of the sample, such as uniformity, strength, durability, etc., in the technical solution of the present application, the set of semantically enhanced coded feature maps of local image blocks of the sample performance is subjected to global feature aggregation based on energy distribution to obtain a global sample performance significantly aggregated representation vector as a global sample performance significantly aggregated representation.

[0043] In detail, first, each sample performance local image block semantic enhancement coding feature map in the set of the sample performance local image block semantic enhancement coding feature map is pooled along the channel dimension to retain the key information in each feature while speeding up the processing efficiency, and a set of sample performance local image block semantic enhancement coding feature vectors is obtained. And the static energy factor of each sample performance local image block semantic enhancement coding feature vector is calculated to evaluate the importance of each feature, and a set of sample performance semantic static energy factors is obtained. Then, in order to find and select a representative feature, the sample performance local image block semantic enhancement coding feature vector corresponding to the largest sample performance semantic static energy factor is selected as the initial center vector of sample performance semantic clustering. Then, in order to consider the importance of the intrinsic features of each data point and to more accurately characterize the interaction between features, the dynamic aggregation energy factor of each sample performance local image block semantic enhancement coding feature vector is calculated based on the spatial span and static energy factor of each sample performance local image block semantic enhancement coding feature vector and the initial center vector. Subsequently, the set of each dynamic aggregation energy factor is input into the gated mask unit to convert it into a set of dynamic aggregation weight factors, so as to dynamically adjust these energy factors and retain the key ones with higher contribution. Finally, the set of semantic enhancement encoding feature vectors of local image blocks of sample performance is weighted and summed with each weight factor in the set of weight factors as weights to aggregate the material state features in the entire area, and the final global sample performance significant aggregation representation vector is obtained.

[0044] Specifically, Figure 5 FIG. 4 is a block diagram of a sample performance global aggregation module in a quality control system for foamed asphalt mixture according to an embodiment of the present application. Figure 5As shown, the sample performance global aggregation module 140 includes: a sample performance local image block semantic enhancement coding feature dimension reduction unit 141, which is used to perform pooling processing of each feature matrix along the channel dimension on each sample performance local image block semantic enhancement coding feature map in the set of the sample performance local image block semantic enhancement coding feature map to obtain a set of sample performance local image block semantic enhancement coding feature vectors; a sample performance semantic static energy factor calculation unit 142, which is used to calculate the static energy factor of each sample performance local image block semantic enhancement coding feature vector in the set of the sample performance local image block semantic enhancement coding feature vector to obtain a set of sample performance semantic static energy factors; a sample performance semantic clustering initial center vector selection unit 143, which is used to select the sample performance local image block semantic enhancement coding feature vector corresponding to the largest sample performance semantic static energy factor in the set of the sample performance semantic static energy factors as the sample performance semantic clustering initial center vector; a sample performance semantic dynamic energy factor calculation unit 144, which is used to calculate the sample performance local image block semantic enhancement coding feature vector based on the sample performance local image block semantic enhancement coding feature vector. The spatial span between each sample performance local image block semantic enhancement coding feature vector in the set of strong coding feature vectors and the initial center vector of the sample performance semantic clustering, as well as the static energy factor of each sample performance local image block semantic enhancement coding feature vector and the static energy factor of the initial center vector of the sample performance semantic clustering, calculate the dynamic aggregation energy factor of each sample performance local image block semantic enhancement coding feature vector in the set of sample performance local image block semantic enhancement coding feature vectors to obtain a set of sample performance semantic dynamic aggregation energy factors; a sample performance semantic dynamic aggregation weight factor calculation unit 145 is used to input the set of sample performance semantic dynamic aggregation energy factors into a gated mask unit to obtain a set of sample performance semantic dynamic aggregation weight factors; a global sample performance significant aggregation representation generation unit 146 is used to calculate the weighted sum of the set of sample performance local image block semantic enhancement coding feature vectors with the set of sample performance semantic dynamic aggregation weight factors to obtain a global sample performance significant aggregation representation vector as the global sample performance significant aggregation representation.

[0045] More specifically, in an embodiment of the present application, the sample performance semantic static energy factor calculation unit is used to: calculate the mean and variance of the sample performance local image block semantic enhancement coding feature vector to obtain the sample performance semantic mean and the sample performance semantic variance; subtract the sample performance local image block semantic enhancement coding feature vector from the sample performance semantic mean by position, and calculate the fourth power of each position of the subtracted feature vector to obtain the sample performance semantic difference modulation vector; calculate the expected value of the sample performance semantic difference modulation vector to obtain the sample performance semantic expected value; divide the sample performance semantic expected value by the square of the sample performance semantic variance, and input the divided value into the sigmoid function to obtain the sample performance semantic static energy factor.

[0046] More specifically, in an embodiment of the present application, the sample performance semantic dynamic energy factor calculation unit is used to: multiply the static energy factor of the sample performance local image block semantic enhancement coding feature vector and the static energy factor of the sample performance semantic clustering initial center vector by a first weighting parameter to obtain a first sample performance semantic dynamic aggregation energy factor; multiply the square of the spatial span between the sample performance local image block semantic enhancement coding feature vector and the sample performance semantic clustering initial center vector by a second weighting parameter to obtain a second sample performance semantic dynamic aggregation energy factor; divide the first sample performance semantic dynamic aggregation energy factor by the second sample performance semantic dynamic aggregation energy factor to obtain the sample performance semantic dynamic aggregation energy factor.

[0047] More specifically, in an embodiment of the present application, the sample performance semantic dynamic aggregation weight factor calculation unit is used to: use a sigmoid function to normalize the set of sample performance semantic dynamic aggregation energy factors to obtain a set of normalized sample performance semantic dynamic aggregation energy factors; mask the set of normalized sample performance semantic dynamic aggregation energy factors to obtain the set of sample performance semantic dynamic aggregation weight factors, wherein, in response to each normalized sample performance semantic dynamic aggregation energy factor in the set of normalized sample performance semantic dynamic aggregation energy factors being greater than a predetermined threshold, the normalized sample performance semantic dynamic aggregation energy factor is set to the original value, and the rest are set to zero, so as to obtain the set of sample performance semantic dynamic aggregation weight factors.

[0048] In the embodiment of the present application, specifically, the sample performance global aggregation module is used to: perform global feature aggregation processing based on energy distribution on the set of semantic enhancement coding feature maps of the sample performance local image blocks according to the following formula; wherein the formula is expressed as: ;in, A set of sample performance local image block semantic enhancement coding feature vectors obtained by performing pooling processing on each feature matrix along the channel dimension for each sample performance local image block semantic enhancement coding feature map in the set of sample performance local image block semantic enhancement coding feature maps, are the first, second, and third in the set of semantic enhancement coding feature vectors of the sample performance local image block. and Sample performance local image block semantic enhancement encoding feature vector, It is The eigenvalues ​​of each position in the semantic enhancement encoding feature vector of the local image block of the sample performance are To calculate the expected value, and They are The mean and variance of yes function, yes The corresponding sample performance semantic static energy factor, To select the maximum value corresponding to value, is the maximum matching value, is the initial center vector of sample performance semantic clustering, yes The corresponding sample performance semantic static energy factor, express and The space span between and is the weighting parameter, for The corresponding sample performance semantic dynamic aggregation energy factor, yes The corresponding normalized sample performance semantic dynamic aggregation energy factor, For masking, is a predetermined threshold, yes The corresponding sample performance semantic dynamic aggregation weight factor, is the number of vectors in the set of semantic enhancement encoding feature vectors of the local image block of the sample performance, is the global sample performance significant aggregate representation vector.

[0049] In an embodiment of the present application, the material performance judgment module 150 is used to obtain a quality control result based on the global sample performance significant aggregation representation, and the quality control result is used to indicate whether the material performance meets the preset standard. Specifically, in an embodiment of the present application, the material performance judgment module is used to: input the global sample performance significant aggregation representation vector into a classifier-based quality control module to obtain the quality control result, and the quality control result is used to indicate whether the material performance meets the preset standard. That is, the global sample performance significant aggregation representation obtained by global feature aggregation using the set of semantically enhanced coded feature maps of the sample performance local image blocks is used for classification processing, so as to intelligently determine whether the material performance meets the preset standard. In this way, by collecting images of the entire sample area through the camera, it can be ensured that each area can be detected. At the same time, subtle defects that may be overlooked by traditional methods can be detected, thereby achieving comprehensive, accurate and intelligent quality control of foamed asphalt mixtures. In particular, in an embodiment of the present application, the global sample performance significant aggregation representation vector is input into a quality control module based on a classifier to obtain the quality control result, and the quality control result is used to indicate whether the material performance meets the preset standard. One possible implementation method may be: using the fully connected layer of the classifier to fully connect encode the global sample performance significant aggregation representation vector to obtain a global sample performance significant aggregation fully connected encoding feature vector; inputting the global sample performance significant aggregation fully connected encoding feature vector into the Softmax classification function of the classifier to obtain the probability value of the global sample performance significant aggregation representation vector belonging to each classification label, and the classification labels include those for indicating that the material performance meets the preset standard and those for indicating that the material performance does not meet the preset standard; and determining the classification label corresponding to the largest of the probability values ​​as the quality control result.

[0050] In particular, when each sample performance local image block semantic enhancement coding feature map in the set of sample performance local image block semantic enhancement coding feature maps respectively represents the local spatial distribution enhanced image semantic features of the corresponding sample image block in the local image semantic space domain, when performing global feature aggregation based on energy distribution, the global image semantic space domain aggregated image semantic representation of the global sample performance significant aggregation representation vector will also have a long-range class regression understanding representation imbalance relative to the feature distribution field selective state space of each local image semantic space domain, thereby deviating from the source image semantic feature representation class understanding and affecting the accuracy of the classification result.

[0051] Preferably, the global sample performance significant aggregation representation vector is input into a classifier-based quality control module to obtain a quality control result, including: determining the number of zero eigenvalues ​​in the global sample performance significant aggregation representation 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 global sample performance significant aggregation representation value and a second global sample performance significant aggregation representation value; calculating a power function of each eigenvalue of the global sample performance significant aggregation representation vector with the reciprocal of the number of zero eigenvalues ​​as an exponent to obtain a global sample performance significant aggregation structural state vector; and inputting the global sample performance significant aggregation representation vector into a classifier-based quality control module to obtain a quality control result. The structural state vector is dot-multiplied with the first global sample performance significant aggregation representation value to obtain a global sample performance significant aggregation information representation vector; the autocorrelation matrix of the global sample performance significant aggregation structural state vector is dot-multiplied with the second global sample performance significant aggregation representation value and the weight hyperparameter respectively to obtain a global sample performance significant aggregation regression understanding matrix; the global sample performance significant aggregation information representation vector is matrix-multiplied with the global sample performance significant aggregation regression understanding matrix to obtain an optimized global sample performance significant aggregation representation vector; the optimized global sample performance significant aggregation representation vector is input into the classifier-based quality control module to obtain the quality control result.

[0052] Here, the first global sample performance significant aggregate representation value and the second global sample performance significant aggregate representation value are expressed as: ;in, is the number of zero eigenvalues ​​in the global sample performance significant aggregation representation vector, represents the logarithmic function with base 2, represents an exponential function with the natural constant e as the base, The significant aggregated representation value for the first global sample performance is, The performance for the second global sample significantly aggregates the indicated values.

[0053] And the optimized global sample performance significant aggregation representation vector is expressed as: ;in, is the global sample performance significant aggregate representation vector, and is a row vector, is the number of zero eigenvalues ​​in the global sample performance significant aggregation representation vector, The significant aggregated representation value for the first global sample performance is, To calculate each eigenvalue of the global sample performance significant aggregation representation vector, a power function with the inverse of the number of zero eigenvalues ​​as the exponent, It means point multiplication by position. is the global sample performance significant aggregation information representation vector, is the weight hyperparameter, The performance of the second global sample shows significant aggregated values, Expressed as The transposed vector of The significant aggregate regression comprehension matrix for the global sample performance, represents matrix multiplication, The global sample performance for the optimization is significantly aggregated into representation vectors.

[0054] That is, the present application realizes effective long-range modeling of the global sample performance significant aggregation representation vector under the field linear complexity by modeling the vector field zero dimension of the global sample performance significant aggregation representation vector as the global field redundant dependency, so as to perform structured correlation self-distillation between information representation and class understanding based on the selective state space of feature distribution in the feature regression process of the global sample performance significant aggregation representation vector, thereby maintaining the overall information complexity of the feature distribution of the global sample performance significant aggregation representation 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 processes, and improve the accuracy of the quality control result obtained by inputting the global sample performance significant aggregation representation vector into the classifier-based quality control module.

[0055] In summary, the quality control system 100 of the foamed asphalt mixture based on the embodiment of the present application is explained, which uses image recognition and detection technology based on machine vision to grid the sample image after the foamed asphalt mixture is compacted and semantically encode the local image block of the sample performance, and then perform spatial semantic enhancement on the semantic encoding features of the local image block of each sample performance, so as to intelligently judge whether the material performance meets the preset standard according to the energy global aggregation of the semantic enhancement features of the local image block of each sample performance after the spatial semantic enhancement. In this way, by analyzing the image of the entire sample area, the quality of the overall material can be fully reflected. At the same time, it can detect subtle defects that may be overlooked by traditional methods, so as to achieve comprehensive, accurate and intelligent quality control of the foamed asphalt mixture.

[0056] Figure 6 FIG. 1 is a flow chart of a method for controlling the quality of foamed asphalt mixture according to an embodiment of the present application. Figure 6As shown, in the quality control method of foamed asphalt mixture, it includes: S110, obtaining a sample image of the foamed asphalt mixture after compaction collected by a camera; S120, gridding and segmenting the sample image of the foamed asphalt mixture after compaction and extracting sample state features to obtain a set of semantic coding feature maps of local image blocks of sample performance; S130, performing spatial semantic perception feature enhancement on the set of semantic coding feature maps of local image blocks of sample performance to obtain a set of semantic enhancement coding feature maps of local image blocks of sample performance; S140, performing global feature aggregation based on energy distribution on the set of semantic enhancement coding feature maps of local image blocks of sample performance to obtain a global sample performance significant aggregation representation; S150, obtaining a quality control result based on the global sample performance significant aggregation representation, and the quality control result is used to indicate whether the material performance meets the preset standard.

[0057] Here, those skilled in the art will appreciate that the specific operations of each step in the above-mentioned method for controlling the quality of foamed asphalt mixture have been described in detail above. Figures 1 to 5 The quality control system of the foamed asphalt mixture has been described in detail, and therefore, its repeated description will be omitted.

[0058] In summary, the quality control method of the foamed asphalt mixture based on the embodiment of the present application is explained, which uses image recognition and detection technology based on machine vision to grid the sample image after the foamed asphalt mixture is compacted and semantically encode the local image blocks of the sample performance, and then perform spatial semantic enhancement on the semantic encoding features of the local image blocks of each sample performance, so as to intelligently judge whether the material performance meets the preset standards based on the energy global aggregation of the semantic enhancement features of the local image blocks of each sample performance after spatial semantic enhancement. In this way, by analyzing the image of the entire sample area, the quality of the overall material can be fully reflected. At the same time, it can detect subtle defects that may be overlooked by traditional methods, thereby achieving comprehensive, accurate and intelligent quality control of the foamed asphalt mixture.

[0059] In other embodiments of the present application, there is also provided an in-situ cold regeneration process for asphalt pavement, which utilizes the oil surface and base material of the original pavement, mills in situ, adds special materials (foamed asphalt, cement), mixes in situ and directly spreads and compacts, and uses it as the base of the highway. Here, by adding special materials, the stability of the recycled materials can be effectively improved, thereby improving the durability and performance of the pavement. In simple terms, the original pavement materials, foamed asphalt, cement, etc. are used to make a foamed asphalt recycled mixture, and the prepared foamed asphalt recycled mixture is used to make a foamed asphalt cold recycled layer to replace the semi-rigid material paving the base of the road, so as to effectively ensure the quality of highway maintenance. Through the cold regeneration process, the recycling of the stone and asphalt of the original pavement asphalt concrete is realized, and the use of new materials for road maintenance is greatly reduced without generating any construction waste. And because the cold recycled mixture can be mixed under normal temperature and paved and rolled at normal temperature, it greatly reduces the dependence on energy and reduces energy consumption, while also reducing the emission of carbon dioxide and reducing environmental pollution.

Claims

1. A quality control system for foamed asphalt mixture, characterized in that: include: A mixture sample image acquisition module is used to acquire a sample image of the foamed asphalt mixture after compaction collected by a camera; A sample image segmentation feature extraction module is used to perform grid segmentation and sample state feature extraction on the sample image after the foamed asphalt mixture is compacted to obtain a set of semantic coding feature maps of local image blocks of sample performance; A sample performance spatial enhancement module, used for performing spatial semantic perception feature enhancement on the set of sample performance local image block semantic coding feature maps to obtain a set of sample performance local image block semantic enhancement coding feature maps; A sample performance global aggregation module, used for performing global feature aggregation based on energy distribution on a set of semantic enhancement coding feature maps of local image blocks of the sample performance to obtain a significant aggregation representation of global sample performance; A material performance judgment module, used to obtain a quality control result based on the significant aggregation expression of the global sample performance, and the quality control result is used to indicate whether the material performance meets the preset standard; Wherein, the sample performance space enhancement module includes: A sample feature dispersion unit, used for performing feature dispersion on the semantic coding feature map of the sample performance local image block along the channel dimension to obtain a set of semantic pixel granularity feature vectors of the sample performance local image block; A sample image block semantic association score calculation unit, used to calculate the semantic association score between any two sample performance local image block semantic pixel granularity feature vectors in the set of sample performance local image block semantic pixel granularity feature vectors to obtain a sample performance local image block semantic association score topological matrix; A sample image block semantic space modulation unit is used to perform spatial attention attenuation modulation on the sample performance local image block semantic association score topology matrix based on the spatial distance between any two sample performance local image block semantic pixel granularity feature vectors in the set of the sample performance local image block semantic pixel granularity feature vectors to obtain the sample performance local image block semantic association space soft constraint score topology matrix; The sample image block semantic enhancement unit is used to perform context enhancement based on convolution coding on the set of the sample performance local image block semantic association space soft constraint score topology matrix and the sample performance local image block semantic pixel granularity feature vector to obtain the sample performance local image block semantic enhancement coding feature map.

2. The quality control system of foamed asphalt mixture according to claim 1, characterized in that: The sample image segmentation feature extraction module comprises: A sample image gridding and segmenting unit is used to grid-divide the sample image after the foamed asphalt mixture is compacted to obtain a set of sample image blocks; The sample image semantic coding unit is used to input each sample image block in the set of sample image blocks into a sample state feature extractor based on a deep separable convolutional neural network model to obtain a set of semantic coding feature maps of the sample performance local image blocks.

3. The quality control system of foamed asphalt mixture according to claim 2, characterized in that: The sample image block semantic association score calculation unit includes: A sample pixel feature mapping subunit, used for mapping each sample performance local image block semantic pixel granularity feature vector in the set of sample performance local image block semantic pixel granularity feature vectors to a Poincare space to obtain a set of mapped sample performance local image block semantic pixel granularity feature vectors; The sample pixel feature semantic association subunit is used to calculate the semantic association score between any two mapped sample performance local image block semantic pixel granularity feature vectors in the set of mapped sample performance local image block semantic pixel granularity feature vectors to obtain the sample performance local image block semantic association score topological matrix composed of multiple sample performance local image block semantic association score values.

4. The quality control system of foamed asphalt mixture according to claim 3, characterized in that: The sample image block semantic enhancement unit is used to: Performing dilated convolution coding on the sample performance local image block semantic association space soft constraint score topology matrix to obtain the sample performance local image block semantic association space soft constraint score topology feature matrix; Inputting the set of semantic pixel granularity feature vectors of the sample performance local image block and the semantic association space soft constraint score topological feature matrix of the sample performance local image block into a graph convolutional coding module to obtain a set of context pixel granularity feature vectors of the sample performance local image block; The set of context pixel granularity feature vectors of the sample performance local image block is reshaped to obtain a semantic enhancement coding feature map of the sample performance local image block.

5. The quality control system of foamed asphalt mixture according to claim 4, characterized in that: The sample performance global aggregation module includes: A sample performance local image block semantic enhancement coding feature dimensionality reduction unit, used for performing pooling processing of each feature matrix along the channel dimension on each sample performance local image block semantic enhancement coding feature map in the set of sample performance local image block semantic enhancement coding feature maps to obtain a set of sample performance local image block semantic enhancement coding feature vectors; A sample performance semantic static energy factor calculation unit, used to calculate the static energy factor of each sample performance local image block semantic enhancement coding feature vector in the set of sample performance local image block semantic enhancement coding feature vectors to obtain a set of sample performance semantic static energy factors; A sample performance semantic clustering initial center vector selection unit is used to select a sample performance local image block semantic enhancement coding feature vector corresponding to the largest sample performance semantic static energy factor in the set of sample performance semantic static energy factors as the sample performance semantic clustering initial center vector; A sample performance semantic dynamic energy factor calculation unit, used to calculate the dynamic aggregation energy factor of each sample performance local image block semantic enhancement coding feature vector in the set of the sample performance local image block semantic enhancement coding feature vector based on the spatial span between each sample performance local image block semantic enhancement coding feature vector in the set of the sample performance local image block semantic enhancement coding feature vector and the sample performance semantic clustering initial center vector and the static energy factor of each sample performance local image block semantic enhancement coding feature vector and the static energy factor of the sample performance semantic clustering initial center vector to obtain a set of sample performance semantic dynamic aggregation energy factors; A sample performance semantic dynamic aggregation weight factor calculation unit, used for inputting the set of sample performance semantic dynamic aggregation energy factors into a gated mask unit to obtain a set of sample performance semantic dynamic aggregation weight factors; A global sample performance significant aggregation representation generation unit is used to calculate the weighted sum of the set of sample performance local image block semantic enhancement coding feature vectors with the set of sample performance semantic dynamic aggregation weight factors to obtain a global sample performance significant aggregation representation vector as the global sample performance significant aggregation representation.

6. The quality control system of foamed asphalt mixture according to claim 5, characterized in that: The sample performance semantic static energy factor calculation unit is used to: Calculating the mean and variance of the semantic enhancement coding feature vector of the sample performance local image block to obtain the sample performance semantic mean and the sample performance semantic variance; Subtracting the semantic enhancement coding feature vector of the local image block of the sample performance from the semantic mean of the sample performance by position, and calculating the fourth power of each position of the feature vector after the subtraction to obtain a sample performance semantic difference modulation vector; Calculating the expected value of the sample performance semantic difference modulation vector to obtain the sample performance semantic expected value; The expected value of the sample performance semantics is divided by the square of the sample performance semantics variance, and the value obtained by the division is input into a sigmoid function to obtain the sample performance semantics static energy factor.

7. The quality control system of foamed asphalt mixture according to claim 6, characterized in that: The sample performance semantic dynamic energy factor calculation unit is used to: Multiplying the static energy factor of the sample performance local image block semantic enhancement coding feature vector and the static energy factor of the sample performance semantic clustering initial center vector by a first weighting parameter to obtain a first sample performance semantic dynamic aggregation energy factor; Multiplying the square of the spatial span between the sample performance local image block semantic enhancement coding feature vector and the sample performance semantic clustering initial center vector by a second weighting parameter to obtain a second sample performance semantic dynamic aggregation energy factor; The first sample performance semantic dynamic aggregation energy factor is divided by the second sample performance semantic dynamic aggregation energy factor to obtain the sample performance semantic dynamic aggregation energy factor.

8. The quality control system of foamed asphalt mixture according to claim 7, characterized in that: The material performance judgment module is used to: input the global sample performance significant aggregation representation vector into a classifier-based quality control module to obtain the quality control result, and the quality control result is used to indicate whether the material performance meets the preset standard.

9. A method for quality control of foamed asphalt mixture, using the quality control system of foamed asphalt mixture according to claim 1, characterized in that: include: Acquire a sample image of the foamed asphalt mixture after compaction, which is collected by a camera; Performing grid segmentation and sample state feature extraction on the sample image of the compacted foamed asphalt mixture to obtain a set of semantic coding feature maps of local image blocks of sample performance; Performing spatial semantic perception feature enhancement on the set of sample performance local image block semantic coding feature maps to obtain a set of sample performance local image block semantic enhancement coding feature maps; Performing global feature aggregation based on energy distribution on a set of semantically enhanced encoded feature maps of local image blocks of the sample performance to obtain a significant aggregation representation of the global sample performance; Based on the significant aggregation representation of the global sample performance, a quality control result is obtained, and the quality control result is used to indicate whether the material performance meets the preset standard.

Images