A road dust detection method and related equipment

By detecting the camera on the vehicle and advanced dust recognition algorithm, multi-dimensional classification and matrix construction of road dust are solved, and the problems of low recognition accuracy and single classification dimensions in the existing technology are solved, and efficient road dust monitoring and management strategies are realized It has been formulated to improve the cleaning and maintenance effect of urban roads.

CN118823720BActive Publication Date: 2024-11-19XINHONGCHANG TECH (SHENZHEN) CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411317478.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-20
Publication Date
2024-11-19
Estimated Expiration
2044-09-20

AI Technical Summary

Technical Problem

The existing road dust detection methods have low recognition accuracy, single classification dimensions, and it is difficult to comprehensively analyze dust types and distribution, which cannot meet the needs of large-scale urban road dust monitoring.

Method used

The detection vehicle is equipped with a camera, and the road image is recognized and classified in multiple dimensions through the preset dust recognition algorithm, the road dust matrix is ​​built, and whether it exceeds the normal range is determined. The management strategy is formulated, using the C3Ghost module, GS bottleneck module, and GS Conv module. , attention mechanism layer, Res2NetBlock module and YOLOv5s module perform dust feature extraction and identification, and feature fusion is performed by combining Transformer module and cascade group attention mechanism.

Benefits of technology

It improves the efficiency and coverage of inspection, provides more comprehensive and accurate road dust data support, optimizes urban road cleaning and maintenance strategies, and reduces the impact of dust on the environment and traffic.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118823720B_ABST
    Figure CN118823720B_ABST
Patent Text Reader

Abstract

The invention relates to a road dust detection method, comprising the following steps: collecting a road image through a camera; performing dust recognition on the road image through a dust recognition algorithm to obtain a road recognition dust image; performing multi-dimensional classification on the road recognition dust image to obtain a classification result of road dust in the road recognition dust image; constructing a road dust matrix based on the dust type, dust concentration level, dust distribution area and main components of dust, and judging whether the road dust exceeds a preset normal range based on the road dust matrix, and if exceeded, marking a vector in the road dust matrix to obtain a marked abnormal vector; formulating a road dust management strategy based on the marked abnormal vector and the road dust matrix, thereby solving the technical problems that the traditional method automatic detection means often has low recognition accuracy, single classification dimension and difficulty in comprehensive analysis of dust type and distribution.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of road dust, and in particular to a road dust detection method and related equipment. Background Art

[0002] With the acceleration of urbanization, the problem of dust pollution on urban roads is becoming increasingly serious. Road dust not only affects the air quality of the city, but may also have an adverse effect on the driving safety of vehicles and the life of the road. Therefore, timely and accurate detection and monitoring of road dust has become an important task of urban management. At present, common road dust detection methods mainly rely on manual inspections and sampling analysis. These methods are not only time-consuming and laborious, but also have low timeliness and coverage of detection, which can hardly meet the needs of large-scale urban road dust monitoring. In addition, existing automated detection methods often have problems such as low recognition accuracy, single classification dimension, and difficulty in comprehensive analysis of dust types and distribution, which cannot provide comprehensive and effective data support for urban road cleaning and maintenance. Summary of the invention

[0003] The main purpose of the present invention is to provide a road dust detection method and related equipment, which solves the technical problems that traditional automated detection methods often have low recognition accuracy, single classification dimension, and difficulty in comprehensive analysis of dust types and distribution.

[0004] To achieve the above object, the present invention provides a road dust detection method, which is applied to a detection vehicle, on which a camera is installed, and comprises the following steps:

[0005] When the detection vehicle is traveling on the road to be detected, an image of the road to be detected is collected by a camera arranged on the detection vehicle to obtain a road image;

[0006] Performing dust recognition on the road image using a preset dust recognition algorithm to obtain a road recognition dust image;

[0007] Performing multi-dimensional classification on the road recognition dust image to obtain a classification result of road dust in the road recognition dust image; wherein the classification result includes road dust type, road dust concentration level, road dust distribution area and main components of road dust;

[0008] A road dust matrix is ​​constructed based on the dust type, dust concentration level, dust distribution area and main components of dust, and whether the road dust exceeds a preset normal range is determined based on the road dust matrix. If exceeded, vectors in the road dust matrix are marked to obtain marked abnormal vectors; wherein the road dust matrix is ​​a square matrix, the rows of the square matrix represent different detection sections, the columns of the square matrix represent road dust detection labels, and the road dust detection labels include dust type labels, concentration level labels, distribution area labels, and main component labels;

[0009] A road dust management strategy is formulated based on the anomaly vector and the road dust matrix.

[0010] Furthermore, the image of the road to be inspected is collected by a camera installed on the inspection vehicle to obtain a road image, including:

[0011] Acquire ambient light of the road to be inspected, and calibrate a camera disposed on the inspection vehicle based on the ambient light to obtain a calibrated camera;

[0012] Under different viewing angles, the road to be detected is imaged based on the calibrated camera to obtain images captured under different viewing angles;

[0013] Seamlessly stitching images collected from different viewing angles to obtain a seamless stitching image;

[0014] The seamless splicing image is standardized to obtain a road image.

[0015] Further, dust recognition is performed on the road image using a preset dust recognition algorithm to obtain a road recognition dust image, including:

[0016] Input the road image into a preset dust recognition algorithm for dust recognition; wherein the dust recognition algorithm includes a C3Ghost module, a GS bottleneck module, a GSConv module and an attention mechanism layer, a Res2NetBlock module and a YOLOv5s module, the C3Ghost module replaces the C3 module in the YOLOv5s module, the attention mechanism layer includes a Transformer module and a cascade group attention mechanism module, the cascade group attention mechanism module is embedded in the encoding layer of the Transformer module, and the Res2NetBlock module replaces the original residual block Bottleneck of YOLOv5s;

[0017] Extracting dust features from the road image using the C3Ghost module to obtain dust features;

[0018] The dust features are compressed and channels are reorganized by the GS bottleneck module to reduce the redundancy of the dust features and obtain optimized features;

[0019] Capturing detail features of the optimized features through the GSConv module to obtain detail capture features;

[0020] Attention is paid to the detail capture feature through the attention mechanism layer to obtain the focused detail capture feature;

[0021] The Res2NetBlock module is used to perform multi-scale feature fusion on the detail capture features to obtain fused features;

[0022] The fused features are identified through the YOLOv5s module to determine the road recognition dust image in the road image.

[0023] Furthermore, the road recognition dust image is subjected to multi-dimensional classification to obtain a classification result of road dust in the road recognition dust image, including:

[0024] Clustering the road recognition dust image to obtain a plurality of initial clusters, and calculating the cluster center of each of the initial clusters to obtain the cluster center of the corresponding initial cluster;

[0025] defining a clustering objective function corresponding to the road recognition dust image, and minimizing the sum of squares of intra-cluster distances through the clustering objective function;

[0026] Calculating the intra-cluster variance of each of the initial clusters and the cluster center of each initial cluster to obtain the variance of each initial cluster;

[0027] According to the variance of each initial clustering cluster, centroid trajectory smoothing processing is performed on the multiple initial clustering clusters and the cluster center of each initial clustering cluster to obtain each target clustering cluster;

[0028] A classification result of road dust in the road recognition dust image is obtained based on each of the target clusters; wherein the classification result includes dust type, dust concentration level, dust distribution area and main components of dust.

[0029] Further, a road dust matrix is ​​constructed based on the dust type, dust concentration level, dust distribution area and main components of dust, and it is determined based on the road dust matrix whether the road dust exceeds a preset normal range. If so, the vectors in the road dust matrix are marked to obtain marked abnormal vectors, including:

[0030] Performing vector transformation on the dust type, dust concentration level, dust distribution area and main components of dust respectively, and obtaining corresponding transformation vectors; wherein the transformation vectors include dust type vector, dust concentration level vector, dust distribution area vector and main component vector of dust;

[0031] constructing a road dust matrix based on the conversion vector;

[0032] By using a preset anomaly recognition algorithm, a probability density distribution calculation is performed on the conversion vectors in the road dust matrix to identify the central tendency and the degree of dispersion of the conversion vectors, and a corresponding probability density distribution value is obtained;

[0033] Comparing the preset normal range with the probability density distribution value, if the probability density distribution value exceeds the preset normal range, marking the probability density distribution value exceeding the preset normal range to obtain the marked probability density distribution value;

[0034] The annotated probability density distribution value is matched with the transformed vector to obtain an abnormal multi-dimensional feature vector in the transformed vector, and the abnormal multi-dimensional feature vector is used as the annotated abnormal vector.

[0035] Furthermore, formulating a road dust management strategy based on the anomaly vector and the road dust matrix includes:

[0036] Performing a spatiotemporal distribution analysis on the anomaly vectors to obtain a road dust accumulation pattern; wherein the road dust accumulation pattern includes an abnormally high incidence area of ​​road dust and an abnormal distribution pattern of road dust within a preset time period; wherein the abnormal distribution pattern includes a peak concentration of road dust within a preset time period and an abnormal frequency within a preset time period;

[0037] Acquire road environment parameters, and perform association rule mining between the abnormal high-incidence area and the road environment parameters through the Apriori algorithm to obtain potential association rules; wherein the potential association rules are the relationship between the abnormal high-incidence area and the road environment parameters;

[0038] Based on the Bayesian network model, the potential association rules are analyzed for causality based on the abnormal distribution law to obtain the causal relationship corresponding to the abnormal high-incidence area;

[0039] By using the analytic hierarchy process, a preliminary management strategy for road dust is developed based on the cause-effect relationship;

[0040] Based on the time series prediction model, the road dust matrix is ​​used to perform real-time monitoring and trend prediction of the road dust to obtain a road dust prediction trend result;

[0041] Dynamically adjusting the preliminary management strategy based on the road dust prediction trend result to obtain a dynamic adjustment plan;

[0042] Implementing management measures on the abnormally high-incidence area through the dynamic adjustment plan to obtain road dust data after the management measures are implemented;

[0043] Based on the road dust data after the implementation of the management measures, the effect of the dynamic adjustment plan is evaluated to obtain an effective evaluation result;

[0044] The effective evaluation results are used to feed back and optimize the dynamic adjustment plan to obtain a road dust management strategy.

[0045] Furthermore, the road environment parameters are obtained, and the association rules between the abnormal high-incidence area and the road environment parameters are mined through the Apriori algorithm to obtain potential association rules, including:

[0046] Through multi-source sensors, data of the road environment corresponding to the road dust is collected to obtain collected data; wherein the collected data includes temperature, humidity, wind speed, traffic flow, and industrial emissions;

[0047] Through the preset remote sensing technology, the road dust is monitored in real time to obtain the spatiotemporal distribution data set of the road dust;

[0048] Based on the abnormal high-incidence area, the road environmental parameters are defined as environmental variables to obtain defined variables; wherein the defined variables include temperature variables, humidity variables, wind speed variables, traffic flow variables, and industrial emissions variables;

[0049] Performing frequent item set mining on the defined variables to obtain frequent item sets that satisfy the minimum support; the frequent item sets are permutations and combinations between any defined variables;

[0050] By using the Apriori algorithm, association rules are mined based on the frequent item sets and the abnormal high-incidence areas to obtain preliminary potential association rules; wherein there are multiple preliminary potential association rules;

[0051] Calculating the confidence and lift of the preliminary potential association rules to obtain a confidence value and a lift value;

[0052] The preliminary potential association rules are screened based on the confidence value and the lift value to obtain potential association rules.

[0053] The present invention also provides a road dust detection device, wherein the road dust detection method is applied to a detection vehicle, and a camera is installed on the detection vehicle, comprising:

[0054] A collection module, used for collecting images of the road to be detected by using a camera arranged on the detection vehicle when the detection vehicle is traveling on the road to be detected, so as to obtain a road image;

[0055] A recognition module, used to perform dust recognition on the road image by using a preset dust recognition algorithm to obtain a road recognition dust image;

[0056] A classification module, used for performing multi-dimensional classification on the road recognition dust image to obtain a classification result of road dust in the road recognition dust image; wherein the classification result includes road dust type, road dust concentration level, road dust distribution area and main components of road dust;

[0057] a labeling module, for constructing a road dust matrix based on the dust type, dust concentration level, dust distribution area and main components of dust, and judging whether the road dust exceeds a preset normal range based on the road dust matrix, and if so, labeling the vectors in the road dust matrix to obtain a labeling abnormal vector; wherein the road dust matrix is ​​a square matrix, the rows of the square matrix represent different detection sections, the columns of the square matrix represent road dust detection labels, and the road dust detection labels include dust type labels, concentration level labels, distribution area labels, and main component labels;

[0058] A strategy module is used to formulate a road dust management strategy based on the anomaly vector and the road dust matrix.

[0059] The present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the above methods when executing the computer program.

[0060] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of any of the above-mentioned methods are implemented.

[0061] The road dust detection method provided by the present invention comprises the following steps: collecting road images through a camera; performing dust recognition on the road images through a dust recognition algorithm to obtain a road recognition dust image; performing multi-dimensional classification on the road recognition dust image to obtain a classification result of road dust in the road recognition dust image; constructing a road dust matrix based on the dust type, dust concentration level, dust distribution area and main components of dust, and judging whether the road dust exceeds a preset normal range based on the road dust matrix, if exceeded, marking the vectors in the road dust matrix to obtain a marked abnormal vector; formulating a road dust management strategy based on the marked abnormal vector and the road dust matrix, and solving the technical problems that the traditional method automatic detection means often have low recognition accuracy, single classification dimension and difficulty in comprehensively analyzing the dust type and distribution through the above technical means; improving the detection efficiency and coverage, and being able to provide more comprehensive and accurate road dust data support for urban management departments, thereby optimizing urban road cleaning and maintenance strategies and reducing the impact of dust on the environment and traffic. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 is a schematic diagram of the steps of a road dust detection method according to an embodiment of the present invention;

[0063] Figure 2 is a structural block diagram of a road dust detection device in one embodiment of the present invention;

[0064] Figure 3 It is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention.

[0065] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0066] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0067] like Figure 1 As shown, Figure 1 is a schematic diagram of the steps of a road dust detection method in one embodiment of the present invention;

[0068] In one embodiment of the present invention, a road dust detection method is provided. The road dust detection method is applied to a detection vehicle, and a camera is installed on the detection vehicle. The method comprises the following steps:

[0069] Step S1, when the detection vehicle is traveling on the road to be detected, an image of the road to be detected is collected by a camera arranged on the detection vehicle to obtain a road image.

[0070] Specifically, in this step, when the inspection vehicle is driving on the road to be inspected, the camera on the inspection vehicle starts working to collect images. Specifically, the camera is a device installed on the inspection vehicle, which can capture the road scene in front of the inspection vehicle in real time and generate corresponding road images. This image is a visual record of the road to be inspected, containing detailed information on the surface of the road section, including but not limited to the distribution of road dust, surface conditions, etc. Through this step, the system can obtain the road images required for subsequent analysis, providing a basis for subsequent dust identification and classification analysis.

[0071] Step S2, performing dust recognition on the road image by using a preset dust recognition algorithm to obtain a road recognition dust image.

[0072] Specifically, in this step, the system uses a pre-set dust recognition algorithm to process and analyze the collected road images. Specifically, the dust recognition algorithm is an algorithm specifically used to detect and identify road dust in images. The algorithm can identify dust-related features in road images and extract these features from the images. Through this processing, the system can generate a new image, which is called a "road recognition dust image." This image not only retains the key information in the road image, but also clearly identifies the location and shape of dust on the road surface, providing a data basis for subsequent multi-dimensional classification.

[0073] Step S3, performing multi-dimensional classification on the road recognition dust image to obtain a classification result of the road dust in the road recognition dust image; wherein the classification result includes the road dust type, the road dust concentration level, the distribution area of ​​the road dust and the main components of the road dust.

[0074] Specifically, first of all, the collected road dust images need to be preprocessed, which includes but is not limited to image denoising, contrast enhancement, correction of uneven lighting, etc., to ensure the accuracy of subsequent analysis. Computer vision technology is used to extract key features from the preprocessed images, such as the shape, size, color, texture, etc. of dust particles, which are crucial for subsequent classification. Multidimensional classification includes: determining the source and type of dust by identifying features, such as industrial dust, soil dust, and particulate matter generated by automobile exhaust. Road dust concentration level: Based on the number, area or volume of dust particles, combined with pre-defined standards, the dust concentration is divided into different levels, such as low, medium, and high. Distribution area of ​​road dust: By analyzing the distribution pattern of dust in the image, it is determined in which areas the dust is mainly concentrated, which helps to understand the pollution source and diffusion path. The main components of road dust: Using chemical analysis methods or combining known dust composition characteristics, the main chemical elements or compounds in dust are identified, which is crucial for understanding the pollution source and potential health risks. Application of classification results: The obtained classification results can be used for a variety of purposes, including but not limited to environmental monitoring, pollution source tracking, policy making, health risk assessment, etc. For example, by understanding the main components and concentration distribution of dust, targeted emission reduction measures or protective measures can be taken to reduce the risk of public exposure to harmful substances. This process not only relies on advanced image processing and machine learning algorithms, but also requires interdisciplinary knowledge, including but not limited to computer science, environmental science, chemistry, public health and other fields to ensure the accuracy and practicality of the analysis.

[0075] Step S4, constructing a road dust matrix based on the dust type, dust concentration level, dust distribution area and main components of dust, and judging whether the road dust exceeds a preset normal range based on the road dust matrix, if exceeded, marking the vectors in the road dust matrix to obtain a marked abnormal vector; wherein the road dust matrix is ​​a square matrix, the rows of the square matrix represent different detection sections, the columns of the square matrix represent road dust detection labels, and the road dust detection labels include dust type labels, concentration level labels, distribution area labels, and main component labels.

[0076] Specifically, the previously classified data on road dust types, concentration levels, distribution areas, and main components are integrated into a two-dimensional table. Each row represents an independent test, and each column corresponds to a specific attribute or label. Matrix construction: Fill the integrated data into the matrix so that the structure of the matrix is ​​clear, easy to understand and analyze. For example, if 10 different road sections are tested, the matrix will have 10 rows; if each label has 5 possible categories, the number of columns in the matrix will be 5. Determine the normal range: Define the "preset normal range" based on established industry standards or health and safety guidelines. Based on historical data, environmental standards, or public health guidelines. Once the normal range is defined, the results of each test can be compared with it. Anomaly annotation: By comparing the data in the matrix with the preset normal range, any test results that are outside the normal range are identified. These results are marked as "abnormal" in the matrix, forming the so-called "annotated anomaly vector". Analysis and action: The anomaly vector provides specific information indicating which road sections may need additional attention or action for dust problems. This can be the basis for further field investigations, targeted cleaning measures, or policy adjustments. This process not only helps us quantify and visualize the problem of road dust, but also provides a data-driven basis for subsequent decision-making and management. It is an important tool for environmental monitoring, pollution control, and public health protection.

[0077] Step S5: formulating a road dust management strategy based on the anomaly vector and the road dust matrix.

[0078] Specifically, by analyzing the anomaly vectors, we can quickly identify those sections where dust concentration exceeds the standard or has special problems. These areas usually have obvious outliers in the matrix and are the focus of attention when formulating strategies. Deepen the cause: For each section marked as anomaly, we need to further analyze the reasons for the high dust concentration. This may involve multiple factors, such as specific industrial emissions, traffic flow, natural environmental conditions, etc. Through the road dust matrix, we can review historical data to understand the trend of dust concentration over time and its correlation with factors such as weather and seasons. Formulate intervention measures: Based on a deep understanding of the problem areas and causes, formulate specific intervention measures. This may include but is not limited to: Strengthen cleaning operations: Increase the frequency of cleaning, especially in areas with serious dust problems. Optimize traffic management: Adjust traffic flow lines and reduce congestion to reduce vehicle emissions. Implement pollution source control: Regulate industrial activities or construction sites that generate a lot of dust and require them to take measures to reduce dust. Greening and vegetation cover: Increase trees and vegetation to absorb dust and pollutants in a natural way. Public education and awareness raising: Through media, community activities and other means, we can raise public awareness of the importance of air quality and encourage environmental protection actions. Implementation and monitoring: After the strategy is formulated, there needs to be a clear implementation plan, including responsibility allocation, resource allocation, timetable, etc. At the same time, we need to establish an effective monitoring mechanism to regularly check the effectiveness of the strategy implementation and adjust the strategy in time to respond to new situations or unexpected challenges. Evaluation and feedback: After the strategy has been implemented for a period of time, it needs to be evaluated to determine its effectiveness. This includes direct measurement of dust concentration, changes in public health indicators, and analysis of economic cost-effectiveness. Based on the evaluation results, the strategy can be adjusted and optimized as necessary. Through this series of steps, we can effectively transform data into practical actions, propose solutions to specific road dust problems, and ultimately achieve the goal of improving air quality and public health.

[0079] In a specific embodiment, the image of the road to be inspected is collected by a camera arranged on the inspection vehicle to obtain a road image, including:

[0080] Acquire ambient light of the road to be inspected, and calibrate a camera disposed on the inspection vehicle based on the ambient light to obtain a calibrated camera;

[0081] Under different viewing angles, the road to be detected is imaged based on the calibrated camera to obtain images captured under different viewing angles;

[0082] Seamlessly stitching images collected from different viewing angles to obtain a seamless stitching image;

[0083] The seamless splicing image is standardized to obtain a road image.

[0084] Specifically, obtain ambient light: First, the system will detect and record the current ambient light conditions, including parameters such as light intensity and color temperature, in order to ensure the quality of subsequent image acquisition. Calibrate the camera: Based on the acquired ambient light information, the system will calibrate the camera on the inspection vehicle. The purpose of calibration is to adjust the parameters of the camera so that it can provide the best image quality under specific lighting conditions. This step ensures the stability and consistency of the camera under different lighting conditions. Image acquisition: The calibrated camera acquires images of the road to be inspected at multiple different perspectives. By changing the position and angle of the inspection vehicle, the system can capture all-round image information of the road, thereby providing a variety of perspectives for subsequent analysis. Seamlessly stitch images: The images collected from different perspectives are stitched together to generate a seamless panoramic image. The stitching process needs to solve the boundary fusion problem between images to ensure that the stitched image has no obvious seams and provides a continuous and complete overview. Standardization: Finally, the seamlessly stitched image is standardized. This may include adjusting the brightness, contrast, color saturation, etc. of the image to unify the image style and improve the consistency and readability of the visual effect. Standardization also helps with subsequent image analysis and processing to ensure the accuracy and reliability of the data. Through the above steps, high-quality images that fully cover the road to be inspected can be obtained, providing basic data support for subsequent tasks such as dust identification and road trajectory prediction.

[0085] In a specific embodiment, dust recognition is performed on the road image by a preset dust recognition algorithm to obtain a road recognition dust image, including:

[0086] Input the road image into a preset dust recognition algorithm for dust recognition; wherein the dust recognition algorithm includes a C3Ghost module, a GS bottleneck module, a GSConv module and an attention mechanism layer, a Res2NetBlock module and a YOLOv5s module, the C3Ghost module replaces the C3 module in the YOLOv5s module, the attention mechanism layer includes a Transformer module and a cascade group attention mechanism module, the cascade group attention mechanism module is embedded in the encoding layer of the Transformer module, and the Res2NetBlock module replaces the original residual block Bottleneck of YOLOv5s;

[0087] Extracting dust features from the road image using the C3Ghost module to obtain dust features;

[0088] The dust features are compressed and channels are reorganized by the GS bottleneck module to reduce the redundancy of the dust features and obtain optimized features;

[0089] Capturing detail features of the optimized features through the GSConv module to obtain detail capture features;

[0090] Attention is paid to the detail capture feature through the attention mechanism layer to obtain the focused detail capture feature;

[0091] The Res2NetBlock module is used to perform multi-scale feature fusion on the detail capture features to obtain fused features;

[0092] The fused features are identified through the YOLOv5s module to determine the road recognition dust image in the road image.

[0093] Specifically, image input and dust recognition algorithm activation: First, the road image is used as input data and input into the preset dust recognition algorithm. The algorithm integrates multiple core modules to efficiently and accurately identify dust features in the image. C3Ghost module application: As part of the algorithm, the C3Ghost module replaces the C3 module in the traditional YOLOv5s module. The C3Ghost module improves the efficiency and accuracy of feature extraction through improved structural design, laying a solid foundation for subsequent dust recognition. Dust feature extraction: Through the C3Ghost module, the algorithm can deeply explore the dust features in the road image and generate a preliminary dust feature set. Feature compression and channel reorganization: Next, the GS bottleneck module intervenes to perform deep feature compression and channel reorganization on the dust features extracted by the C3Ghost module. This process aims to remove redundant information, retain the most critical dust features, and improve the accuracy and efficiency of recognition. Detail feature capture: The GSConv module then comes into play, focusing on capturing the subtleties of the compressed and reorganized features to ensure that the algorithm can recognize various detailed features of dust, such as size, shape, position, etc. Attention mechanism enhancement: The introduced attention mechanism layer, including the Transformer module and the cascade group attention mechanism module, intelligently pays attention to detail capture features. The Transformer module processes sequence data in an efficient way, while the cascade group attention mechanism further enhances the recognition ability of local features, ensuring that the algorithm can focus on the key areas that are most likely to contain dust. Multi-scale feature fusion: The addition of the Res2NetBlock module realizes the multi-scale feature fusion of detail capture features enhanced by the attention mechanism. This process integrates feature information at different levels, improves the generalization ability and recognition accuracy of the model, and ensures comprehensive and accurate recognition of dust. Recognition and positioning: Finally, the fused features are finally recognized and positioned through the YOLOv5s module to determine the road recognition dust image in the road image. With its efficient detection performance, the YOLOv5s module ensures a good balance between speed and accuracy of the algorithm, making the entire recognition process fast and accurate. Through this series of complex and precise steps, the preset dust recognition algorithm successfully recognizes dust in the road image, providing an accurate and reliable basis for subsequent analysis and processing.

[0094] In a specific embodiment, the road recognition dust image is subjected to multi-dimensional classification to obtain a classification result of road dust in the road recognition dust image, including:

[0095] Clustering the road recognition dust image to obtain a plurality of initial clusters, and calculating the cluster center of each of the initial clusters to obtain the cluster center of the corresponding initial cluster;

[0096] defining a clustering objective function corresponding to the road recognition dust image, and minimizing the sum of squares of intra-cluster distances through the clustering objective function;

[0097] Calculating the intra-cluster variance of each of the initial clusters and the cluster center of each initial cluster to obtain the variance of each initial cluster;

[0098] According to the variance of each initial clustering cluster, centroid trajectory smoothing processing is performed on the multiple initial clustering clusters and the cluster center of each initial clustering cluster to obtain each target clustering cluster;

[0099] A classification result of road dust in the road recognition dust image is obtained based on each of the target clusters; wherein the classification result includes dust type, dust concentration level, dust distribution area and main components of dust.

[0100] Specifically, clustering division and calculation of initial clustering clusters: First, we cluster the road recognition dust image, which divides the image into multiple initial clustering clusters. Each cluster represents an area in the image where similar dust features may exist. Next, by calculating the center point of each cluster (i.e., cluster center), we define a representative location for each cluster for the subsequent classification process. Definition and optimization of clustering objective function: In order to maximize the similarity within the cluster, a clustering objective function is defined, which aims to minimize the sum of squared distances between elements within the cluster. By optimizing this function, we can more accurately classify similar dust features into the same cluster, thereby improving the accuracy of classification. Calculation of intra-cluster variance: Next, we calculate the variance of each initial clustering cluster, which helps us understand the degree of heterogeneity within the cluster. The smaller the variance, the more concentrated the dust features within the cluster, and vice versa, it means that the features are dispersed. Centroid trajectory smoothing: Based on the variance information of each cluster, we perform centroid trajectory smoothing on the initial clustering cluster and its cluster center. This process aims to reduce the noise within the cluster and improve the stability of the cluster, so as to obtain a clearer and more accurate target clustering cluster. Obtaining classification results: Finally, by analyzing each target cluster, we can obtain the classification results of road dust in the road dust recognition image. These results usually include the type, concentration level, distribution area and main components of dust, providing us with a comprehensive insight into the characteristics of dust. Through this series of steps, we can not only accurately identify and classify road dust, but also deeply understand the characteristics of dust, providing a scientific basis for subsequent environmental protection, urban planning, etc.

[0101] In a specific embodiment, a road dust matrix is ​​constructed based on the dust type, dust concentration level, dust distribution area and main components of dust, and it is determined based on the road dust matrix whether the road dust exceeds a preset normal range. If so, the vectors in the road dust matrix are marked to obtain the marked abnormal vectors, including:

[0102] Performing vector transformation on the dust type, dust concentration level, dust distribution area and main components of dust respectively, and obtaining corresponding transformation vectors; wherein the transformation vectors include dust type vector, dust concentration level vector, dust distribution area vector and main component vector of dust;

[0103] constructing a road dust matrix based on the conversion vector;

[0104] By using a preset anomaly recognition algorithm, a probability density distribution calculation is performed on the conversion vectors in the road dust matrix to identify the central tendency and the degree of dispersion of the conversion vectors, and a corresponding probability density distribution value is obtained;

[0105] Comparing the preset normal range with the probability density distribution value, if the probability density distribution value exceeds the preset normal range, marking the probability density distribution value exceeding the preset normal range to obtain the marked probability density distribution value;

[0106] The annotated probability density distribution value is matched with the transformed vector to obtain an abnormal multi-dimensional feature vector in the transformed vector, and the abnormal multi-dimensional feature vector is used as the annotated abnormal vector.

[0107] Specifically, the road dust matrix is ​​constructed: First, the dust type, concentration level, distribution area and main component information extracted from the data are converted into a form that can be processed by the mathematical model through the method of vector transformation. The purpose of this is to simplify complex multi-dimensional data into a vector form that is easy for computers to understand and analyze. The obtained vectors include dust type vectors, dust concentration level vectors, distribution area vectors and main component vectors. Matrix construction based on transformation vectors: These transformed vectors are organized into a matrix, namely the road dust matrix. The matrix form allows us to view and compare dust data of different attributes in a structured way, making it easier to find patterns and anomalies. Probability density distribution calculation: Next, the probability density distribution of the transformation vectors in the matrix is ​​calculated using the preset anomaly recognition algorithm. This step helps us understand the distribution of the data by analyzing the central tendency and dispersion of the transformation vectors. The probability density distribution value reflects the central tendency and distribution range of the data and is an important indicator for identifying anomalies. Anomaly recognition and annotation: We set a normal range as a benchmark, and any probability density distribution value outside this range is considered anomaly. By comparing the preset normal range with the calculated probability density distribution value, we can identify which data points belong to abnormal situations. Annotate anomaly vectors: Finally, all probability density distribution values ​​that exceed the preset normal range are annotated to form annotated anomaly vectors. These vectors represent data points in the road dust matrix that show abnormal behaviors or characteristics. They are caused by specific environmental factors, pollution sources, or management issues. Through such steps, we can not only quantitatively identify the anomalies of road dust problems, but also further analyze and understand the specific factors that cause these anomalies, providing a scientific basis for subsequent environmental pollution control and road maintenance strategies.

[0108] In a specific embodiment, formulating a road dust management strategy based on the anomaly vector and the road dust matrix includes:

[0109] Performing a spatiotemporal distribution analysis on the anomaly vectors to obtain a road dust accumulation pattern; wherein the road dust accumulation pattern includes an abnormally high incidence area of ​​road dust and an abnormal distribution pattern of road dust within a preset time period; wherein the abnormal distribution pattern includes a peak concentration of road dust within a preset time period and an abnormal frequency within a preset time period;

[0110] Acquire road environment parameters, and perform association rule mining between the abnormal high-incidence area and the road environment parameters through the Apriori algorithm to obtain potential association rules; wherein the potential association rules are the relationship between the abnormal high-incidence area and the road environment parameters;

[0111] Based on the Bayesian network model, the potential association rules are analyzed for causality based on the abnormal distribution law to obtain the causal relationship corresponding to the abnormal high-incidence area;

[0112] By using the analytic hierarchy process, a preliminary management strategy for road dust is developed based on the cause-effect relationship;

[0113] Based on the time series prediction model, the road dust matrix is ​​used to perform real-time monitoring and trend prediction of the road dust to obtain a road dust prediction trend result;

[0114] Dynamically adjusting the preliminary management strategy based on the road dust prediction trend result to obtain a dynamic adjustment plan;

[0115] Implementing management measures on the abnormally high-incidence area through the dynamic adjustment plan to obtain road dust data after the management measures are implemented;

[0116] Based on the road dust data after the implementation of the management measures, the effect of the dynamic adjustment plan is evaluated to obtain an effective evaluation result;

[0117] The effective evaluation results are used to feed back and optimize the dynamic adjustment plan to obtain a road dust management strategy.

[0118] Specifically, first, the spatiotemporal distribution of the anomaly vectors is analyzed to identify the road areas with high abnormality and the abnormal distribution patterns within a specific time. These analysis results include the identification of abnormal high-incidence areas and the peak concentration and abnormal frequency within a preset time period, which provide basic data for our subsequent decision-making. Association rule mining: The Apriori algorithm is used to mine the relationship between abnormal high-incidence areas and road environmental parameters. The Apriori algorithm is a commonly used association rule learning algorithm that can discover potential associations between different parameters and help us understand which environmental conditions may cause dust problems. Causal relationship analysis: Based on the Bayesian network model, the potential association rules under the abnormal distribution pattern are analyzed for causal relationships. The Bayesian network model allows us to build a probabilistic graphical model to depict the dependencies between different variables, and then infer the possible reasons behind the abnormal high-incidence areas. Preliminary management strategy formulation: Through the hierarchical analysis method, combined with the results of the causal relationship analysis, a preliminary management strategy is formulated. The hierarchical analysis method is a multi-criteria decision-making method that can help us systematically evaluate the pros and cons of different management measures and select the most effective strategy. Real-time monitoring and trend prediction: Use time series prediction models to monitor road dust in real time and predict trends to obtain predicted trends for future dust development. This step is crucial for dynamically adjusting management strategies because it provides an expectation of future conditions, allowing measures to be taken in advance. Dynamic adjustment plan generation: Based on the predicted trend results of road dust, the preliminary management strategy is dynamically adjusted to form a set of management plans that adapt to current and future changes. The dynamic adjustment plan takes into account current abnormal conditions as well as future trends, aiming to achieve more effective dust management. Implementation effect evaluation: The effect of the dynamic adjustment plan is evaluated through the road dust data collected after the implementation of management measures. This evaluation is a key part of the circular improvement strategy, which helps us verify the effectiveness of the plan and make necessary adjustments accordingly. Feedback and optimization: Based on the evaluation results, the dynamic adjustment plan is fed back and optimized to obtain the final road dust management strategy. This process is continuously iterated, aiming to gradually improve management efficiency and effectiveness, and ultimately achieve a long-term solution to the road dust problem. Through the above steps, we can not only formulate targeted management strategies for specific abnormal situations, but also continuously optimize according to environmental changes and data feedback to achieve refined management and efficient response to road dust problems.

[0119] In a specific embodiment, the road environment parameters are obtained, and the association rules between the abnormal high-incidence area and the road environment parameters are mined through the Apriori algorithm to obtain potential association rules, including:

[0120] Through multi-source sensors, data of the road environment corresponding to the road dust is collected to obtain collected data; wherein the collected data includes temperature, humidity, wind speed, traffic flow, and industrial emissions;

[0121] Through the preset remote sensing technology, the road dust is monitored in real time to obtain the spatiotemporal distribution data set of the road dust;

[0122] Based on the abnormal high-incidence area, the road environmental parameters are defined as environmental variables to obtain defined variables; wherein the defined variables include temperature variables, humidity variables, wind speed variables, traffic flow variables, and industrial emissions variables;

[0123] Performing frequent item set mining on the defined variables to obtain frequent item sets that satisfy the minimum support; the frequent item sets are permutations and combinations between any defined variables;

[0124] By using the Apriori algorithm, association rules are mined based on the frequent item sets and the abnormal high-incidence areas to obtain preliminary potential association rules; wherein there are multiple preliminary potential association rules;

[0125] Calculating the confidence and lift of the preliminary potential association rules to obtain a confidence value and a lift value;

[0126] The preliminary potential association rules are screened based on the confidence value and the lift value to obtain potential association rules.

[0127] Specifically, first, we collect data through multi-source sensors installed around the road, including but not limited to environmental parameters such as temperature, humidity, wind speed, traffic flow, and industrial emissions. The data from these sensors provide basic information for subsequent analysis. Real-time monitoring: At the same time, we use the preset remote sensing technology to monitor road dust in real time. The collected spatiotemporal distribution data set contains the changes in dust in different time and space, which helps us identify areas with abnormally high incidence. Definition of environmental variables: According to the characteristics of abnormally high incidence areas, we define the collected environmental parameters as different variables, such as temperature variables, humidity variables, wind speed variables, traffic flow variables, and industrial emissions variables. These definitions help us describe the environmental state more accurately. Frequent item set mining: Next, we perform frequent item set mining on the defined environmental variables. Frequent item sets refer to combinations that appear frequently in the data set and meet the minimum support condition. For example, frequently occurring environmental combinations such as "high temperature + high humidity" or "low wind speed + high traffic flow" will be found. Association rule mining: Association rule mining is performed based on frequent item sets and abnormally high incidence areas through the Apriori algorithm. Apriori algorithm is a commonly used data mining algorithm, which is used to find association rules between frequent item sets. In this process, we will get multiple preliminary potential association rules, such as "high temperature + high traffic flow" leads to "increased dust concentration". Rule evaluation: Calculate the confidence and lift of the association rules mined initially. Confidence measures the reliability of the rule, and lift measures the enhancement effect of the rule relative to the background probability. Through these two indicators, we can evaluate the practicality and effectiveness of the rules. Rule screening: Finally, based on the evaluation results of confidence and lift, screen out those association rules that are both reliable and practical as potential association rules. These rules can help us understand the reasons for the abnormally high incidence of road dust, thereby providing a scientific basis for subsequent management strategies. Through the above steps, we can systematically explore and reveal the complex relationship between road environmental parameters and abnormally high incidence areas, and provide data support for the effective management of road dust problems.

[0128] The road dust detection method in the embodiment of the present invention is described above. The road dust detection device in the embodiment of the present invention is described below. Figure 2 , an embodiment of the road dust detection device in the embodiment of the present invention includes:

[0129] The acquisition module 21 is used to acquire an image of the road to be detected by using a camera arranged on the detection vehicle when the detection vehicle is traveling on the road to be detected, so as to obtain a road image;

[0130] The recognition module 22 is used to perform dust recognition on the road image by using a preset dust recognition algorithm to obtain a road recognition dust image;

[0131] A classification module 23 is used to perform multi-dimensional classification on the road recognition dust image to obtain a classification result of road dust in the road recognition dust image; wherein the classification result includes road dust type, road dust concentration level, road dust distribution area and main components of road dust;

[0132] The labeling module 24 is used to construct a road dust matrix based on the dust type, dust concentration level, dust distribution area and main components of dust, and determine whether the road dust exceeds a preset normal range based on the road dust matrix. If it exceeds, the vector in the road dust matrix is ​​labeled to obtain a labeled abnormal vector; wherein the road dust matrix is ​​a square matrix, the rows of the square matrix represent different detection sections, the columns of the square matrix represent road dust detection labels, and the road dust detection labels include dust type labels, concentration level labels, distribution area labels, and main component labels;

[0133] The strategy module 25 is used to formulate a road dust management strategy based on the anomaly vector and the road dust matrix.

[0134] In this embodiment, for the specific implementation of each unit in the above system embodiment, please refer to the above method embodiment, which will not be repeated here.

[0135] Reference Figure 3 The present invention also provides a computer device in an embodiment, wherein the internal structure of the computer device can be as follows: Figure 3 As shown. The computer device includes a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. Among them, the processor designed by the computer is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above method is implemented.

[0136] Those skilled in the art will understand that Figure 3 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0137] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0138] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided by the present invention and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM.

[0139] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0140] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A road dust detection method, characterized in that: The road dust detection method is applied to a detection vehicle, on which a camera is installed, and the method comprises the following steps: When the detection vehicle is traveling on the road to be detected, an image of the road to be detected is collected by a camera arranged on the detection vehicle to obtain a road image; Performing dust recognition on the road image using a preset dust recognition algorithm to obtain a road recognition dust image; Performing multi-dimensional classification on the road recognition dust image to obtain a classification result of road dust in the road recognition dust image; wherein the classification result includes road dust type, road dust concentration level, road dust distribution area and main components of road dust; A road dust matrix is ​​constructed based on the dust type, dust concentration level, dust distribution area and main components of dust, and whether the road dust exceeds a preset normal range is determined based on the road dust matrix. If exceeded, vectors in the road dust matrix are marked to obtain marked abnormal vectors; wherein the road dust matrix is ​​a square matrix, the rows of the square matrix represent different detection sections, the columns of the square matrix represent road dust detection labels, and the road dust detection labels include dust type labels, concentration level labels, distribution area labels, and main component labels; formulating a road dust management strategy based on the anomaly vector and the road dust matrix; Performing dust recognition on the road image by using a preset dust recognition algorithm to obtain a road recognition dust image includes: Input the road image into a preset dust recognition algorithm for dust recognition; wherein the dust recognition algorithm includes a C3Ghost module, a GS bottleneck module, a GSConv module and an attention mechanism layer, a Res2NetBlock module and a YOLOv5s module, the C3Ghost module replaces the C3 module in the YOLOv5s module, the attention mechanism layer includes a Transformer module and a cascade group attention mechanism module, the cascade group attention mechanism module is embedded in the encoding layer of the Transformer module, and the Res2NetBlock module replaces the original residual block Bottleneck of YOLOv5s; Extracting dust features from the road image using the C3Ghost module to obtain dust features; The dust features are compressed and channels are reorganized by the GS bottleneck module to reduce the redundancy of the dust features and obtain optimized features; Capturing detail features of the optimized features through the GSConv module to obtain detail capture features; Attention is paid to the detail capture feature through the attention mechanism layer to obtain the focused detail capture feature; The Res2NetBlock module is used to perform multi-scale feature fusion on the detail capture features to obtain fused features; Identify the fused features through the YOLOv5s module to determine a road recognition dust image in the road image; A road dust matrix is ​​constructed based on the dust type, dust concentration level, dust distribution area and main components of dust, and it is determined based on the road dust matrix whether the road dust exceeds a preset normal range. If so, the vectors in the road dust matrix are marked to obtain a marked abnormal vector, including: Performing vector transformation on the dust type, dust concentration level, dust distribution area and main components of dust respectively, and obtaining corresponding transformation vectors; wherein the transformation vectors include dust type vector, dust concentration level vector, dust distribution area vector and main component vector of dust; constructing a road dust matrix based on the conversion vector; By using a preset anomaly recognition algorithm, a probability density distribution calculation is performed on the conversion vectors in the road dust matrix to identify the central tendency and the degree of dispersion of the conversion vectors, and a corresponding probability density distribution value is obtained; Comparing the preset normal range with the probability density distribution value, if the probability density distribution value exceeds the preset normal range, marking the probability density distribution value exceeding the preset normal range to obtain the marked probability density distribution value; Matching the annotated probability density distribution value with the transformed vector to obtain an abnormal multi-dimensional feature vector in the transformed vector, and using the abnormal multi-dimensional feature vector as an annotated abnormal vector; Formulating a road dust management strategy based on the anomaly vector and the road dust matrix includes: Performing a spatiotemporal distribution analysis on the anomaly vectors to obtain a road dust accumulation pattern; wherein the road dust accumulation pattern includes an abnormally high incidence area of ​​road dust and an abnormal distribution pattern of road dust within a preset time period; wherein the abnormal distribution pattern includes a peak concentration of road dust within a preset time period and an abnormal frequency within a preset time period; Acquire road environment parameters, and perform association rule mining between the abnormal high-incidence area and the road environment parameters through the Apriori algorithm to obtain potential association rules; wherein the potential association rules are the relationship between the abnormal high-incidence area and the road environment parameters; Based on the Bayesian network model, the potential association rules are analyzed for causality based on the abnormal distribution law to obtain the causal relationship corresponding to the abnormal high-incidence area; By using the analytic hierarchy process, a preliminary management strategy for road dust is developed based on the cause-effect relationship; Based on the time series prediction model, the road dust matrix is ​​used to perform real-time monitoring and trend prediction of the road dust to obtain a road dust prediction trend result; Dynamically adjusting the preliminary management strategy based on the road dust prediction trend result to obtain a dynamic adjustment plan; Implementing management measures on the abnormally high-incidence area through the dynamic adjustment plan to obtain road dust data after the management measures are implemented; Based on the road dust data after the implementation of the management measures, the effect of the dynamic adjustment plan is evaluated to obtain an effective evaluation result; The effective evaluation results are used to feed back and optimize the dynamic adjustment plan to obtain a road dust management strategy.

2. The road dust detection method according to claim 1, characterized in that: The image of the road to be inspected is collected by a camera arranged on the inspection vehicle to obtain a road image, including: Acquire ambient light of the road to be inspected, and calibrate a camera disposed on the inspection vehicle based on the ambient light to obtain a calibrated camera; Under different viewing angles, the road to be detected is imaged based on the calibrated camera to obtain images captured under different viewing angles; Seamlessly stitching images collected from different viewing angles to obtain a seamless stitching image; The seamless splicing image is standardized to obtain a road image.

3. The road dust detection method according to claim 1, characterized in that: The road recognition dust image is subjected to multi-dimensional classification to obtain a classification result of road dust in the road recognition dust image, including: Clustering the road recognition dust image to obtain a plurality of initial clusters, and calculating the cluster center of each of the initial clusters to obtain the cluster center of the corresponding initial cluster; defining a clustering objective function corresponding to the road recognition dust image, and minimizing the sum of squares of intra-cluster distances through the clustering objective function; Calculating the intra-cluster variance of each of the initial clusters and the cluster center of each initial cluster to obtain the variance of each initial cluster; According to the variance of each initial clustering cluster, centroid trajectory smoothing processing is performed on the multiple initial clustering clusters and the cluster center of each initial clustering cluster to obtain each target clustering cluster; A classification result of road dust in the road recognition dust image is obtained based on each of the target clusters; wherein the classification result includes dust type, dust concentration level, dust distribution area and main components of dust.

4. The road dust detection method according to claim 1, characterized in that: Obtain road environment parameters, and use the Apriori algorithm to perform association rule mining on the abnormal high-incidence area and the road environment parameters to obtain potential association rules, including: Through multi-source sensors, data of the road environment corresponding to the road dust is collected to obtain collected data; wherein the collected data includes temperature, humidity, wind speed, traffic flow, and industrial emissions; Through the preset remote sensing technology, the road dust is monitored in real time to obtain the spatiotemporal distribution data set of the road dust; Based on the abnormal high-incidence area, the road environmental parameters are defined as environmental variables to obtain defined variables; wherein the defined variables include temperature variables, humidity variables, wind speed variables, traffic flow variables, and industrial emissions variables; Performing frequent item set mining on the defined variables to obtain frequent item sets that satisfy the minimum support; the frequent item sets are permutations and combinations between any defined variables; By using the Apriori algorithm, association rules are mined based on the frequent item sets and the abnormal high-incidence areas to obtain preliminary potential association rules; wherein there are multiple preliminary potential association rules; Calculating the confidence and lift of the preliminary potential association rules to obtain a confidence value and a lift value; The preliminary potential association rules are screened based on the confidence value and the lift value to obtain potential association rules.

5. A road dust detection device, characterized in that: Used to implement a road dust detection method as claimed in claim 1, the road dust detection device is applied to a detection vehicle, and a camera is installed on the detection vehicle, comprising: A collection module, used for collecting images of the road to be detected by using a camera arranged on the detection vehicle when the detection vehicle is traveling on the road to be detected, so as to obtain a road image; A recognition module, used to perform dust recognition on the road image by using a preset dust recognition algorithm to obtain a road recognition dust image; A classification module, used for performing multi-dimensional classification on the road recognition dust image to obtain a classification result of road dust in the road recognition dust image; wherein the classification result includes road dust type, road dust concentration level, road dust distribution area and main components of road dust; a labeling module, for constructing a road dust matrix based on the dust type, dust concentration level, dust distribution area and main components of dust, and judging whether the road dust exceeds a preset normal range based on the road dust matrix, and if so, labeling the vectors in the road dust matrix to obtain a labeling abnormal vector; wherein the road dust matrix is ​​a square matrix, the rows of the square matrix represent different detection sections, the columns of the square matrix represent road dust detection labels, and the road dust detection labels include dust type labels, concentration level labels, distribution area labels, and main component labels; A strategy module, configured to formulate a road dust management strategy based on the anomaly vector and the road dust matrix; Performing dust recognition on the road image by using a preset dust recognition algorithm to obtain a road recognition dust image includes: Input the road image into a preset dust recognition algorithm for dust recognition; wherein the dust recognition algorithm includes a C3Ghost module, a GS bottleneck module, a GSConv module and an attention mechanism layer, a Res2NetBlock module and a YOLOv5s module, the C3Ghost module replaces the C3 module in the YOLOv5s module, the attention mechanism layer includes a Transformer module and a cascade group attention mechanism module, the cascade group attention mechanism module is embedded in the encoding layer of the Transformer module, and the Res2NetBlock module replaces the original residual block Bottleneck of YOLOv5s; Extracting dust features from the road image using the C3Ghost module to obtain dust features; The dust features are compressed and channels are reorganized by the GS bottleneck module to reduce the redundancy of the dust features and obtain optimized features; Capturing detail features of the optimized features through the GSConv module to obtain detail capture features; Attention is paid to the detail capture feature through the attention mechanism layer to obtain the focused detail capture feature; The Res2NetBlock module is used to perform multi-scale feature fusion on the detail capture features to obtain fused features; Identify the fused features through the YOLOv5s module to determine a road recognition dust image in the road image; A road dust matrix is ​​constructed based on the dust type, dust concentration level, dust distribution area and main components of dust, and it is determined based on the road dust matrix whether the road dust exceeds a preset normal range. If so, the vectors in the road dust matrix are marked to obtain a marked abnormal vector, including: Performing vector transformation on the dust type, dust concentration level, dust distribution area and main components of dust respectively, and obtaining corresponding transformation vectors; wherein the transformation vectors include dust type vector, dust concentration level vector, dust distribution area vector and main component vector of dust; constructing a road dust matrix based on the conversion vector; By using a preset anomaly recognition algorithm, a probability density distribution calculation is performed on the conversion vectors in the road dust matrix to identify the central tendency and the degree of dispersion of the conversion vectors, and a corresponding probability density distribution value is obtained; Comparing the preset normal range with the probability density distribution value, if the probability density distribution value exceeds the preset normal range, marking the probability density distribution value exceeding the preset normal range to obtain the marked probability density distribution value; Matching the annotated probability density distribution value with the transformed vector to obtain an abnormal multi-dimensional feature vector in the transformed vector, and using the abnormal multi-dimensional feature vector as an annotated abnormal vector; Formulating a road dust management strategy based on the anomaly vector and the road dust matrix includes: Performing a spatiotemporal distribution analysis on the anomaly vectors to obtain a road dust accumulation pattern; wherein the road dust accumulation pattern includes an abnormally high incidence area of ​​road dust and an abnormal distribution pattern of road dust within a preset time period; wherein the abnormal distribution pattern includes a peak concentration of road dust within a preset time period and an abnormal frequency within a preset time period; Acquire road environment parameters, and perform association rule mining between the abnormal high-incidence area and the road environment parameters through the Apriori algorithm to obtain potential association rules; wherein the potential association rules are the relationship between the abnormal high-incidence area and the road environment parameters; Based on the Bayesian network model, the potential association rules are analyzed for causality based on the abnormal distribution law to obtain the causal relationship corresponding to the abnormal high-incidence area; By using the analytic hierarchy process, a preliminary management strategy for road dust is developed based on the cause-effect relationship; Based on the time series prediction model, the road dust matrix is ​​used to perform real-time monitoring and trend prediction of the road dust to obtain a road dust prediction trend result; Dynamically adjusting the preliminary management strategy based on the road dust prediction trend result to obtain a dynamic adjustment plan; Implementing management measures on the abnormally high-incidence area through the dynamic adjustment plan to obtain road dust data after the management measures are implemented; Based on the road dust data after the implementation of the management measures, the effect of the dynamic adjustment plan is evaluated to obtain an effective evaluation result; The effective evaluation results are used to feed back and optimize the dynamic adjustment plan to obtain a road dust management strategy.

6. A computer device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 4 are implemented.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 4 are implemented.