Road infrastructure evaluation processing method and system

Abstract

The invention discloses a highway infrastructure evaluation processing method and system, and relates to the technical field of highway management. Multi-modal data are collected through satellite remote sensing, unmanned aerial vehicle aerial photography and a sensor, and data preprocessing and feature extraction are carried out. Fusing the spatial features and the detection features by adopting a hierarchical fusion method, carrying out secondary fusion with GIS data, and establishing association between the spatial features and attribute information; constructing an evaluation prediction model based on a support vector machine, a random forest, a gradient boosting decision tree and logistic regression; and outputting a road infrastructure evaluation grade according to the evaluation index and the scoring standard in combination with the association fusion result and the prediction model. According to the method, factors such as road conditions and traffic flow are comprehensively considered, and a scientific basis is provided for road infrastructure assessment.

Description

Technical Field

[0001] The present invention relates to the technical field of highway management, and in particular to a highway infrastructure assessment and processing method and system. Background Art

[0002] With the rapid development of my country's economy and the advancement of urbanization, remarkable achievements have been made in the construction of highway infrastructure. As an important national infrastructure, the safe and efficient operation of highways is of great significance to economic development and social stability. my country's highway network is relatively large, and there are a large number of various traffic safety facilities. Due to the rapid update speed of information on traffic safety facilities and the unreasonable management system, the quality and level of vehicle operation on the road have a large-scale adverse impact. In order to better manage and maintain highway infrastructure and realize the refined management of its entire life cycle, digital processing technology has been widely used.

[0003] In order to manage and maintain highway infrastructure in a timely and effective manner, it is necessary to be able to obtain the specific situation of highway infrastructure at any time. At present, the data of highway infrastructure is obtained from various channels, and after screening and classification, the current status of highway infrastructure is evaluated, and maintenance decisions are provided after the evaluation results are determined. However, in the data processing and evaluation methods, screening, classification and evaluation still require manual participation, which is not only time-consuming and labor-intensive, but also difficult to fully tap the full potential of the data, limiting the accuracy of the evaluation results and the effectiveness of decision support, making it difficult to meet the needs of rapid response, resulting in untimely provision of maintenance decision support, and making it difficult to achieve accurate maintenance and optimization of highway facilities. Summary of the invention

[0004] In order to overcome the shortcomings of the data processing and evaluation methods in the above-mentioned prior art that screening, classification and evaluation still require manual participation, which is not only time-consuming and labor-intensive, but also difficult to fully tap the full potential of the data, limiting the accuracy of the evaluation results and the effectiveness of decision support, making it difficult to meet the needs of rapid response, and making it difficult to provide maintenance decision support in a timely manner, and difficult to achieve accurate maintenance and optimization of highway facilities, the main purpose of the present invention is to provide a highway infrastructure evaluation and processing method, system and device.

[0005] To achieve the above object, the present invention adopts the following technical solution, a highway infrastructure assessment and processing method, comprising the following steps:

[0006] Acquire multimodal highway infrastructure data, including spatial data, inspection data, and GIS data;

[0007] After preprocessing the spatial data and the detection data and extracting features, spatial features and detection features are obtained;

[0008] Use hierarchical fusion to fuse spatial features and detection features to obtain fusion features;

[0009] After preprocessing, GIS data is spatially registered with the fusion feature, the feature association between spatial features and attribute information is established, and weighted splicing is performed to obtain the association fusion result;

[0010] Based on the association fusion results, support vector machine is selected as the basic model, random forest and gradient boosting decision tree are used as the first layer model, logistic regression is grouped as the second layer model, evaluation indicators are set, and an evaluation prediction model is constructed;

[0011] The associated fusion results of the highway infrastructure to be evaluated are obtained and input into the evaluation prediction model to obtain the probability of the evaluation indicators. Combined with the weighted scoring and grading rules set for the evaluation indicators, the highway infrastructure evaluation results are obtained.

[0012] Furthermore, hierarchical fusion is used to fuse the spatial features and the detection features to obtain effective spatial features, including the following steps:

[0013] Preprocess the effective spatial features and detection features, then use spatial pyramid pooling to extract spatial features, and use edge detection and corner detection to extract detection features;

[0014] The extracted spatial features and detection features are combined, and the principal component analysis is used to perform feature transformation to obtain the first layer of fusion features;

[0015] According to the first layer of fusion features, select the features that contribute most to the task, use the gradient boosting machine to sort the feature importance, obtain the important features, fuse the features based on different scales, and obtain the second layer of fusion features;

[0016] The second layer fusion features are subjected to feature dimension reduction and enhancement to obtain fusion features.

[0017] Furthermore, the GIS data is pre-processed and spatially registered with the fusion feature, a feature association between the spatial feature and the attribute information is established, and weighted splicing is performed, including the following steps:

[0018] Preprocessing the GIS data, obtaining the preprocessed GIS data, and performing spatial registration with the fusion features to obtain registered fusion features;

[0019] The fused features are associated with the GIS data according to the attribute information to obtain the attribute association results, and the registered fused features are associated with the preprocessed GIS spatial data using spatial connection to obtain the spatial association features;

[0020] The principal component analysis method is used to confirm the respective weights, and the results of spatial correlation features and attribute correlation are weightedly spliced ​​to construct a new feature vector, that is, to obtain the correlation fusion result.

[0021] Furthermore, obtaining the highway infrastructure assessment result comprises the following steps:

[0022] According to the characteristics of highway infrastructure, define evaluation indicators, set scoring standards for each evaluation indicator, and assign weights according to the importance of the indicator. The evaluation indicators include road conditions, traffic flow, maintenance history, and accident rate;

[0023] According to the relationship between the fusion results and the evaluation indicators, the conversion relationship between the output of the prediction model and the evaluation score is determined;

[0024] The relevant data of highway infrastructure is input into the evaluation prediction model to obtain the prediction result, that is, the probability of each evaluation indicator; based on the prediction result and the scoring criteria, a score is given to each evaluation indicator, and the score of each indicator is multiplied by its weight, and the sum is obtained to obtain the total score; based on the total score and the set evaluation rules, the highway infrastructure evaluation score is divided into different levels to obtain the highway infrastructure evaluation result.

[0025] Furthermore, obtaining effective spatial features comprises the following steps:

[0026] Correcting, clipping and denoising the spatial data to obtain preprocessed spatial data;

[0027] The pre-processed spatial data are respectively subjected to gray level co-occurrence matrix, edge detection, spectrum analysis, image segmentation and pattern recognition to obtain a variety of spatial features, including texture features, shape features, spectrum features and structural features;

[0028] Use principal component analysis to select multiple features and obtain effective spatial features;

[0029] The method of obtaining effective detection features comprises the following steps:

[0030] Cleaning, normalizing and filling missing values ​​of the detection data to obtain preprocessed detection data;

[0031] Through descriptive statistical analysis, Fourier transform, time series analysis, image processing and machine learning, a variety of detection features are obtained, including statistical features, frequency domain features, time domain features and damage features;

[0032] The multiple detection features are selected by recursive feature elimination to obtain effective detection features. Furthermore, the spatial data is obtained through satellite remote sensing and drone aerial photography, which is image and video data of the highway;

[0033] The detection data uses various sensors and detection equipment to collect road surface condition and structure data;

[0034] The GIS data utilizes a geographic information system to obtain the geographical location, topography, and traffic flow data of the highway.

[0035] A highway infrastructure assessment and processing system, comprising:

[0036] The multimodal feature acquisition module is used to acquire multimodal highway infrastructure data, including spatial data, detection data, and GIS data; the spatial data and detection data are preprocessed and features are extracted to obtain effective spatial features and detection features;

[0037] The feature fusion module uses hierarchical fusion to fuse spatial features and detection features to obtain fusion features; the GIS data is preprocessed and spatially registered with the fusion features to establish feature associations between spatial features and attribute information, and weighted splicing is performed to obtain association fusion results;

[0038] The evaluation and analysis module is used to select the support vector machine as the basic model based on the association fusion results, use the random forest and gradient boosting decision tree as the first-layer model, group the logistic regression as the second-layer model, set the evaluation indicators, and construct an evaluation prediction model; obtain the association fusion results of the highway infrastructure to be evaluated, input them into the evaluation prediction model, obtain the evaluation indicator probability, and obtain the highway infrastructure evaluation results by combining the weighted scoring and grading rules set for the evaluation indicators.

[0039] Compared with the prior art, the beneficial effects of the present invention are as follows: by integrating multimodal data, a more comprehensive data perspective is provided, the diversity and completeness of data are enhanced, the accuracy and reliability of evaluation are improved, and the deviation caused by the singleness of data is reduced. The multi-layer integrated learning model is adopted, combined with basic models such as support vector machines, the generalization ability of the model is enhanced, and the accuracy and robustness of prediction are improved. Through spatial registration and spatial connection operations, the association between spatial features and attribute information is established. Richer contextual information is provided for the evaluation, and the interpretability and practicality of the evaluation results are enhanced. The automation and systematization of the evaluation process are realized, and human intervention is reduced. The efficiency of the evaluation is improved, the possibility of human error is reduced, and the evaluation process is made more standardized.

[0040] Improve the accuracy of assessment results through multimodal data fusion and advanced machine learning models. Provide data support and decision-making basis for the maintenance, transformation and planning of highway infrastructure. Help decision makers to reasonably allocate maintenance resources and budgets through accurate assessment. Automated and systematic assessment process reduces manual intervention and improves management efficiency. Make full use of existing data resources, explore data value, and provide more information for highway management. Achieve comprehensive, efficient and accurate assessment of highway infrastructure, thereby providing support for the long-term health and sustainable development of highways. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The drawings described herein are used to provide further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute improper limitations on the present application.

[0042] Figure 1 It is a schematic diagram of the process structure of the present invention;

[0043] Figure 2 It is a schematic diagram of the structure of the acquisition process of the fusion feature of the present invention;

[0044] Figure 3 It is a schematic diagram of the process structure of obtaining the associated fusion result of the present invention;

[0045] Figure 4 It is a schematic diagram of the process structure of obtaining a kilometer infrastructure assessment structure according to the present invention. DETAILED DESCRIPTION

[0046] The evaluation of highway infrastructure is crucial to ensure traffic safety, improve transportation efficiency and extend the life of roads. Traditional evaluation methods rely on manual inspection and expert experience, which is not only time-consuming and labor-intensive, but also difficult to fully and accurately reflect the real condition of the highway.

[0047] In recent years, with the development of technologies such as the Internet of Things, big data, cloud computing, and artificial intelligence, the digital processing capabilities of highway infrastructure have been significantly improved. The following technological development trends provide new opportunities for the fine digital processing of highway infrastructure.

[0048] The application of high-tech sensor technology makes the monitoring data of highway infrastructure more accurate and real-time. Big data analysis technology provides a powerful tool for the processing and analysis of massive highway infrastructure data. The development of artificial intelligence algorithms, especially deep learning technology, provides new methods for feature extraction and pattern recognition of highway infrastructure. The widespread application of BIM technology in the design, construction and operation and maintenance stages of highway infrastructure provides rich information resources for digital processing.

[0049] With the development of information technology and artificial intelligence technology, a highway infrastructure assessment and processing method based on multimodal data fusion and intelligent prediction model has emerged. The present invention aims to provide a highway infrastructure assessment and processing method and system to solve the problems existing in the prior art. Figure 1 , specifically including:

[0050] Acquire multimodal highway infrastructure data, including spatial data, inspection data, and GIS data;

[0051] After preprocessing the spatial data and the detection data and extracting features, spatial features and detection features are obtained;

[0052] Use hierarchical fusion to fuse spatial features and detection features to obtain fusion features;

[0053] After preprocessing, GIS data is spatially registered with the fusion feature, the feature association between spatial features and attribute information is established, and weighted splicing is performed to obtain the association fusion result;

[0054] Based on the association fusion results, support vector machine is selected as the basic model, random forest and gradient boosting decision tree are used as the first layer model, logistic regression is grouped as the second layer model, evaluation indicators are set, and an evaluation prediction model is constructed;

[0055] The associated fusion results of the highway infrastructure to be evaluated are obtained and input into the evaluation prediction model to obtain the probability of the evaluation indicators. Combined with the weighted scoring and grading rules set for the evaluation indicators, the highway infrastructure evaluation results are obtained.

[0056] Improve the accuracy and real-time performance of highway infrastructure data collection; improve data processing efficiency and reduce manual intervention; realize multi-source data integration and resource sharing; and provide intelligent support for highway infrastructure maintenance decisions. By achieving the above goals, the present invention helps to improve the management level of highway infrastructure, reduce operation and maintenance costs, extend service life, and provide strong guarantees for the sustainable development of my country's highway industry.

[0057] The present invention will be further described below in conjunction with the accompanying drawings and implementation modes.

[0058] Embodiment 1:

[0059] See also Figure 1-Figure 4 , the following is a specific embodiment, which describes how to use the hierarchical fusion method to perform secondary fusion matching of effective spatial features and effective detection features with GIS data to evaluate the status of highway infrastructure:

[0060] The first is data collection. For spatial data, high-resolution images of roads are obtained through satellite remote sensing or drone aerial photography. For detection data, vehicle-mounted detection equipment is used to collect data such as cracks, ruts, and flatness on the road surface. And GIS data, including attribute information such as the geographical location of the road, traffic flow, and historical maintenance records.

[0061] Correct and register image data to ensure the accuracy of spatial data. Clean detection data to remove noise and outliers. Standardize GIS data to ensure a unified format for attribute information.

[0062] Extract texture features, spectral features, etc. from spatial data. Extract crack density, rutting depth, flatness index and other features from detection data. Extract attribute features such as traffic flow and historical maintenance times from GIS data. Use principal component analysis (PCA) to reduce the dimension of the extracted features and select the features with the largest contribution rate. Apply recursive feature elimination (RFE) method to further screen out the most important features for pavement condition assessment.

[0063] The spatial features and detection features are combined through feature-level fusion methods such as feature splicing to perform the first-level fusion:

[0064] After preprocessing, GIS data is spatially registered with the fusion feature, the feature association between spatial features and attribute information is established, and weighted splicing is performed to obtain the association fusion result;

[0065] Use an ensemble learning model, such as random forest, to train the fused features. Apply support vector machine (SVM) as the second layer model to further classify or regress the output of random forest. Evaluate the performance of the model through cross-validation and adjust the model parameters to optimize the prediction results. Use the test set data to verify the generalization ability of the model.

[0066] Analyze the model output to determine the degree of road damage and priority maintenance areas. Use GIS software to visualize the assessment results and show the road condition level in different areas.

[0067] According to the evaluation results, a highway maintenance and repair plan is formulated. Data support is provided for budget allocation and resource planning. Through this example, we can see the application process of the new method in highway infrastructure evaluation and how to improve the accuracy and practicality of the evaluation through multi-level data fusion and machine learning models.

[0068] Example 2

[0069] Highway infrastructure is a key component of the modern transportation system, and its health status is directly related to traffic safety and efficiency. Therefore, it is very important to accurately evaluate and maintain highway infrastructure. This embodiment will introduce in detail an innovative highway infrastructure evaluation and processing method and system, which uses multimodal data fusion and advanced machine learning technology to achieve a comprehensive evaluation of highway health status.

[0070] Data Acquisition and Preprocessing,First, the method involves acquiring multimodal highway infrastructure data,,including spatial data, detection data, and GIS data.,Spatial data are such as satellite images or photos taken by,drones, detection data include information about road conditions collected by various,sensors, and GIS data contain relevant attribute information in,geographic information systems.

[0071] After preprocessing the spatial data and detection data, the features are extracted to obtain effective spatial features and detection features. The spatial data is preprocessed through correction, clipping and denoising, and then the texture features, shape features, spectral features and structural features are extracted through gray-level co-occurrence matrix, edge detection, spectral analysis, image segmentation and pattern recognition. The detection data is cleaned, normalized and filled with missing values, and statistical features, frequency domain features, time domain features and damage features are extracted through descriptive statistical analysis, Fourier transform, time series analysis, image processing and machine learning.

[0072] Next, hierarchical fusion is used to fuse spatial features and detection features to obtain fused features. First, effective spatial features and detection features are preprocessed, and then spatial features are extracted using spatial pyramid pooling, and detection features are extracted using edge detection and corner detection. The extracted spatial features and detection features are combined, and feature transformation is performed using principal component analysis to obtain the first layer of fused features. Based on the first layer of fused features, the features that contribute most to the task are selected, and the gradient boosting machine is used to sort the feature importance to obtain important features. The features are fused based on different scales to obtain the second layer of fused features. The second layer of fused features are subjected to feature dimensionality reduction and enhancement to obtain fused features.

[0073] Perform secondary fusion matching on the fusion features and GIS data, establish the association between spatial features and attribute information, and obtain the associated fusion results. Preprocess the GIS data and perform spatial registration on the fusion features. Use the spatial join operation to associate the registered fusion features with the preprocessed GIS spatial data, and associate the fusion features with the GIS attribute data according to the attribute information. Assign weights to different attribute information.

[0074] Based on the set evaluation rules, the highway infrastructure evaluation results are obtained by combining the association fusion results and the evaluation prediction model. The support vector machine is selected as the basic model, the random forest and gradient boosting decision tree are used as the first layer model, the logistic regression group is used as the second layer model, and the evaluation prediction model is constructed by combining the historical highway infrastructure evaluation data.

[0075] Through multimodal data fusion and advanced machine learning technology, a comprehensive and accurate method for evaluating the health status of highways is provided. This method not only improves the accuracy of the assessment, but also provides strong data support for highway maintenance and management.

[0076] Example 3

[0077] The goal of this embodiment is to evaluate the infrastructure conditions of a certain highway section, including factors such as road flatness, traffic flow, road surface damage, etc., so as to provide managers with maintenance, upgrade and improvement suggestions on the road section.

[0078] Spatial data: Use satellite remote sensing such as Landsat or Sentinel satellite data and drone aerial data to collect geographic information, road surface images and videos of the highway segment.

[0079] Inspection data: road surface condition data collected by road inspection equipment such as LiDAR sensors and ground-based radar (GPR), including road surface morphology, crack information, and structural health data of the road.

[0080] GIS data: Obtain the geographical location, terrain data, traffic flow, road history information, etc. of the road section through the Geographic Information System (GIS).

[0081] 2. The specific steps include:

[0082] Step 1: Data preprocessing and feature extraction

[0083] Spatial data preprocessing:

[0084] The collected satellite remote sensing images and drone aerial images are rectified, cropped and denoised to remove irrelevant information and noise.

[0085] Use image processing techniques such as gray-level co-occurrence matrix, edge detection, spectral analysis, image segmentation, etc. to extract spatial features in the image. You can obtain features such as texture features (reflecting the roughness of the road surface), shape features (shape of road surface cracks), spectral features (spectral reflectance of different ground types), and structural features (road surface cracks, potholes, etc.).

[0086] Detection data preprocessing:

[0087] Clean and normalize sensor data, and fill in missing values.

[0088] Extract detection features from sensor data using methods such as Fourier transform, time series analysis, and image processing. This may include:

[0089] Statistical characteristics: such as mean, standard deviation, etc.

[0090] Frequency domain features: The main signal components in the frequency domain are obtained through Fourier transform.

[0091] Time domain characteristics: the time characteristics of road surface vibration.

[0092] Damage characteristics: such as crack width, depth, road surface subsidence, etc.

[0093] Step 2: Feature fusion and association

[0094] Feature Fusion:

[0095] The spatial features and detection features are combined using a hierarchical fusion method. First, spatial features are extracted by spatial pyramid pooling, and detection features are extracted by edge detection and corner detection.

[0096] Principal component analysis (PCA) is used to reduce the dimension of features, select the main features, and sort the contributions of different features through the gradient boosting machine (GBDT) to obtain the second layer of fusion features.

[0097] Enhance the fused features and use dimensionality reduction techniques such as t-SNE or PCA to further reduce feature redundancy and improve the training efficiency and accuracy of the model.

[0098] GIS data and fusion feature matching:

[0099] Preprocessing of GIS data, including standardization of geographic coordinates and spatial registration, enables GIS data to be spatially matched with other features.

[0100] The fused features are combined with GIS data using spatial join operations, and the geographical features of the road are associated with its structure, damage, traffic and other information through attribute information.

[0101] Weights are assigned to different attribute information such as traffic flow, historical maintenance records, etc., and finally a fused feature vector containing spatial information and attribute information is generated, which is the associated fusion result.

[0102] Step 3: Evaluation Model Building and Evaluation

[0103] Build an evaluation prediction model:

[0104] Support vector machine (SVM) is selected as the base model to handle the evaluation task.

[0105] Random forest (RF) and gradient boosted decision tree (GBDT) are used as the first layer models to make preliminary predictions on the features.

[0106] Logistic Regression was used as the second-layer model and combined with historical highway infrastructure assessment data to establish the final assessment prediction model.

[0107] Define evaluation rules and evaluation indicators:

[0108] Set evaluation metrics, such as:

[0109] Road conditions: assessed by factors such as crack distribution, potholes, and degree of damage.

[0110] Traffic Flow: Evaluate traffic density and traffic volume through traffic flow sensor data.

[0111] Maintenance history: Combine maintenance records in GIS data to evaluate past maintenance frequency and effectiveness.

[0112] Accident rate: Use historical accident data to evaluate the traffic safety of the road section.

[0113] Set scoring criteria for each evaluation indicator and assign weights to each indicator according to its importance.

[0114] Evaluation results calculation:

[0115] The input data includes fusion features and historical evaluation data into the evaluation prediction model, and the predicted value or probability of each evaluation indicator is output.

[0116] Score each evaluation indicator according to the scoring criteria, multiply the score by its respective weight, and obtain the weighted sum of each indicator.

[0117] Finally, based on the total score and preset evaluation rules, highway infrastructure is divided into different levels such as excellent, good, average, and poor.

[0118] Step 4: Final evaluation results

[0119] Combining the assessment model and rules, the final assessment result is the comprehensive health status of highway infrastructure. For example:

[0120] The total score is between 80-100 and the evaluation result is "excellent".

[0121] The total score is between 60-79 and the evaluation result is "good".

[0122] The total score is between 40-59, the evaluation result is "average", and maintenance needs to be arranged.

[0123] The total score is between 0-39, and the assessment result is "poor" and needs repair or upgrading.

[0124] Example 4

[0125] When applied specifically, the evaluation process for a highway section such as the Beijing-Tianjin Expressway is as follows:

[0126] Spatial data: Satellite remote sensing data provided high-definition images of the highway section, showing cracks and damage in the road surface.

[0127] Image processing extracts textural features such as pavement roughness and shape features such as crack morphology.

[0128] Inspection data: Deep cracks and subsidence areas in the road surface were detected using LiDAR and GPR equipment.

[0129] The data analysis results include time domain characteristics such as vibration changes and damage characteristics such as crack depth.

[0130] GIS data: Obtain the traffic flow, traffic accident rate and historical maintenance data of the road section through the GIS system.

[0131] If it is found that the traffic accident rate on this road section is high and there have been multiple repair records in the past five years.

[0132] Evaluation and prediction model: Based on the fusion features such as the number of cracks, traffic flow data, accident history, etc., the evaluation and prediction model outputs the scores of the highway section in terms of road conditions, traffic flow, accident rate and other indicators.

[0133] The traffic flow score was 75 points (good), the accident rate score was 60 points (average), and the road condition score was 50 points (average).

[0134] Final evaluation result: Based on weighted calculation, the total score was 68 points, so the infrastructure assessment result of this highway section was "good", but it was recommended to strengthen maintenance work, especially to make improvements in areas with high accident incidence.

[0135] This embodiment demonstrates a highway infrastructure assessment method based on multimodal data and multi-level fusion. By integrating spatial data, detection data, and GIS data, and using machine learning models for prediction and assessment, it can provide accurate highway health status assessment results, provide valuable information for decision makers, and help optimize highway maintenance and management strategies. It can comprehensively and accurately assess the status of highway infrastructure and provide a scientific basis for road maintenance and management. With the continuous development of technology, data-driven highway infrastructure assessment methods will play an increasingly important role in the future.

[0136] It should be noted that, in the present invention, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0137] The above embodiments are merely examples of the present invention and do not limit the protection scope of the present invention. All designs that are the same or similar to the present invention fall within the protection scope of the present invention.

Claims

1. A highway infrastructure assessment and processing method, characterized in that: The following steps are involved: Acquire multimodal highway infrastructure data, including spatial data, inspection data, and GIS data; After preprocessing the spatial data and the detection data and extracting features, spatial features and detection features are obtained; Use hierarchical fusion to fuse spatial features and detection features to obtain fusion features; After preprocessing, GIS data is spatially registered with the fusion feature, the feature association between spatial features and attribute information is established, and weighted splicing is performed to obtain the association fusion result; Based on the association fusion results, support vector machine is selected as the basic model, random forest and gradient boosting decision tree are used as the first layer model, logistic regression group is used as the second layer model, evaluation indicators are set, and an evaluation prediction model is constructed; The associated fusion results of the highway infrastructure to be evaluated are obtained and input into the evaluation prediction model to obtain the probability of the evaluation indicators. The evaluation results of the highway infrastructure are obtained by combining the weighted scoring and grading rules set for the evaluation indicators.

2. The highway infrastructure assessment processing method according to claim 1, characterized in that: The method of fusing the spatial features and the detection features by using hierarchical fusion includes the following steps: Preprocess the effective spatial features and detection features, then use spatial pyramid pooling to extract spatial features, and use edge detection and corner detection to extract detection features; The extracted spatial features and detection features are combined, and the principal component analysis is used to perform feature transformation to obtain the first layer of fusion features; According to the first-layer fusion features, select the features that contribute most to the task, use the gradient boosting machine to sort the feature importance, obtain the important features, fuse the features based on different scales, and obtain the second-layer fusion features; The second layer fusion features are subjected to feature dimension reduction and enhancement to obtain fusion features.

3. The highway infrastructure assessment processing method according to claim 1, characterized in that: The GIS data is pre-processed and spatially registered with the fusion feature, a feature association between the spatial feature and the attribute information is established, and weighted splicing is performed, including the following steps: Preprocessing the GIS data, obtaining the preprocessed GIS data, and performing spatial registration with the fusion features to obtain registered fusion features; The fused features are associated with the GIS data according to the attribute information to obtain the attribute association results, and the registered fused features are associated with the preprocessed GIS spatial data using spatial connection to obtain the spatial association features; The principal component analysis method is used to confirm the respective weights, and the results of spatial correlation features and attribute correlation are weightedly spliced ​​to construct a new feature vector, that is, to obtain the correlation fusion result.

4. The highway infrastructure assessment processing method according to claim 1, characterized in that: The method of obtaining the highway infrastructure assessment result comprises the following steps: According to the characteristics of highway infrastructure, evaluation indicators are defined, scoring criteria are set for each evaluation indicator, and weights are assigned according to the importance of the indicator. The evaluation indicators include road conditions, traffic flow, maintenance history, and accident rate; According to the relationship between the fusion results and the evaluation indicators, the conversion relationship between the output of the prediction model and the evaluation score is determined; The relevant data of highway infrastructure is input into the evaluation prediction model to obtain the prediction result, that is, the probability of each evaluation indicator; based on the prediction result and the scoring criteria, a score is given to each evaluation indicator, and the score of each indicator is multiplied by its weight, and the sum is obtained to obtain the total score; based on the total score and the set evaluation rules, the highway infrastructure evaluation score is divided into different levels to obtain the highway infrastructure evaluation result.

5. The highway infrastructure assessment processing method according to claim 1, characterized in that: The method of obtaining effective spatial features comprises the following steps: Correcting, clipping and denoising the spatial data to obtain preprocessed spatial data; The pre-processed spatial data are respectively subjected to gray level co-occurrence matrix, edge detection, spectrum analysis, image segmentation and pattern recognition to obtain a variety of spatial features, including texture features, shape features, spectrum features and structural features; Use principal component analysis to select multiple features and obtain effective spatial features; The method of obtaining effective detection features comprises the following steps: Cleaning, normalizing and filling missing values ​​of the detection data to obtain preprocessed detection data; Through descriptive statistical analysis, Fourier transform, time series analysis, image processing and machine learning, a variety of detection features are obtained, including statistical features, frequency domain features, time domain features and damage features; The multiple detection features are selected by recursive feature elimination to obtain effective detection features.

6. The highway infrastructure assessment processing method according to claim 1, characterized in that: The spatial data is obtained through satellite remote sensing and drone aerial photography, and is image and video data of the highway; The detection data uses various sensors and detection equipment to collect road surface condition and structure data; The GIS data utilizes a geographic information system to obtain the geographical location, topography, and traffic flow data of the highway.

7. A highway infrastructure assessment and processing system, characterized in that: include: Multimodal feature acquisition module, used to acquire multimodal highway infrastructure data, including spatial data, detection data, and GIS data; After preprocessing the spatial data and detection data and extracting features, effective spatial features and detection features are obtained; The feature fusion module uses hierarchical fusion to fuse spatial features and detection features to obtain fusion features; the GIS data is preprocessed and spatially registered with the fusion features to establish feature associations between spatial features and attribute information, and weighted splicing is performed to obtain association fusion results; The evaluation and analysis module is used to select the support vector machine as the basic model based on the association fusion results, use the random forest and gradient boosting decision tree as the first-layer model, group the logistic regression as the second-layer model, set the evaluation indicators, and construct an evaluation prediction model; obtain the association fusion results of the highway infrastructure to be evaluated, input them into the evaluation prediction model, obtain the evaluation indicator probability, and obtain the highway infrastructure evaluation results by combining the weighted scoring and grading rules set for the evaluation indicators.

8. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

9. A computer device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method described in any one of claims 1 to 6 is implemented.

Images