A road inspection holographic perception system and method based on artificial intelligence

Abstract

The present invention discloses a holographic perception system and method for road inspection based on artificial intelligence, which specifically relates to the field of road inspection and is used to solve the problem of hidden problem detection of road inspection systems under complex environmental conditions. The system collects and constructs multi-sensor data sets under different environmental conditions, defines environmental labels, classifies and stores and records environmental labels, performs denoising processing based on wavelet basis functions and decomposition layers to remove environmental noise, then analyzes time synchronization and space alignment errors and feature change rates, comprehensively calculates multi-sensor overall fusion index and environmental adaptability index, obtains global confidence coefficient, further summarizes global confidence coefficient data, calculates volatility and constructs an abnormal frequency matrix, analyzes evenness, generates a global hidden detection index to quantify hidden problem detection capability, comprehensively evaluates the hidden problem detection capability of the road inspection system in a complex environment, and ensures road safety and inspection quality.

Description

Technical Field

[0001] The present invention relates to the field of road inspection, and more specifically, to an artificial intelligence-based holographic perception system and method for road inspection. Background Art

[0002] With the advancement of smart city construction, road inspection systems play an increasingly important role in urban infrastructure management. Modern road inspection systems usually rely on a variety of sensors, such as cameras, lidars, and temperature sensors, which can collect multi-dimensional data on road conditions in real time. Under different environmental conditions, such as sunny days, rainy days, daytime, and nighttime, the data performance of sensors will be different, which puts high demands on the system's data processing and analysis capabilities. In order to ensure the safety and smooth flow of roads, it is necessary to effectively integrate and analyze these multi-source data to accurately detect and warn of potential hidden problems, such as road damage and traffic facility failures.

[0003] Existing road inspection systems have significant deficiencies in their ability to detect hidden problems. Although these systems are able to collect a large amount of multi-sensor data, they usually lack effective methods to identify and quantify hidden problems when processing and analyzing this data. Traditional hidden problem detection methods often rely on data features of a single sensor or a simple set of rules, and cannot fully utilize the comprehensive information of multi-source data. In addition, the existing technologies lack consistency and reliability in their ability to detect hidden problems under different environmental conditions, and are difficult to adapt to complex and changing actual situations. For example, at night or in bad weather conditions, the data quality of sensors may decline, resulting in a significant decrease in the system's recognition rate of small cracks or potential faults on the road surface. Existing systems lack systematicity and precision in evaluating and quantifying their ability to detect hidden problems, and cannot provide a scientific basis to optimize the reliability and adaptability of the system, affecting the efficiency and quality of road inspections. An improved method is urgently needed to enhance the system's ability to detect hidden problems under various environmental conditions to ensure the efficient operation of the road inspection system and road safety.

[0004] In order to solve the above problems, a technical solution is now provided. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present invention provide a road inspection holographic perception system and method based on artificial intelligence, which collects and constructs multi-sensor data sets under different environmental conditions, defines environmental labels, classifies and stores and records environmental labels, decomposes multi-sensor data based on wavelet basis functions and decomposition layers to extract scale features, denoises high-frequency components and reconstructs low-frequency and high-frequency components, and reconstructs the final signal layer by layer, collects sensor data under various environments after denoising, analyzes time synchronization and spatial alignment errors, and measures the change rate and difference of sensor characteristics in different environments, comprehensively calculates the overall fusion index of multiple sensors and the environmental adaptability index, and obtains the global confidence coefficient to analyze the performance and adaptability under different environmental conditions during road inspection, further summarizes the global confidence coefficient data, calculates the volatility and constructs the abnormal frequency matrix, analyzes the pairing, and comprehensively generates the global hidden detection index to quantify the hidden problem detection capability, thereby comprehensively evaluating the detection capability and maintenance efficiency of the system in a complex environment, ensuring road safety and inspection quality, improving the reliability and adaptability of the system, and providing a scientific basis for optimizing the road inspection system to solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] S1, collects and constructs multi-sensor data sets under different environmental conditions, defines environmental labels, classifies and stores data, and records environmental labels;

[0008] S2, decomposes the collected multi-sensor data based on wavelet basis functions and decomposition layers to extract scale features, denoises high-frequency components and reconstructs low-frequency and high-frequency components, and reconstructs the final signal layer by layer;

[0009] S3, collects sensor data in various environments after denoising, analyzes the time synchronization and spatial alignment errors of sensor data in different environments, measures the change rate and difference of sensor features in different environments, comprehensively calculates the corresponding multi-sensor overall fusion index and environmental adaptability index, and obtains the global confidence coefficient to analyze the performance and adaptability of road inspection under different environmental conditions;

[0010] S4, summarizes the global confidence coefficient data, calculates the volatility and constructs the abnormal frequency matrix, analyzes the evenness, and comprehensively generates the global hidden detection index to quantify the hidden problem detection capability.

[0011] In a preferred embodiment, step S2 includes the following contents:

[0012] S2.1, based on the selected wavelet basis function, the number of decomposition layers is set to decompose the road data collected by the multi-sensor into components of different frequencies, including low-frequency components and high-frequency components;

[0013] S2.2, represents the signal at different resolutions, retaining the feature information at each scale;

[0014] S2.3, use wavelet transform to denoise the high frequency components, retain the main features of the signal and remove the noise components;

[0015] Perform soft threshold processing on high-frequency components, that is, the part smaller than the corresponding threshold is considered as noise and removed or weakened;

[0016] Recombine the processed high-frequency components and low-frequency components to obtain a denoised signal;

[0017] S2.4, reconstructing the denoised low-frequency and high-frequency components to restore the main features of the original signal; restoring the processed low-frequency components to a time domain signal; restoring the processed high-frequency components to a time domain signal;

[0018] The low-frequency and high-frequency components are recombined to obtain a reconstructed signal, which specifically includes the following steps:

[0019] First, the low-frequency and high-frequency components of each layer are obtained from the wavelet decomposition process, such as the low-frequency components of the third layer and the high-frequency components of the first, second, and third layers;

[0020] The low-frequency components of the third layer are used as the initial reconstructed signal;

[0021] The low-frequency components of the third layer and the high-frequency components of the third layer are upsampled and convolved through a wavelet reconstruction filter, and then the convolution results are added to obtain the reconstructed signal of the second layer;

[0022] The reconstructed signal of the second layer and the high-frequency components of the second layer are upsampled and convolved through a wavelet reconstruction filter, and then the convolution results are added to obtain the reconstructed signal of the first layer;

[0023] The reconstructed signal of the first layer and the high-frequency components of the first layer are upsampled and convolved through a wavelet reconstruction filter, and then the convolution results are added to obtain the final reconstructed signal;

[0024] Through layer-by-layer combination and reconstruction, the low-frequency and high-frequency components of all levels are combined to obtain the final reconstructed signal.

[0025] In a preferred embodiment, step S3 includes the following contents:

[0026] The logic for obtaining the multi-sensor overall fusion index is:

[0027] a1, classify the sensor data after denoising in step S2 to ensure that each set of data has a corresponding environmental label ; Indicates in the environment The sensor data set collected below; Indicates environmental label ;

[0028] a2, perform feature matching on each set of environmental data and calculate the time synchronization error and spatial alignment error;

[0029] ;in, Represents the feature matching function Sensor data and Perform feature matching; Indicates sensor and In the environment The feature matching matrix under

[0030] ;in, Indicates in the environment Next, sensor and sensors The absolute value of the time difference of the collected data is calculated by and The difference between the two sensors can be obtained and In the environment Is there a time synchronization error in the data collection under Indicates sensor and In the environment The time synchronization error under

[0031] ;in, Indicates in the environment Next, sensor and sensors The spatial position difference is measured by Euclidean distance; Indicates sensor and In the environment The spatial alignment error under

[0032] a3, normalize the time synchronization error and the space alignment error, and mark the normalized time synchronization error as , the normalized spatial alignment error is marked as ;

[0033] a4, combine the normalized time synchronization error and spatial alignment error to calculate the error of each pair of sensors in the environment The fusion error under: ,in, Indicates sensor and In the environment The fusion error under

[0034] a5, based on each pair of sensors in the environment The fusion error under different environmental conditions is used to calculate the fusion stability index of each pair of sensors: ,in, Indicates sensor and In the environment Fusion stability index under ;

[0035] a6, comprehensively consider the fusion stability index of all sensor pairs under each environmental condition, calculate the multi-sensor overall fusion index, and obtain the multi-sensor overall fusion index through multiple iterations and combinations: ,in, Represents the multi-sensor overall fusion index; Indicates the total number of environmental conditions; In each environment The number of sensors used.

[0036] In a preferred embodiment, the logic for obtaining the environmental perception sensitivity index is:

[0037] b1, preprocess the multi-sensor data and extract feature information, including and , Indicates sensor In the environment The preprocessed data set under ; Indicates sensor In the environment The following feature set; Indicates sensor In the environment The following feature set;

[0038] b2, calculate the rate of change of sensor characteristics under different environmental conditions to evaluate the sensitivity of the sensor to environmental changes; ,in Indicates sensor and In the environment The characteristic change rate under

[0039] b3, calculate the environmental difference index under different environmental conditions: ,in, Indicates the environment Environmental difference index under ;

[0040] b4, normalize the environmental difference index and mark the normalized environmental difference index as ;

[0041] b5, based on the normalized environmental difference index, calculate the environmental perception sensitivity under each environmental condition, and combine the sensitivity under all environmental conditions to calculate the system's environmental perception sensitivity index. By combining the sensitivity under each environmental condition through the geometric mean method, the environmental perception sensitivity index is obtained. : .

[0042] In a preferred embodiment, the global confidence coefficient is generated by combining the environmental perception sensitivity index and the multi-sensor overall fusion index: ,in, represents the global confidence coefficient; and They are the environmental perception sensitivity index and the multi-sensor overall fusion index respectively; and They are the preset proportional coefficients corresponding to the environmental perception sensitivity index and the multi-sensor overall fusion index, and both are greater than zero.

[0043] In a preferred embodiment, step S4 includes the following contents:

[0044] S4.1, under different environmental conditions, continuously collect the time series data of the global confidence coefficient, record the global confidence coefficient value of the system at each time point, and use the global confidence coefficient value collected each time as part of the time series data to ensure that different time points under various environmental conditions are covered;

[0045] S4.2, calculate the volatility of the global confidence coefficient time series data to quantify the change of the global confidence coefficient during the observation period; first, calculate the change of the global confidence coefficient between adjacent time points, that is, the difference of the global confidence coefficient between time t and t-1, then calculate the absolute value of all changes, sum and divide by the difference between the maximum change and the minimum change, so as to obtain the volatility;

[0046] S4.3, construct an abnormal frequency matrix to record the frequency of occurrence of each time when the global confidence coefficient is less than the threshold, and analyze the frequency and interval below the threshold; first, detect the abnormal value in the time series, that is, the situation where the global confidence coefficient is less than the preset threshold, and record its time point each time an abnormal value is detected, then calculate the time interval between each pair of abnormal values, and construct the abnormal frequency matrix;

[0047] S4.4, based on the abnormal frequency matrix, calculate the parallelism index to quantify whether the situation where the global confidence coefficient is less than the threshold occurs regularly, reflecting the parallelism of the hidden problems of the system; first, perform nonlinear transformation on each frequency value in the abnormal frequency matrix and calculate its parallelism index, then calculate the mean and variance of the frequency matrix to quantify the distribution of the frequency values;

[0048] Parallelism index : ;

[0049] Mean of the frequency matrix : ;

[0050] Variance of the frequency matrix : ;

[0051] Indicates The second abnormality and The frequency between sub-anomalies; Represents an exception counter;

[0052] S4.5, combining volatility and evenness index, generates the comprehensive hidden problem detection capability evaluation index of the system. First, the volatility and evenness index are nonlinearly combined to generate the intermediate index. Then, the intermediate index is logarithmically transformed to enhance the nonlinear characteristics of the index. Finally, the global hidden detection index is calculated. : ,in, represents volatility; Represents the mean of the pair-parity index.

[0053] In a preferred embodiment, the global hidden detection index and the hidden detection critical threshold are compared to distinguish the detection capabilities of the road inspection system for various hidden problems under different environmental conditions. The specific comparison results are as follows: when the global hidden detection index is greater than or equal to the hidden detection critical threshold, an available signal is generated; when the global hidden detection index is less than the hidden detection critical threshold, a hidden danger signal is generated.

[0054] A road inspection holographic perception system based on artificial intelligence, including: an environment acquisition module, a feature extraction module, a performance analysis module and a hidden inspection evaluation module;

[0055] The environment acquisition module collects and constructs multi-sensor data sets under different environmental conditions, defines environmental labels, classifies and stores data and records environmental labels, and passes the collected multi-sensor data and environmental labels to the feature extraction module;

[0056] The feature extraction module decomposes the collected multi-sensor data based on the wavelet basis function and the number of decomposition layers to extract scale features, denoises the high-frequency components and reconstructs the low-frequency and high-frequency components, reconstructs the final signal layer by layer, and passes the processed feature data to the performance analysis module;

[0057] The performance analysis module collects sensor data in various environments after denoising, analyzes the time synchronization and spatial alignment errors of sensor data in different environments, measures the change rate and difference of sensor features in different environments, and comprehensively calculates the corresponding multi-sensor overall fusion index and environmental adaptability index to obtain the global confidence coefficient, so as to analyze the performance and adaptability under different environmental conditions during road inspection, and send the global confidence coefficient to the hidden inspection evaluation module;

[0058] The hidden detection assessment module summarizes the global confidence coefficient data, calculates the volatility and constructs the abnormal frequency matrix, analyzes the evenness, and comprehensively generates the global hidden detection index to quantify the hidden problem detection capability.

[0059] The technical effects and advantages of the road inspection holographic perception system and method based on artificial intelligence of the present invention are as follows:

[0060] 1. The present invention decomposes multi-sensor data based on wavelet basis functions and decomposition layers to extract scale features and denoise high-frequency components, which can significantly improve the data quality of the road inspection system, reduce noise interference, and enhance the resolution of the signal, so that more details can be observed at high resolution and the overall trend can be grasped at low resolution. The final signal is reconstructed layer by layer to ensure the integrity and accuracy of the data after processing, providing a more reliable data basis for subsequent road status assessment and decision-making. This method not only optimizes the data processing process, but also improves the system's adaptability under complex environmental conditions, helps to timely discover and locate road defects, improve inspection efficiency and accuracy, and ultimately ensure road safety and maintenance quality.

[0061] 2. The present invention can comprehensively identify and quantify the hidden problem detection capability of the road inspection system under different environmental conditions through in-depth analysis and processing of the global confidence coefficient data. First, by continuously collecting the global confidence coefficient time series data and calculating its volatility, the stability of the system at each time point can be dynamically evaluated, revealing its performance differences under different environmental conditions. Secondly, by constructing an abnormal frequency matrix and calculating the evenness index, the occurrence law of hidden problems can be systematically analyzed, the complexity and frequency of these problems can be quantified, and the accuracy and comprehensiveness of the detection results can be ensured. Finally, by generating a global hidden inspection index by combining volatility and evenness index, an objective and quantitative evaluation index can be provided, which intuitively reflects the hidden problem detection capability of the system under different environmental conditions. It not only improves the recognition accuracy and response speed of hidden problems, ensures the stability and reliability of the system, but also helps to timely discover and prevent potential road safety hazards, optimize the system's maintenance and management strategies, thereby extending the service life of road facilities, reducing maintenance costs, and ensuring the safety and smoothness of road traffic. Through this comprehensive and detailed analysis, the efficiency and effect of the road inspection system in practical applications can be greatly improved, providing solid technical support for the construction of smart cities. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 A schematic diagram of a process of a road inspection holographic perception method based on artificial intelligence according to the present invention;

[0063] Figure 2 The present invention is a schematic structural diagram of an artificial intelligence-based road inspection holographic perception system. DETAILED DESCRIPTION

[0064] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0065] Embodiment 1: Figure 1 The present invention provides a road inspection holographic perception method based on artificial intelligence, comprising:

[0066] S1, collects and constructs multi-sensor data sets under different environmental conditions, defines environmental labels, classifies and stores data, and records environmental labels;

[0067] S2, decomposes the collected multi-sensor data based on wavelet basis functions and decomposition layers to extract scale features, denoises high-frequency components and reconstructs low-frequency and high-frequency components, and reconstructs the final signal layer by layer;

[0068] S3, collects sensor data in various environments after denoising, analyzes the time synchronization and spatial alignment errors of sensor data in different environments, measures the change rate and difference of sensor features in different environments, comprehensively calculates the corresponding multi-sensor overall fusion index and environmental adaptability index, and obtains the global confidence coefficient to analyze the performance and adaptability of road inspection under different environmental conditions;

[0069] S4, summarizes the global confidence coefficient data, calculates the volatility and constructs the abnormal frequency matrix, analyzes the evenness, and comprehensively generates the global hidden detection index to quantify the hidden problem detection capability.

[0070] In order to ensure the stability and accuracy of the road inspection system in various complex environments, it is necessary to collect and analyze multi-sensor data under different environmental conditions. Modern road inspection systems usually rely on a variety of sensors, such as lidar, cameras, and ultrasonic sensors, which can provide detailed road status information. However, different environmental conditions, such as lighting, weather, speed, and terrain, will have different degrees of impact on the data collection of sensors. In order to comprehensively evaluate the performance of the system, it is necessary to collect data under various environmental conditions and accurately classify and store these data. In this process, it is crucial to define clear environmental labels, which can help classify and store sensor data, making subsequent data processing and analysis more efficient and accurate. By systematically collecting and constructing multi-sensor data sets under different environmental conditions, and classifying, storing, and labeling the data, a solid data foundation can be provided for subsequent multi-resolution analysis and performance evaluation, thereby ensuring the reliability and adaptability of the system in practical applications.

[0071] Step S1 includes the following contents:

[0072] First, it is necessary to clarify and refine the classification of various environmental conditions, including light (such as strong light during the day, shadows, and low light at night), weather (such as rain, snow, fog, and haze), speed (such as highways and urban roads), and terrain (such as plains, mountains, and urban streets). The purpose of this step is to ensure the comprehensiveness and systematicness of data collection and provide accurate environmental basis for subsequent analysis.

[0073] Secondly, develop a detailed data collection plan based on different environmental conditions. Determine the collection period (such as morning, noon, afternoon, night) and collection frequency to ensure sufficient data under each environmental condition. In addition, prepare the required multi-sensor equipment (such as lidar, camera, ultrasonic sensor, etc.) and ensure the equipment is prepared and calibrated to ensure the quality and consistency of data collection.

[0074] During the actual road inspection process, data collection is carried out according to the plan, and each set of data is environmentally labeled. The current environmental conditions are recorded in real time and marked with corresponding environmental labels to ensure accurate data labeling. At the same time, the time and space synchronization of all sensors is ensured, and time synchronization is performed using the Network Time Protocol (NTP) or Precision Time Protocol (PTP), and spatial calibration is performed using calibration technology to provide an accurate data basis for subsequent time-space alignment and fusion.

[0075] The collected data is preprocessed, including data cleaning (removing noise and invalid data), format conversion (converting data into a unified format), etc., to ensure data quality. Then, the data is classified according to environmental labels and stored in a structured database. Each set of data should contain sensor raw data, timestamp, spatial location, and environmental labels. A high-quality multi-sensor dataset is established to facilitate subsequent analysis and processing.

[0076] Finally, establish an efficient data management system and design the database structure (including table structure, indexes, and relationships) to ensure efficient storage and fast retrieval of data. Back up data regularly to prevent data loss, and set up permission management to ensure data security and privacy protection. Through systematic and structured data management, a solid foundation is provided for subsequent complex data analysis and calculations.

[0077] In modern road inspection systems, the accuracy and reliability of sensor data directly affect road safety and maintenance efficiency. However, due to the diversity of sensor data and the complexity of environmental conditions, how to effectively process and analyze this data has become a major challenge. The data collected by sensors are often affected by various noises and interferences, such as changes in lighting, weather conditions, road conditions, etc. These factors cause the data to contain a large amount of high-frequency noise and low-frequency interference. In addition, the data fusion and alignment of different sensors also face technical difficulties, which further increases the complexity of data processing. When dealing with these complex problems, existing signal processing methods are usually unable to take into account the information of different frequency components at the same time, which affects the integrity and accuracy of the data. Traditional denoising and feature extraction methods often rely on simple filtering and statistical methods, and cannot fully explore and utilize the rich information contained in multi-frequency components. In the face of these problems, there is an urgent need for a more sophisticated and efficient method to decompose and reconstruct multi-sensor data in order to extract useful feature information at different resolutions. By decomposing the data at multiple levels and scales, we can better understand and analyze the various components of the signal, remove noise and extract key features, thereby improving the quality of the data and the accuracy of the analysis. This lays the foundation for subsequent performance evaluation and decision support, making the road inspection system more efficient and reliable in practical applications. The multi-resolution analysis method based on wavelet transform is proposed to meet this demand. By selecting appropriate wavelet basis functions and setting the number of decomposition layers, multi-sensor data is decomposed into low-frequency and high-frequency components, retaining the feature information at each scale, and then realizing refined data processing and analysis. This method can not only significantly reduce the noise in the data, but also provide clearer feature representation at different resolutions, providing strong technical support for the subsequent performance evaluation and optimization of the road inspection system.

[0078] Step S2 includes the following contents:

[0079] S2.1, based on the selected wavelet basis function (such as Daubechies wavelet or Haar wavelet), the number of decomposition layers is then set to decompose the road data collected by the multi-sensors into components of different frequencies, including low-frequency components and high-frequency components.

[0080] In this way, complex signals can be decomposed into parts of different frequencies, making it easier to analyze the various components in the signal and helping to observe the characteristics of the data at different resolutions.

[0081] When selecting wavelet basis functions, multiple factors such as signal characteristics, application requirements, and computational efficiency should be considered. For multi-sensor data processing in road inspection systems, Daubechies wavelets and Haar wavelets are more common and applicable choices. Daubechies wavelets have good time-frequency localization characteristics and are suitable for capturing mutations and short-term features in signals. They are very suitable for processing sensor data containing noise and irregular changes. Haar wavelets have the advantages of simple and fast calculations and are suitable for scenarios with high real-time processing requirements. Therefore, according to specific needs, Daubechies wavelets can be given priority to ensure the refinement of signal feature extraction, while Haar wavelets can be selected when computing resources are limited or real-time processing is required to balance accuracy and efficiency.

[0082] The number of decomposition layers should be set based on the characteristic frequency range of the signal and the analysis objectives. Usually, by pre-analyzing the data, the main frequency distribution range of the signal can be determined, so as to set the appropriate number of decomposition layers. The number of decomposition layers should be large enough to separate feature information of different scales, but not too many to avoid information redundancy and computational overhead caused by over-refinement. For multi-sensor data processing in road inspection systems, the typical number of decomposition layers is generally between 3 and 5 layers, which can effectively separate the main features and detailed changes of road conditions while controlling the computational complexity. The specific number of layers can be verified through experiments and adjusted according to data characteristics and analysis requirements to ensure the optimal decomposition effect while balancing feature extraction accuracy and computational efficiency.

[0083] S2.2, represents the signal at different resolutions, retaining the feature information at each scale, so that details can be observed at high resolution and the overall trend can be observed at low resolution.

[0084] The low-frequency components represent the main structure of the signal and usually contain most of the energy and main features.

[0085] The high frequency components contain both details and noise in the signal.

[0086] By combining the components at different levels, the original signal can be fully represented.

[0087] Through multi-resolution representation, the overall trend and details of the signal can be observed simultaneously, which makes it easier to discover anomalies and important features in road data and improves the comprehensiveness of the analysis.

[0088] S2.3, use wavelet transform to denoise the high-frequency components, retain the main features of the signal and remove the noise components.

[0089] Set an appropriate threshold to remove noise from high-frequency components.

[0090] The high-frequency components are subjected to soft threshold processing, that is, the part smaller than the corresponding threshold is considered as noise and is removed or weakened.

[0091] The processed high-frequency components and low-frequency components are recombined to obtain the denoised signal.

[0092] Through denoising processing, the noise in sensor data can be significantly reduced, the quality of the signal can be improved, and subsequent analysis can be made more accurate and reliable.

[0093] S2.4, reconstructs the denoised low-frequency and high-frequency components to restore the main features of the original signal, so that the signal still retains its main information after denoising.

[0094] Restore the processed low-frequency components to time domain signals.

[0095] Restore the processed high-frequency components to time domain signals.

[0096] The low-frequency and high-frequency components are recombined to obtain the reconstructed signal. The specific steps include:

[0097] First, the low-frequency and high-frequency components of each layer are obtained from the wavelet decomposition process, such as the low-frequency components of the third layer and the high-frequency components of the first, second, and third layers;

[0098] The low-frequency components of the third layer are used as the initial reconstructed signal;

[0099] The low-frequency components of the third layer and the high-frequency components of the third layer are upsampled (interpolated), and convolved through a wavelet reconstruction filter, and then the convolution results are added to obtain the reconstructed signal of the second layer;

[0100] The reconstructed signal of the second layer and the high-frequency components of the second layer are upsampled (interpolated), and convolved through a wavelet reconstruction filter, and then the convolution results are added to obtain the reconstructed signal of the first layer;

[0101] The reconstructed signal of the first layer and the high-frequency components of the first layer are upsampled (interpolated), and convolved through a wavelet reconstruction filter, and then the convolution results are added to obtain the final reconstructed signal;

[0102] By combining and reconstructing layer by layer, the low-frequency and high-frequency components of all levels are combined to obtain the final reconstructed signal, which contains the main features of the original signal while removing noise and other unnecessary components.

[0103] The present invention decomposes multi-sensor data based on wavelet basis functions and decomposition layers to extract scale features and denoise high-frequency components, which can significantly improve the data quality of the road inspection system, reduce noise interference, and enhance the resolution of the signal, so that more details can be observed at high resolution and the overall trend can be grasped at low resolution. The final signal is reconstructed layer by layer to ensure the integrity and accuracy of the data after processing, providing a more reliable data basis for subsequent road status assessment and decision-making. This method not only optimizes the data processing process, but also improves the system's adaptability under complex environmental conditions, helps to timely discover and locate road defects, improve inspection efficiency and accuracy, and ultimately ensure road safety and maintenance quality.

[0104] Step S3 includes the following contents:

[0105] The logic for obtaining the multi-sensor overall fusion index is:

[0106] a1, classify the sensor data after denoising in step S2 to ensure that each set of data has a corresponding environmental label .

[0107] Indicates in the environment The sensor data set collected below.

[0108] Indicates environmental label , such as lighting conditions, weather conditions, etc.

[0109] a2, perform feature matching on each set of environmental data and calculate the time synchronization error and spatial alignment error.

[0110] ;in, Represents the feature matching function Sensor data and Perform feature matching; Indicates sensor and In the environment The feature matching matrix below.

[0111] ;in, Indicates in the environment Next, sensor and sensors The absolute value of the time difference of the collected data is calculated by and The difference between the two sensors can be obtained and In the environment Is there a time synchronization error in the data collection under Indicates sensor and In the environment The time synchronization error is as follows.

[0112] ;in, Indicates in the environment Next, sensor and sensors The spatial position difference is measured by Euclidean distance; Indicates sensor and In the environment The spatial alignment error.

[0113] a3, in order to compare different types of errors on the same scale, it is necessary to normalize the time synchronization error and the spatial alignment error. Normalization can eliminate the influence of different error magnitudes, making the fusion calculation more fair and accurate.

[0114] Normalization processing maps the errors to the same scale so that the subsequent fusion error calculation can comprehensively consider the effects of temporal and spatial errors.

[0115] The normalized time synchronization error is marked as , the normalized spatial alignment error is marked as .

[0116] a4, combine the normalized time synchronization error and spatial alignment error to calculate the error of each pair of sensors in the environment The fusion error takes into account both temporal and spatial errors and reflects the consistency of sensor data.

[0117] Each pair of sensors in the environment The fusion error under: ;

[0118] in, Indicates sensor and In the environment The fusion error is .

[0119] By calculating the fusion error, the consistency of sensor data under different environmental conditions can be quantified, providing a basis for calculating the fusion stability index.

[0120] a5, based on each pair of sensors in the environment The fusion error under different environmental conditions is calculated, and the fusion stability index of each pair of sensors under different environmental conditions is calculated. The fusion stability index reflects the stability and consistency of sensor data under specific environmental conditions.

[0121] Fusion stability index of each pair of sensors under different environmental conditions: ,in, Indicates sensor and In the environment The fusion stability index under .

[0122] a6, comprehensively consider the fusion stability index of all sensor pairs under each environmental condition and calculate the multi-sensor overall fusion index. Through multiple iterations and combinations, the multi-sensor overall fusion index is obtained: ;

[0123] in, Represents the multi-sensor overall fusion index; Indicates the total number of environmental conditions; In each environment The number of sensors used.

[0124] The multi-sensor overall fusion index is used to represent and reflect the consistency and stability of sensor data of the road inspection system under different environmental conditions. The index quantifies the fusion effect of sensor data in a variety of complex environments by comprehensively considering the time synchronization error and spatial alignment error between each pair of sensors under multiple environmental conditions. The larger the value, the better the synchronization and alignment of multi-sensor data under various environmental conditions, the more stable the system's performance in a changing environment, and the stronger its adaptability. Conversely, the smaller the value, the poorer the consistency of sensor data, the lower the stability and reliability of the system when facing different environmental conditions, and further optimization and improvement are needed to improve its adaptability and performance.

[0125] The logic for obtaining the environmental perception sensitivity index is:

[0126] b1, pre-process the multi-sensor data and extract feature information, including the original data of the sensor and environmental characteristics, including and . Ensure data quality and consistency through data cleaning, denoising and normalization.

[0127] Indicates sensor In the environment The preprocessed data set under ; Indicates sensor In the environment The following feature set; Indicates sensor In the environment The following feature set.

[0128] Preprocess the multi-sensor data and extract feature information. The specific steps include: first, clean the raw data collected by the sensor to remove noise and invalid data to ensure the quality and consistency of the data. Secondly, normalize the data and map the data of different sensors to the same scale for comparison and analysis. Then, use signal processing methods such as wavelet transform and Fourier transform to extract features from the preprocessed data, including key information such as frequency domain features, time domain features, and statistical features. In addition, combined with environmental labels, classify the data under different environmental conditions, and further extract feature information related to environmental changes, such as temperature, humidity, light intensity, etc. Through the above steps, a feature set containing multi-dimensional and multi-scale features is obtained, providing basic data for subsequent sensitivity analysis.

[0129] Preprocessing and feature extraction can reduce noise and interference, improve data quality, and provide a reliable data basis for subsequent sensitivity analysis.

[0130] b2, calculate the rate of change of sensor characteristics under different environmental conditions to evaluate the sensitivity of the sensor to environmental changes. The larger the rate of change, the higher the sensor's sensitivity to environmental changes.

[0131] ,in Indicates sensor and In the environment The characteristic change rate under .

[0132] The characteristic change rate can quantify the response degree of the sensor under different environmental conditions and reflect the sensitivity of the sensor to environmental changes.

[0133] b3, calculate the environmental difference index under different environmental conditions to evaluate the impact of environmental changes on sensor characteristics. The higher the environmental difference index, the greater the impact of environmental changes on sensor characteristics.

[0134] ,in, Indicates the environment The environmental difference index under .

[0135] The environmental difference index can quantify the impact of environmental changes on sensor characteristics and provide a reference for sensitivity analysis.

[0136] b4, normalize the environmental difference index so that it can be compared and analyzed on a unified scale. Normalization can eliminate the magnitude differences of different environmental difference indices, making the sensitivity calculation more fair and accurate.

[0137] Normalization ensures the rationality and consistency of subsequent sensitivity calculations by mapping the environmental difference index to the same scale. The normalized environmental difference index is marked as ;

[0138] b5, based on the normalized environmental difference index, calculate the environmental perception sensitivity under each environmental condition, and combine the sensitivity under all environmental conditions to calculate the environmental perception sensitivity index of the system. By combining the sensitivity under each environmental condition through the geometric mean method, a comprehensive index reflecting the overall perception ability of the system is obtained.

[0139] Environmental Sensitivity Index : ;

[0140] The Environmental Awareness Sensitivity Index (EASI) is used to represent and reflect the ability of a multi-sensor system to perceive environmental changes under different environmental conditions. It evaluates the system's response to environmental changes by quantifying the sensor feature change rate and environmental difference index. The larger the value, the higher the system's perception sensitivity under various environmental conditions, and the more accurate and rapid it can detect environmental changes; the smaller the value, the weaker the system's environmental perception ability, the slower or insensitive response to environmental changes, and further optimization is needed to improve the system's adaptability and stability under different environmental conditions. This index provides an important basis for performance evaluation and optimization of road inspection systems in complex environments.

[0141] By combining the environmental perception sensitivity index and the multi-sensor overall fusion index, the global confidence coefficient is generated to evaluate the accuracy and credibility of the data acquisition of the system during the road inspection process. For example, the calculation formula is as follows: ;

[0142] in, represents the global confidence coefficient; and They are the environmental perception sensitivity index and the multi-sensor overall fusion index respectively; and They are the preset proportional coefficients corresponding to the environmental perception sensitivity index and the multi-sensor overall fusion index, and both are greater than zero.

[0143] The global confidence coefficient is used to represent and reflect the road inspection system's ability to detect hidden problems such as data anomalies, sensor failures, and errors caused by environmental changes under different environmental conditions, and comprehensively analyze the consistency, stability, and sensitivity to environmental changes of multi-sensor data. The larger the value, the stronger the system's ability to detect various hidden problems in a complex environment, and the more accurate it can identify potential risks; the smaller the value, the weaker the system's ability to detect various hidden problems in a complex environment, and the slower response to environmental changes and data anomalies, which requires further optimization to improve the reliability and adaptability of the system.

[0144] Although step S3 has analyzed and quantified the fusion performance and environmental adaptability of multiple sensors under different environmental conditions by calculating the multi-sensor overall fusion index and environmental perception sensitivity index, this is only a preliminary evaluation. In order to gain a deeper understanding of the system's ability to detect hidden problems under different environmental conditions, step S4 needs to be further carried out. In step S4, by summarizing the global trust coefficient data under multiple environmental conditions, calculating the volatility, constructing the abnormal frequency matrix, and analyzing the pairwiseness, the stability of the system under different environmental conditions and the regularity of the occurrence of hidden problems can be identified. The pairwiseness analysis is particularly important because it can reveal whether the system's detection of hidden problems is regular and consistent, thereby providing a more comprehensive and detailed evaluation and providing an important basis for system optimization and risk management. In this way, a global hidden detection index can be generated to quantify the system's ability to detect hidden problems under different environmental conditions, ensuring the efficient operation and safety of the road inspection system in various complex environments.

[0145] Step S4 includes the following contents:

[0146] S4.1, under different environmental conditions, continuously collect the time series data of the global confidence coefficient, and record the global confidence coefficient value of the system at each time point. The global confidence coefficient value collected each time is used as part of the time series data to ensure that different time points under various environmental conditions are covered. In this way, the global confidence coefficient data over a period of time is accumulated to provide a basis for subsequent analysis.

[0147] By collecting and recording the global confidence coefficient time series data, we can fully understand the performance of the system at different time points and identify the system's ability to detect hidden problems under different environmental conditions.

[0148] S4.2, calculate the volatility of the global confidence coefficient time series data to quantify the change of the global confidence coefficient during the observation period. First, calculate the change of the global confidence coefficient between adjacent time points, that is, the difference of the global confidence coefficient between time t and t-1. Then, calculate the absolute value of these changes, sum them and divide them by the difference between the maximum change and the minimum change to obtain the volatility.

[0149] By quantifying the fluctuation of the global confidence coefficient, we can identify the stability of the system under different environmental conditions and distinguish between occasional and frequent situations. The higher the fluctuation rate, the more drastic the change of the global confidence coefficient and the lower the stability of the system.

[0150] S4.3, construct an abnormal frequency matrix to record the frequency of occurrence each time the global confidence coefficient is less than the threshold, and analyze the frequency and interval below the threshold. First, detect abnormal values ​​in the time series, that is, the situation where the global confidence coefficient is less than the preset threshold. Each time an abnormal value is detected, its time point is recorded. Then, the time interval between each pair of abnormal values ​​is calculated to construct the abnormal frequency matrix.

[0151] Through the frequency matrix, we can analyze whether the situation below the threshold occurs frequently, quantify its frequency and interval, and identify occasional and frequent situations. The construction of the frequency matrix can intuitively show the distribution and rules of outliers.

[0152] S4.4, based on the abnormal frequency matrix, calculate the parity index to quantify whether the situation where the global confidence coefficient is less than the threshold occurs regularly, reflecting the parity of the hidden problems of the system. First, perform a nonlinear transformation on each frequency value in the abnormal frequency matrix and calculate its parity index. Then, calculate the mean and variance of the frequency matrix to quantify the distribution of the frequency values.

[0153] Parallelism index : ;

[0154] Mean of the frequency matrix : ;

[0155] Variance of the frequency matrix : ;

[0156] Indicates The second abnormality and The frequency between sub-anomalies; Represents the exception counter.

[0157] By calculating the parity index, we can quantify the complexity of the system's ability to detect hidden problems under different environmental conditions and identify occasional and frequent situations. The higher the parity index, the stronger the system's ability to detect hidden problems under different environmental conditions.

[0158] S4.5, combining volatility and evenness index, generates the system's comprehensive hidden problem detection capability evaluation index. First, the volatility and evenness index are nonlinearly combined to generate an intermediate index. Then, the intermediate index is logarithmically transformed to enhance the nonlinear characteristics of the index. Finally, the global hidden detection index is calculated.

[0159] Global Hidden Detection Index : ;

[0160] in, represents volatility; Represents the mean of the pair-parity index.

[0161] By comprehensively considering the volatility and pair index of the global confidence coefficient, a global hidden inspection index is generated to reflect the stability and reliability of the system under different environmental conditions. The comprehensive evaluation index enhances the complexity and sensitivity of the index through nonlinear combination and logarithmic transformation, and can more accurately identify and deal with potential hidden dangers of the road inspection system.

[0162] The global hidden detection index is used to represent and reflect the road inspection system's ability to detect hidden problems such as data anomalies, sensor failures, and errors caused by environmental changes under different environmental conditions. By analyzing the volatility and pairing of the global confidence coefficient, the stability and regularity of the system in different environments are quantified. The larger the value, the stronger the system's ability to detect hidden problems in complex environments, the more accurately it can identify and respond to potential risks, and the system exhibits higher stability and regularity; the smaller the value, the weaker the system's ability to detect hidden problems, the slower it responds to environmental changes and data anomalies, the lower the stability and regularity of the system, and further optimization is needed to improve its reliability and adaptability.

[0163] Analyzing the strength of the ability to detect hidden problems is of great significance for the road inspection holographic perception system. By evaluating the system's ability to detect hidden problems such as data anomalies, sensor failures, and errors caused by environmental changes under different environmental conditions, it is possible to ensure the stability and reliability of the system, improve the accuracy and efficiency of road inspections, timely discover and prevent potential road safety hazards, and optimize the system's maintenance and management strategies, thereby extending the service life of road facilities, reducing maintenance costs, ensuring the safety and smoothness of road traffic, and providing solid technical support for the construction of smart cities.

[0164] The global hidden detection index and the hidden detection critical threshold are compared to distinguish the detection capabilities of the road inspection system for various hidden problems under different environmental conditions. The specific comparison results are as follows: When the global hidden detection index is greater than or equal to the hidden detection critical threshold, it indicates that the road inspection system has a strong hidden problem detection capability under the current environmental conditions, and can effectively identify errors caused by data anomalies, sensor failures and environmental changes. The system shows high stability and regularity, indicating that the road inspection system has high operating reliability in this environment and generates usable signals; when the global hidden detection index is less than the hidden detection critical threshold, it indicates that the system is insufficient in the hidden problem detection capability and may not be able to fully identify errors caused by data anomalies, sensor failures and environmental changes. The system has low stability and regularity, indicating that the road inspection system needs to be optimized and improved to improve its reliability and adaptability under complex environmental conditions and generate hidden danger signals. Through this comparison, the performance evaluation and optimization strategy of the road inspection system under different environmental conditions can be guided to improve inspection efficiency and road safety.

[0165] The present invention can comprehensively identify and quantify the hidden problem detection capability of the road inspection system under different environmental conditions through in-depth analysis and processing of the global confidence coefficient data. First, by continuously collecting the global confidence coefficient time series data and calculating its volatility, the stability of the system at each time point can be dynamically evaluated, revealing its performance differences under different environmental conditions. Secondly, by constructing an abnormal frequency matrix and calculating the evenness index, the occurrence law of hidden problems can be systematically analyzed, the complexity and frequency of these problems can be quantified, and the accuracy and comprehensiveness of the detection results can be ensured. Finally, by generating a global hidden inspection index by combining volatility and evenness index, an objective and quantitative evaluation index can be provided, which intuitively reflects the hidden problem detection capability of the system under different environmental conditions. It not only improves the recognition accuracy and response speed of hidden problems, ensures the stability and reliability of the system, but also helps to timely discover and prevent potential road safety hazards, optimize the system's maintenance and management strategies, thereby extending the service life of road facilities, reducing maintenance costs, and ensuring the safety and smoothness of road traffic. Through this comprehensive and detailed analysis, the efficiency and effect of the road inspection system in practical applications can be greatly improved, providing solid technical support for the construction of smart cities.

[0166] Figure 2 The present invention provides a road inspection holographic perception system based on artificial intelligence, including: an environment acquisition module, a feature extraction module, a performance analysis module and a hidden inspection evaluation module;

[0167] The environment acquisition module collects and constructs multi-sensor data sets under different environmental conditions, defines environmental labels, classifies and stores data and records environmental labels, and passes the collected multi-sensor data and environmental labels to the feature extraction module;

[0168] The feature extraction module decomposes the collected multi-sensor data based on the wavelet basis function and the number of decomposition layers to extract scale features, denoises the high-frequency components and reconstructs the low-frequency and high-frequency components, reconstructs the final signal layer by layer, and passes the processed feature data to the performance analysis module;

[0169] The performance analysis module collects sensor data in various environments after denoising, analyzes the time synchronization and spatial alignment errors of sensor data in different environments, measures the change rate and difference of sensor features in different environments, and comprehensively calculates the corresponding multi-sensor overall fusion index and environmental adaptability index to obtain the global confidence coefficient, so as to analyze the performance and adaptability under different environmental conditions during road inspection, and send the global confidence coefficient to the hidden inspection evaluation module;

[0170] The hidden detection assessment module summarizes the global confidence coefficient data, calculates the volatility and constructs the abnormal frequency matrix, analyzes the evenness, and comprehensively generates the global hidden detection index to quantify the hidden problem detection capability.

[0171] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0172] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

[0173] It should be noted that, in this article, if there are relational terms such as first and second, etc., they 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", "comprise" or any other variants 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. In the absence of further restrictions, the elements defined by the sentence "including a..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0174] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A road inspection holographic perception method based on artificial intelligence, characterized in that: Includes steps: S1, collects and constructs multi-sensor data sets under different environmental conditions, defines environmental labels, classifies and stores data, and records environmental labels; S2, decomposes the collected multi-sensor data based on wavelet basis functions and decomposition layers to extract scale features, denoises high-frequency components and reconstructs low-frequency and high-frequency components, and reconstructs the final signal layer by layer; S3, collects sensor data in various environments after denoising, analyzes the time synchronization and spatial alignment errors of sensor data in different environments, measures the change rate and difference of sensor features in different environments, comprehensively calculates the corresponding multi-sensor overall fusion index and environmental perception sensitivity index, and obtains the global confidence coefficient to analyze the performance and adaptability of road inspections under different environmental conditions; S4, summarizes the global confidence coefficient data, calculates the volatility and constructs the abnormal frequency matrix, analyzes the pairing, and comprehensively generates the global hidden detection index to quantify the hidden problem detection capability; Step S2 includes the following contents: S2.1, based on the selected wavelet basis function, the number of decomposition layers is set to decompose the road data collected by the multi-sensor into components of different frequencies, including low-frequency components and high-frequency components; S2.2, represents the signal at different resolutions, retaining the feature information at each scale; S2.3, use wavelet transform to denoise the high frequency components, retain the main features of the signal and remove the noise components; Perform soft threshold processing on high-frequency components, that is, the part smaller than the corresponding threshold is considered as noise and removed or weakened; Recombine the processed high-frequency components and low-frequency components to obtain a denoised signal; S2.4, reconstructing the denoised low-frequency and high-frequency components to restore the main features of the original signal; restoring the processed low-frequency components to a time domain signal; restoring the processed high-frequency components to a time domain signal; The low-frequency and high-frequency components are recombined to obtain a reconstructed signal, which specifically includes the following steps: First, the low-frequency and high-frequency components of each layer are obtained from the wavelet decomposition process, such as the low-frequency components of the third layer and the high-frequency components of the first, second, and third layers; The low-frequency components of the third layer are used as the initial reconstructed signal; The low-frequency components of the third layer and the high-frequency components of the third layer are upsampled and convolved through a wavelet reconstruction filter, and then the convolution results are added to obtain the reconstructed signal of the second layer; The reconstructed signal of the second layer and the high-frequency components of the second layer are upsampled and convolved through a wavelet reconstruction filter, and then the convolution results are added to obtain the reconstructed signal of the first layer; The reconstructed signal of the first layer and the high-frequency components of the first layer are upsampled and convolved through a wavelet reconstruction filter, and then the convolution results are added to obtain the final reconstructed signal; By combining and reconstructing layer by layer, the low-frequency and high-frequency components of all levels are combined to obtain the final reconstructed signal; Step S3 includes the following contents: The logic for obtaining the multi-sensor overall fusion index is: a1, classify the sensor data after denoising in step S2 to ensure that each set of data has a corresponding environmental label ; Indicates in the environment The sensor data set collected below; Indicates environmental label ; a2, perform feature matching on each set of environmental data and calculate the time synchronization error and spatial alignment error; ;in, Represents the feature matching function Sensor data and Perform feature matching; Indicates sensor and In the environment The feature matching matrix under ;in, Indicates in the environment Next, sensor and sensors The absolute value of the time difference of the collected data is calculated by and The difference between the two sensors can be obtained and In the environment Is there any time synchronization error in the data collection under Indicates sensor and In the environment The time synchronization error under ;in, Indicates in the environment Next, sensor and sensors The spatial position difference is measured by Euclidean distance; Indicates sensor and In the environment The spatial alignment error under a3, normalize the time synchronization error and the space alignment error, and mark the normalized time synchronization error as , the normalized spatial alignment error is marked as ; a4, combine the normalized time synchronization error and spatial alignment error to calculate the error of each pair of sensors in the environment The fusion error under: ,in, Indicates sensor and In the environment The fusion error under a5, based on each pair of sensors in the environment The fusion error under different environmental conditions is used to calculate the fusion stability index of each pair of sensors: ,in, Indicates sensor and In the environment Fusion stability index under ; a6, comprehensively consider the fusion stability index of all sensor pairs under each environmental condition, calculate the multi-sensor overall fusion index, and obtain the multi-sensor overall fusion index through multiple iterations and combinations: ,in, Represents the multi-sensor overall fusion index; Indicates the total number of environmental conditions; In each environment The number of sensors used below; The logic for obtaining the environmental perception sensitivity index is: b1, preprocess the multi-sensor data and extract feature information, including and , Indicates sensor In the environment The preprocessed data set under ; Indicates sensor In the environment The following feature set; Indicates sensor In the environment The following feature set; b2, calculate the rate of change of sensor characteristics under different environmental conditions to evaluate the sensitivity of the sensor to environmental changes; ,in Indicates sensor and In the environment The characteristic change rate under b3, calculate the environmental difference index under different environmental conditions: ,in, Indicates the environment Environmental difference index under ; b4, normalize the environmental difference index and mark the normalized environmental difference index as ; b5, based on the normalized environmental difference index, calculate the environmental perception sensitivity under each environmental condition, and combine the sensitivity under all environmental conditions to calculate the system's environmental perception sensitivity index. By combining the sensitivity under each environmental condition through the geometric mean method, the environmental perception sensitivity index is obtained. : ; By combining the environmental perception sensitivity index and the multi-sensor overall fusion index, the global confidence coefficient is generated: ,in, represents the global confidence coefficient; and They are the environmental perception sensitivity index and the multi-sensor overall fusion index respectively; and are preset proportional coefficients corresponding to the environmental perception sensitivity index and the multi-sensor overall fusion index, respectively, and both are greater than zero; Step S4 includes the following contents: S4.1, under different environmental conditions, continuously collect the time series data of the global confidence coefficient, record the global confidence coefficient value of the system at each time point, and use the global confidence coefficient value collected each time as part of the time series data to ensure that different time points under various environmental conditions are covered; S4.2, calculate the volatility of the global confidence coefficient time series data to quantify the change of the global confidence coefficient during the observation period; first, calculate the change of the global confidence coefficient between adjacent time points, that is, the difference of the global confidence coefficient between time t and t-1, then calculate the absolute value of all changes, sum and divide by the difference between the maximum change and the minimum change, so as to obtain the volatility; S4.3, construct an abnormal frequency matrix to record the frequency of occurrence of each time when the global confidence coefficient is less than the threshold, and analyze the frequency and interval below the threshold; first, detect the abnormal value in the time series, that is, the situation where the global confidence coefficient is less than the preset threshold, and record its time point each time an abnormal value is detected, then calculate the time interval between each pair of abnormal values, and construct the abnormal frequency matrix; S4.4, based on the abnormal frequency matrix, calculate the parallelism index to quantify whether the situation where the global confidence coefficient is less than the threshold occurs regularly, reflecting the parallelism of the hidden problems of the system; first, perform nonlinear transformation on each frequency value in the abnormal frequency matrix and calculate its parallelism index, then calculate the mean and variance of the frequency matrix to quantify the distribution of the frequency values; Parallelism index : ; Mean of the frequency matrix : ; Variance of the frequency matrix : ; Indicates The second abnormality and The frequency between sub-anomalies; Represents an exception counter; S4.5, combining volatility and evenness index, generates the comprehensive hidden problem detection capability evaluation index of the system. First, the volatility and evenness index are nonlinearly combined to generate the intermediate index. Then, the intermediate index is logarithmically transformed to enhance the nonlinear characteristics of the index. Finally, the global hidden detection index is calculated. : ,in, represents volatility; Represents the mean of the evenness index.

2. The method for road inspection holographic perception based on artificial intelligence according to claim 1 is characterized by: The global hidden detection index and the hidden detection critical threshold are compared to distinguish the detection capabilities of the road inspection system for various hidden problems under different environmental conditions. The specific comparison results are as follows: when the global hidden detection index is greater than or equal to the hidden detection critical threshold, an available signal is generated; when the global hidden detection index is less than the hidden detection critical threshold, a hidden danger signal is generated.

3. A road inspection holographic perception system based on artificial intelligence, used to implement the road inspection holographic perception method based on artificial intelligence as described in any one of claims 1-2, characterized in that: include: Environment acquisition module, feature extraction module, performance analysis module and hidden detection evaluation module; The environment acquisition module collects and constructs multi-sensor data sets under different environmental conditions, defines environmental labels, classifies and stores data and records environmental labels, and passes the collected multi-sensor data and environmental labels to the feature extraction module; The feature extraction module decomposes the collected multi-sensor data based on the wavelet basis function and the number of decomposition layers to extract scale features, denoises the high-frequency components and reconstructs the low-frequency and high-frequency components, reconstructs the final signal layer by layer, and passes the processed feature data to the performance analysis module; The performance analysis module collects sensor data in various environments after denoising, analyzes the time synchronization and spatial alignment errors of sensor data in different environments, measures the change rate and difference of sensor features in different environments, and comprehensively calculates the corresponding multi-sensor overall fusion index and environmental perception sensitivity index to obtain the global confidence coefficient, so as to analyze the performance and adaptability under different environmental conditions during road inspection, and send the global confidence coefficient to the hidden inspection evaluation module; The hidden detection assessment module summarizes the global confidence coefficient data, calculates the volatility and constructs the abnormal frequency matrix, analyzes the evenness, and comprehensively generates the global hidden detection index to quantify the hidden problem detection capability.

Images