A method and system for monitoring large-scale transportation on highways

Abstract

The present invention relates to the field of logistics management technology, and in particular to a method for in-transit supervision of large-scale transportation on highways. The method comprises the following steps: obtaining transportation speed data and transportation vibration data; extracting transportation speed features from the transportation speed data to obtain transportation speed feature data, and extracting transportation vibration features from the transportation vibration data to obtain transportation vibration feature data; constructing feature graphs from the transportation speed feature data and the transportation vibration feature data to obtain transportation speed vibration feature graph data; collecting mutation features and extracting the degree of change according to the transportation speed vibration feature graph data to obtain transportation mutation feature data and transportation change degree data, so as to perform large-scale transportation abnormality detection operations. The present invention effectively improves the safety and management level of the large-scale transportation process.

Description

Technical Field

[0001] The invention relates to the technical field of logistics management, and in particular to an in-transit supervision method for large-scale transportation on highways. Background Art

[0002] With the rapid development of the logistics and transportation industry, especially in the transportation of large cargo, the demand for real-time monitoring and abnormal detection of the transportation process is becoming more and more urgent. Large cargo usually has the characteristics of large volume, heavy weight, and high transportation difficulty. Traditional transportation management methods can no longer meet the needs of the modern logistics industry.

[0003] Traditional transportation monitoring methods mainly rely on manual records and regular inspections, which are not only inefficient but also difficult to achieve real-time monitoring and accurate anomaly detection. Manual records are prone to errors and cannot respond to various problems that may arise during transportation in a timely manner. Although regular inspections can detect some obvious anomalies, they are unable to detect sudden and hidden anomalies. Summary of the invention

[0004] In order to solve the above technical problems, the present invention proposes a method for in-transit supervision of large-scale transportation on highways to solve at least one of the above technical problems.

[0005] The present application provides a method for in-transit supervision of large-scale transportation on highways, the method comprising:

[0006] S1. Acquire transportation speed data and transportation vibration data;

[0007] S2, extracting the transport speed feature from the transport speed data to obtain transport speed feature data, and extracting the transport vibration feature from the transport vibration data to obtain transport vibration feature data;

[0008] S3, constructing a characteristic graph of the transport speed characteristic data and the transport vibration characteristic data to obtain transport speed vibration characteristic graph data;

[0009] S4. Collect mutation features and extract the degree of change based on the transport speed vibration characteristic diagram data to obtain transport mutation feature data and transport change degree data to perform large-scale transport anomaly detection operations.

[0010] In the present invention, the transportation speed and vibration data are obtained in real time through sensors and virtual positioning systems to ensure the timeliness and accuracy of the data. Speed ​​and vibration data are obtained at the same time to provide a comprehensive information basis for subsequent analysis. High-precision sensors and positioning systems are used to ensure the accuracy of data, which helps to improve the accuracy of subsequent feature extraction and anomaly detection. Feature extraction is performed on speed and vibration data respectively to capture multi-dimensional information of the data, including time and frequency characteristics. The extracted data features include statistical features, frequency domain features and nonlinear features, which enrich the feature space and are conducive to a more comprehensive description of the transportation status. In the feature extraction process, noise is reduced by filtering and other methods to improve the quality of feature data. Constructing speed and vibration feature data into feature graphs is helpful for data visualization and intuitive analysis. The speed and vibration data are integrated through feature graphs, which is convenient for analysis under a unified framework. The feature graph integrates the multi-dimensional feature data of speed and vibration, improves the information density, and provides a good basis for mutation feature collection and change degree extraction. By performing mutation feature collection and change degree extraction on feature graph data, abnormal changes in the transportation process can be effectively detected. Based on the mutation characteristics and degree of change data, it can accurately identify abnormal situations, such as severe vibration or speed changes during transportation, and provide early warning and take measures. Real-time monitoring and anomaly detection can help improve the safety of large-scale transportation and reduce accidents. It provides detailed mutation characteristics and degree of change data to support data-driven decision-making and optimize transportation routes and methods.

[0011] Optionally, S1 includes:

[0012] The position of the large-scale transport vehicle is collected through the virtual positioning system to obtain the position data of the large-scale transport vehicle, and the speed is calculated based on the position data of the large-scale transport vehicle to obtain the speed data of the first large-scale transport vehicle;

[0013] The acceleration data of the vehicle is collected by using an acceleration sensor preset on the large-scale transport vehicle to obtain the acceleration data of the large-scale transport vehicle, and the speed is calculated based on the acceleration data of the large-scale transport vehicle to obtain the speed data of the second large-scale transport vehicle;

[0014] Perform time series calibration and construct a speed graph based on the speed data of the first large-piece transport vehicle and the speed data of the second large-piece transport vehicle to obtain transport speed data;

[0015] The first vibration sensor preset on the large-cargo transport vehicle is used to collect vibration data of the large-cargo transport vehicle to obtain the first vibration data of the large-cargo transport vehicle, and the second vibration sensor preset on the large-cargo transport vehicle is used to collect vibration data of the large-cargo transport vehicle to obtain the second vibration data of the large-cargo transport vehicle;

[0016] A vibration graph is constructed based on the vibration data of the first large-piece transport vehicle and the vibration data of the second large-piece transport vehicle to obtain transport vibration data.

[0017] Optionally, S2 includes:

[0018] Perform time series decomposition on the transport speed data to obtain transport speed time series decomposition data;

[0019] Perform dimension reduction processing on the time series decomposition data of transportation speed to obtain dimension reduction data of transportation speed;

[0020] Perform nonlinear feature extraction on the transport speed dimension reduction data to obtain transport speed feature data;

[0021] Extract time-frequency features of transportation vibration data to obtain time-frequency vibration feature data;

[0022] Extract the Mel-frequency cepstrum coefficients of the time-frequency vibration feature data to obtain the vibration frequency domain feature data;

[0023] The spectrum diagram is constructed according to the vibration frequency domain characteristic data to obtain the transportation vibration characteristic data.

[0024] In the present invention, a virtual positioning system (such as GPS, GNSS, etc.) is used for high-precision position acquisition to ensure the accuracy of position data. Real-time monitoring of the position of the transport vehicle is achieved, providing a continuous speed data calculation basis. The speed is calculated by the position data to obtain the speed data of the first large-scale transport vehicle, ensuring the continuity and accuracy of the speed data. The speed is calculated by the data collected by the acceleration sensor as the second speed data source, providing multi-source speed data support. Real-time acquisition of acceleration data reflects the dynamic behavior of the vehicle under different road conditions and improves the sensitivity to speed changes. The speed data calculated by acceleration can supplement and verify the speed data obtained by position calculation, and improve the reliability of the overall data. Through the timing calibration technology, the speed data from different sources are aligned to eliminate the errors caused by data asynchrony. Construct a speed map to comprehensively display the vehicle speed changes and provide a complete speed data basis for subsequent feature extraction. The two speed data are integrated to obtain more accurate and reliable transportation speed data, improving the accuracy of subsequent analysis. Vibration data is collected by multiple vibration sensors to comprehensively monitor the vibration of the vehicle during transportation. Vibration sensors at different positions can capture vibration information of different parts of the vehicle, improving the comprehensiveness of the data and the ability to capture details. Multi-sensor data collection provides redundant information to prevent the impact of a single sensor failure on overall monitoring. Vibration data from different locations are integrated to construct a complete vibration map to reflect the overall vibration of the vehicle. The construction of the vibration map makes the changes in vibration data more intuitive, providing an intuitive reference for subsequent feature extraction and anomaly detection. Vibration maps can be used to detect subtle vibration changes, improving sensitivity to abnormal conditions and detection accuracy.

[0025] Optionally, S3 includes:

[0026] Normalizing the transport speed characteristic data and the transport vibration characteristic data to obtain transport speed characteristic normalized data and transport vibration characteristic normalized data;

[0027] Performing two-dimensional spatial mapping according to the normalized data of the transport speed characteristics and the normalized data of the transport vibration characteristics to obtain the two-dimensional mapping data of the transport speed characteristics and the two-dimensional mapping data of the transport speed characteristics;

[0028] Perform feature fusion according to the two-dimensional mapping data of transport speed characteristics and the two-dimensional mapping data of transport speed characteristics to obtain a transport feature fusion point data set;

[0029] The image is constructed based on the transport feature fusion point data set to obtain a preliminary transport feature map;

[0030] The preliminary transport characteristic map is enhanced to obtain the transport speed vibration characteristic map data.

[0031] In the present invention, the normalization process converts different feature data to the same scale, eliminates dimensional differences, and improves the comparability of data. After normalization, the numerical range of the data is consistent, reducing the numerical instability in the process of feature fusion and image construction. The normalization process helps to optimize the subsequent machine learning and image processing algorithms and improve their convergence speed and accuracy. Through two-dimensional space mapping, high-dimensional feature data is converted into two-dimensional space for visualization and analysis. Two-dimensional mapping can help identify and distinguish different feature patterns and improve the intuitiveness of data analysis. Two-dimensional mapping realizes the dimensionality reduction of data, while reducing the computational complexity, retaining the main features of the data. The transport speed feature and the transport vibration feature are fused to comprehensively analyze the multi-dimensional feature information. The feature fusion point data set integrates the feature information of speed and vibration, providing a more comprehensive feature description. Feature fusion reduces the complexity of processing different features separately and improves the efficiency of data processing. Through image construction, the feature data is converted into image form to intuitively display the distribution and changes of the data. The image form is convenient for discovering and analyzing patterns and anomalies in the data, and improves the effect of data analysis. The feature data from different sources are uniformly represented as images, which is convenient for subsequent processing and analysis. Through image enhancement, the clarity and contrast of the preliminary transport feature map are improved, and the image readability is enhanced. Image enhancement technology can retain and highlight important details in the image, which is conducive to more detailed analysis and judgment. The enhanced feature map is easy to identify and detect abnormal points, and improve the accuracy and timeliness of transport status monitoring.

[0032] Optionally, S4 includes:

[0033] Perform edge processing according to the transport speed vibration characteristic diagram data to obtain edge characteristic diagram data;

[0034] Performing gradient calculation on edge feature map data to obtain gradient feature map data;

[0035] Perform mutation point detection based on the gradient characteristic map data to obtain transport mutation characteristic data;

[0036] Extract the degree of change according to the transport mutation characteristic data to obtain the transport change degree data;

[0037] Large-scale transportation anomaly detection operations are carried out based on transportation mutation feature data and transportation change degree data.

[0038] The edge processing in the present invention can extract edge features in the image and highlight important structures and change areas in the feature map. Edge processing can filter out noise in the image, retain the main features of the image, and improve the clarity and availability of the data. Through edge processing, the detailed features in the image are enhanced, which is helpful for subsequent gradient calculation and mutation point detection. Gradient calculation can capture the change information in the image and identify the area with drastic changes in the image. Through gradient calculation, the contrast of the feature edge in the image is enhanced to highlight important feature information. The gradient map can simplify the data representation, making the subsequent mutation point detection more efficient and accurate. Mutation point detection can accurately identify the location where the mutation occurs in the image and capture abnormal changes in the transportation process. Through mutation point detection, possible problems in the transportation process can be discovered in advance and abnormal warnings can be provided. Mutation point detection can improve the sensitivity and detection accuracy of abnormal changes in data, which helps to take countermeasures in time. Change degree extraction can quantify the change amplitude and frequency of mutation points and provide detailed abnormal change information. Through the change degree data, the specific changes in the transportation process can be refined and analyzed to understand the degree and impact of abnormal changes. The change degree data provides quantitative data support for abnormal detection and analysis, which helps to make scientific decisions. By comprehensively utilizing mutation feature data and change degree data, we can conduct comprehensive large-scale transportation anomaly detection to ensure the accuracy and reliability of detection. We can achieve real-time monitoring and anomaly detection of large-scale transportation process to improve the efficiency and safety of transportation management. We can reduce transportation risks and ensure smooth transportation by timely discovering and preventing abnormal problems in the transportation process.

[0039] Optionally, the dimensionality reduction process includes:

[0040] Perform covariance calculation on the transport speed time series decomposition data to obtain the transport speed time series covariance matrix data;

[0041] Perform eigendecomposition on the transport speed time series covariance matrix data to obtain transport speed eigendecomposition data;

[0042] Perform principal component screening based on the transport speed characteristic decomposition data to obtain the transport speed principal component data;

[0043] According to the principal component data of transportation speed and the time series decomposition data of transportation speed, dimension reduction projection is performed to obtain the dimension reduction data of transportation speed.

[0044] The covariance matrix in the present invention can reflect the correlation between each transport speed feature, identify highly correlated or unrelated features, and lay the foundation for feature decomposition. Through the calculation of the covariance matrix, the internal structure and change trend of the data can be better understood, which is helpful for subsequent feature decomposition and principal component analysis. Covariance calculation can help identify and remove noise, extract the main change mode of the data, and improve the quality of the data. Through feature decomposition, high-dimensional data is simplified into a few eigenvectors to reduce data complexity. Feature decomposition can extract the most important characteristic components in the data, retain important information, and remove redundant features. The simplified data structure after feature decomposition is helpful for efficient subsequent calculation and analysis. Through principal component screening, the principal component that can explain the maximum variance is selected to ensure that as much information as possible of the original data is retained. Principal component screening can optimize the dimensionality reduction effect, remove noise and redundant information, and improve the signal-to-noise ratio of the data. The screened principal component data is simplified and has physical meaning, which is convenient for subsequent analysis and interpretation. Through dimensionality reduction projection, high-dimensional data is compressed into low-dimensional space to reduce storage and computing costs. Dimensionality reduction projection can retain the most important feature information while reducing data dimensions and improving computing efficiency. The data after dimensionality reduction removes redundant information and simplifies the data structure, making it easier for subsequent analysis and processing.

[0045] Optionally, the transport speed principal component data includes first transport speed principal component data and second transport speed principal component data, and the principal component screening includes:

[0046] Calculate the sum of eigenvalues ​​of the transport speed characteristic decomposition data to obtain the transport speed characteristic sum data;

[0047] The explained variance ratio is calculated based on the decomposition data of the transport speed characteristics and the sum of the transport speed characteristics to obtain the explained variance ratio data;

[0048] Sort the explained variance proportion data to obtain the explained variance sort proportion data;

[0049] The first principal component data of transportation speed is obtained by screening according to the explained variance sorting ratio data and the transportation speed characteristic decomposition data;

[0050] Sorting the transport speed characteristic decomposition data in descending order to obtain transport speed characteristic decomposition sorting data;

[0051] According to the transport speed characteristic decomposition and sorting data, the interval change rate is calculated to obtain the characteristic interval change rate data;

[0052] The maximum change rate is screened according to the characteristic interval change rate data and the transport speed characteristic decomposition sorting data to obtain the second transport speed principal component data.

[0053] The sum of eigenvalues ​​in the present invention reflects the overall variability of the entire data set, provides global information, and helps to understand the overall structure of the data. The sum of eigenvalues ​​serves as a benchmark to help determine the criteria and thresholds for subsequent feature selection. Calculating the sum of eigenvalues ​​can evaluate the contribution of each feature to the overall variation of the data, providing a basis for calculating the explained variance ratio. The explained variance ratio quantifies the contribution of each feature to the overall variation of the data, helping to identify the most important features. The explained variance ratio data provides a quantitative basis for feature screening, ensuring the scientificity and rationality of the screening process. By explaining the variance ratio, the degree of information retention after the feature is screened can be evaluated to ensure that the main information is not lost. The sorted explained variance ratio data clarifies the importance priority of each feature and provides clear guidance for subsequent feature screening. The sorting process organizes the data, which helps to analyze and understand the importance of each feature more intuitively. The sorted data makes the feature screening process more efficient and reduces the interference of irrelevant or minor features. Screening is performed according to the sorted explained variance ratio data to ensure that the selected principal component data has the largest explained variance. The screened first transport speed principal component data retains the main variability information of the data and reduces information loss. Through screening, the data structure is significantly simplified, and the efficiency of subsequent analysis and processing is improved. Descending sorting can quickly identify the most significant features in the data set for further analysis. Descending sorting arranges the data by importance, making the subsequent processing process clearer and more efficient. The sorted data clarifies the importance priority of each feature, providing a basis for subsequent change rate calculation and screening. Interval change rate calculation can capture the changing trend of feature values ​​in different intervals and identify features with significant changes. Through change rate data, the importance changes of features can be dynamically analyzed to provide more analysis dimensions. Change rate calculation helps optimize the feature screening process and ensures that the selected features have the largest information change rate. Maximum change rate screening ensures that the most significant features in different intervals are retained, enhancing the expressiveness of the data. The screened second transport speed principal component data provides a more diverse feature representation and improves the comprehensiveness of the analysis. Through maximum change rate screening, the dimensionality reduction effect of the data is further optimized, providing a higher quality data foundation for subsequent anomaly detection and analysis.

[0054] Optionally, the transport speed characteristic data includes first transport speed characteristic data and second transport speed characteristic data, and the nonlinear feature extraction includes:

[0055] Performing linear kernel matrix calculation on the first transport speed principal component data to obtain linear kernel matrix data;

[0056] Perform eigenvalue decomposition on the linear kernel matrix data to obtain linear kernel matrix decomposition data;

[0057] The linear kernel matrix decomposition data is screened by a preset characteristic threshold data and projected into the kernel principal component space to obtain the first transport speed characteristic data;

[0058] Performing Gaussian kernel matrix calculation on the first transport speed principal component data to obtain Gaussian kernel matrix data;

[0059] Perform eigenvalue decomposition on the Gaussian kernel matrix data to obtain Gaussian kernel matrix decomposition data;

[0060] Perform eigenvalue decomposition on the Gaussian kernel matrix decomposition data to obtain Gaussian kernel matrix decomposition data;

[0061] The Gaussian kernel matrix decomposition data is screened by a preset characteristic threshold data and projected into the kernel principal component space to obtain the second transport speed characteristic data.

[0062] The linear kernel calculation method in the present invention is simple and efficient, and is suitable for preliminary feature extraction. The linear kernel can effectively retain the linear relationship of the data and provide a basis for subsequent nonlinear analysis. The linear kernel matrix has low computational complexity and is suitable for rapid processing of large-scale data. Eigenvalue decomposition can extract the main characteristic components in the linear kernel matrix and simplify the data structure. Through eigenvalue decomposition, the dimension of the data is reduced and redundant information is reduced. The data representation after eigenvalue decomposition is more compact, which is convenient for subsequent processing and analysis. Through the preset feature threshold, the most important features are screened out to ensure that the main information of the data is retained. The screening and projection process further reduces the data dimension and improves the calculation efficiency. The first transport speed feature data after screening is more optimized, which is convenient for subsequent analysis and processing. The Gaussian kernel can effectively capture the nonlinear relationship in the data and make up for the shortcomings of the linear kernel. The Gaussian kernel parameters can mine the periodic laws in the principal component data of the first transport speed. Since the generation process of the principal component data of the first transport speed is similar to the elbow law, the data performance is relatively stable. The Gaussian kernel matrix can provide richer feature representation and improve the accuracy of data analysis. The eigenvalue decomposition can extract the main nonlinear characteristic components in the Gaussian kernel matrix and enhance the data representation capability. Through eigenvalue decomposition, the data structure is simplified and the main nonlinear features are retained. The data representation after eigenvalue decomposition is more compact, which is convenient for subsequent processing and analysis. The most important nonlinear features are screened out through the preset feature threshold to ensure that the main information of the data is retained. The screening and projection process enhances the diversity of feature representation and improves the comprehensiveness of data analysis. The screening and projection process further reduces the data dimension and improves computational efficiency.

[0063] Optionally, mutation point detection includes:

[0064] Generate an initial mutation feature set according to the transport mutation feature data to obtain initial mutation feature set data;

[0065] According to the transport mutation feature data and the initial mutation feature set data, a primary calculation is performed through the mutation feature calculation function to obtain the primary mutation feature set data; wherein the mutation feature calculation function is specifically:

[0066] ;

[0067] is the primary mutation signature set data, For transport mutation characteristic quantity data, is the transport mutation characteristic sequence data, To characterize mutation intensity data for transport mutations, For the Transport mutation feature data, is the transport mutation intensity error adjustment term, is the transport mutation intensity error adjustment term, is the weighted data of the transport mutation characteristic amplitude coefficient, is the characteristic amplitude coefficient data of transport mutation, is the weight data of transport mutation characteristic change, It is the periodic weight data of transport mutation characteristics;

[0068] Recursively split the primary mutation feature set data to obtain secondary mutation feature set data;

[0069] The recursive segmentation is as follows:

[0070] ;

[0071] is the secondary mutation feature set data, is the first location data to be placed, is the second location data to be placed, Data for the primary mutation signature set is at position The minimum cost data, Data for primary mutation signature collection from position To location The minimum cost data, is the penalty coefficient data;

[0072] The secondary mutation feature set data is pruned to obtain the transport mutation feature data; the pruning operation is specifically as follows:

[0073] ;

[0074] For transport mutation feature data, is the first location data to be placed, Data for the primary mutation signature set is at position The minimum cost data, Data for primary mutation signature collection from position To location The minimum cost data, is the penalty coefficient data, It is the secondary mutation feature set data.

[0075] In the present invention, all transport mutation feature data are integrated through the generation of the initial feature set, laying the foundation for subsequent analysis. The generation of the initial feature set can be regarded as a preprocessing of the original data, removing some noise and improving the data quality. Through the reasonable generation of the feature set, the data organization method is optimized and the efficiency of subsequent processing is improved. Through the mutation feature calculation function, multiple features (such as amplitude, rate of change, and periodicity) are integrated to extract primary features to ensure the multi-dimensional representation of the data. The calculation results provide quantitative feature data, which provides an accurate data basis for subsequent recursive segmentation and pruning operations. Through recursive segmentation, the optimal segmentation point between different positions in the primary feature set is found to ensure the optimality of the segmentation result. The minimum cost is calculated during the recursive segmentation process to ensure that the cost of calculation and processing is reduced while ensuring the quality of the features. The segmentation process optimizes the structure of the feature set, making the data easier to analyze and process. Through the pruning operation, insignificant or redundant feature data is removed to improve the simplicity and effectiveness of the data set. The pruning operation reduces the amount of data, improves the efficiency of subsequent processing, and reduces the computational complexity. It ensures that important mutation features are retained, and improves the accuracy and reliability of anomaly detection.

[0076] Compared with the traditional cost function which usually only considers a single dimension or simple linear features, the mutation feature calculation function adopted in the present invention comprehensively considers multiple dimensions and nonlinear features. Taking into account the mutation intensity and its adjustment terms, it can more sensitively reflect the changes in mutation intensity. This item takes into account the logarithmic change of the amplitude, which can capture the impact of smaller amplitude changes and compress the change amplitude through a logarithmic function, making the mutation characteristics more stable. This item introduces the periodic change of mutation characteristics, and reflects the periodic change rate of mutation characteristics through derivatives, which can more accurately capture the periodic change characteristics of mutations. Compared with the traditional cost function, the calculation formula of the present invention can detect mutation characteristics more sensitively. The various processing in the formula can amplify small changes, and through the comprehensive consideration of multi-dimensional characteristics, the detection is more accurate and reliable. This highly sensitive mutation detection is of great significance for the early detection of potential problems and anomalies in the transportation process.

[0077] Optionally, the present application further provides a highway large-scale transport in-transit supervision system, which is used to execute the highway large-scale transport in-transit supervision method as described above, and the highway large-scale transport in-transit supervision system includes:

[0078] Large-scale transport vehicle supervision data collection module, used to obtain transport speed data and transport vibration data;

[0079] The transport supervision data feature extraction module is used to extract the transport speed feature of the transport speed data to obtain the transport speed feature data, and to extract the transport vibration feature of the transport vibration data to obtain the transport vibration feature data;

[0080] A transport speed vibration frequency map construction module is used to construct a feature map of the transport speed characteristic data and the transport vibration characteristic data to obtain transport speed vibration characteristic map data, and to perform frequency map conversion on the transport speed vibration characteristic map data to obtain transport speed vibration frequency map data;

[0081] The large-piece transport anomaly detection module is used to perform signal decomposition on the transport speed vibration frequency graph data to obtain frequency graph decomposition data, and to perform frequency feature extraction on the frequency graph decomposition data to obtain frequency graph feature data for large-piece transport anomaly detection operations.

[0082] The purpose of the present invention is to obtain the transport speed and vibration data in real time through a virtual positioning system and an acceleration sensor to ensure the comprehensiveness and timeliness of the data. This lays a solid foundation for subsequent feature extraction and analysis. Using a positioning system and a sensor to ensure the accuracy of the data is helpful to improve the accuracy of subsequent feature extraction and anomaly detection. Combining the data of the virtual positioning system and the acceleration sensor, the ability of multi-source data fusion is provided. By comprehensively analyzing data from different sources, more comprehensive and reliable transport status information can be obtained. Through time series decomposition, dimensionality reduction processing and nonlinear feature extraction, the characteristics of the transport speed and vibration data are fully extracted, including time characteristics, frequency characteristics and nonlinear characteristics. These multidimensional features provide rich information for the subsequent feature map construction. By using signal processing technology (such as short-time Fourier transform, Mel frequency cepstral coefficient extraction, etc.), the expression ability of the features is enhanced and the description ability of complex transport states is improved. In the feature extraction process, the noise in the data is removed by dimensionality reduction and screening, the main feature information is retained, and the data quality is improved. By constructing the speed and vibration feature data into a feature map, the visual display of the data is realized. The feature map can intuitively display the state changes during transportation and provide a clear image basis for anomaly detection. The feature graph combines speed and vibration data in a unified framework for analysis, providing a more comprehensive perspective and helping to more accurately identify and analyze abnormal situations. Feature graph construction simplifies the representation of complex data, making subsequent processing and analysis more efficient. The feature graph is fast to build and process, and can meet the needs of real-time monitoring. Through mutation feature collection and change degree extraction, mutation points in the transportation process can be accurately identified and significant changes in transportation status can be captured. This accurate identification capability is crucial for timely detection and handling of abnormal situations. Using algorithms (such as eigenvalue decomposition, change rate calculation, etc.), multi-dimensional change analysis of mutation features is performed to provide more detailed and accurate abnormal change information. Systematically combining mutation features and change degree data for real-time anomaly detection ensures timely warning and response when abnormal situations occur, improving the safety and reliability of the transportation process. Detailed mutation feature and change degree data provide strong data support for transportation managers, assisting in making scientific decisions, optimizing transportation routes and methods, and improving transportation efficiency and safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] Other features, objects and advantages of the present application will become more apparent by reading the detailed description of non-limiting implementations made with reference to the following drawings:

[0084] Figure 1 A flowchart showing a method for supervising in-transit transportation of large items on highways according to an embodiment of the present invention is provided;

[0085] Figure 2A flowchart showing a method for collecting data on large-scale transport vehicles under one embodiment is provided;

[0086] Figure 3 A flowchart showing a method for extracting features from transport supervision data according to an embodiment of the present invention is shown;

[0087] Figure 4 A flow chart showing the steps of a method for constructing a transport speed vibration frequency diagram according to an embodiment of the present invention is shown;

[0088] Figure 5 A flowchart showing a method for detecting abnormalities in large-scale transportation according to an embodiment of the present invention is provided;

[0089] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0090] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are 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 technicians in this field without creative work are within the scope of protection of the present invention.

[0091] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0092] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0093] See also Figures 1 to 5 The present application provides a method for supervising the in-transit transportation of large items on highways, the method comprising:

[0094] S1. Acquire transportation speed data and transportation vibration data;

[0095] S2, extracting the transport speed feature from the transport speed data to obtain transport speed feature data, and extracting the transport vibration feature from the transport vibration data to obtain transport vibration feature data;

[0096] S3, constructing a characteristic graph of the transport speed characteristic data and the transport vibration characteristic data to obtain transport speed vibration characteristic graph data;

[0097] S4. Collect mutation features and extract the degree of change based on the transport speed vibration characteristic diagram data to obtain transport mutation feature data and transport change degree data to perform large-scale transport anomaly detection operations.

[0098] The present invention provides an embodiment, using GPS or GNSS positioning system to collect vehicle position and calculate vehicle speed. Acceleration sensors installed on the vehicle collect vehicle acceleration data in real time and calculate vehicle speed. Vibration sensors are installed at different positions of the vehicle to collect vehicle vibration data in real time.

[0099] The location information of the vehicle is obtained through the virtual positioning system, and the speed between adjacent position points is calculated to obtain the speed data of the first large-scale transport vehicle. The acceleration data is obtained through the acceleration sensor, and the integration calculation is performed to obtain the speed data of the second large-scale transport vehicle. The first and second speed data are time-series calibrated to eliminate the error caused by time asynchrony and construct unified speed data. Vibration data is collected through multiple vibration sensors, and data fusion is performed to obtain complete transportation vibration data.

[0100] Decompose the transport speed data in time series to extract the main components and noise components. Use PCA and other dimensionality reduction techniques to reduce the dimensionality of the data after time series decomposition and retain the main features. Use Kernel PCA and other methods to extract nonlinear features to obtain richer feature data. Perform time-frequency analysis on vibration data to extract time-frequency features. Calculate the Mel-frequency cepstrum coefficients of the vibration signal to obtain frequency domain features.

[0101] The covariance of the transport speed data is calculated to obtain the covariance matrix. The covariance matrix is ​​subjected to eigendecomposition to extract the main components and noise components. According to the eigendecomposition results, the main components are screened to retain the main transport speed characteristics. The transport speed dimension reduction data is obtained through dimensionality reduction projection. The linear kernel and Gaussian kernel matrix are calculated for the reduced-dimensional data, and the eigenvalue decomposition and screening are performed respectively to obtain the first and second transport speed characteristic data. The transport vibration data is subjected to short-time Fourier transform to obtain the time-frequency characteristic data. The Mel-frequency cepstrum coefficients of the time-frequency characteristic data are calculated to obtain the vibration frequency domain characteristic data.

[0102] Normalize the transport speed feature and transport vibration feature data. Perform two-dimensional spatial mapping on the normalized feature data. Fuse the two-dimensionally mapped speed and vibration features to obtain a feature fusion point data set. Perform image construction based on the feature fusion point data set to generate a preliminary transport feature map. Perform image enhancement on the preliminary transport feature map.

[0103] Normalize the transport speed and vibration feature data to ensure that the data is compared and analyzed on the same scale. Use methods such as t-SNE or UMAP to map the normalized feature data to a two-dimensional space for visualization. Fuse the speed and vibration feature points in the two-dimensional space to form a feature fusion point data set. Construct a transport feature map based on the feature fusion point data set, and display the distribution and changes of the feature data in the form of a point cloud map or a heat map. Perform image processing operations such as edge detection and contrast enhancement on the preliminary transport feature map to improve the clarity and readability of the image.

[0104] Perform edge detection on the feature map data to extract edge features. Perform gradient calculation on the edge feature map data to obtain gradient feature map data. Perform mutation point detection based on the gradient feature map data. Extract the degree of change based on the mutation feature data. Perform anomaly detection based on the mutation feature data and the degree of change data.

[0105] Apply the Canny edge detection algorithm to perform edge detection on the transport speed vibration feature map data to obtain edge feature map data. Perform Sobel gradient calculation on the edge feature map data to obtain gradient feature map data, which reflects the area with the most dramatic changes in the feature map. Use the Pelt algorithm to detect mutation points on the gradient feature map data to identify mutation points that occur during transportation. Calculate the change amplitude and frequency based on the detected mutation points, and extract the degree of change data. Comprehensively analyze the mutation feature data and degree of change data (use a visual method to display the data, or perform further data analysis, such as calculation or processing using a preset algorithm), perform anomaly detection, identify abnormal conditions in the transportation process, and generate warning information.

[0106] In the present invention, the transportation speed and vibration data are obtained in real time through sensors and virtual positioning systems to ensure the timeliness and accuracy of the data. Speed ​​and vibration data are obtained at the same time to provide a comprehensive information basis for subsequent analysis. High-precision sensors and positioning systems are used to ensure the accuracy of data, which helps to improve the accuracy of subsequent feature extraction and anomaly detection. Feature extraction is performed on speed and vibration data respectively to capture multi-dimensional information of the data, including time and frequency characteristics. The extracted data features include statistical features, frequency domain features and nonlinear features, which enrich the feature space and are conducive to a more comprehensive description of the transportation status. In the feature extraction process, noise is reduced by filtering and other methods to improve the quality of feature data. Constructing speed and vibration feature data into feature graphs is helpful for data visualization and intuitive analysis. The speed and vibration data are integrated through feature graphs, which is convenient for analysis under a unified framework. The feature graph integrates the multi-dimensional feature data of speed and vibration, improves the information density, and provides a good basis for mutation feature collection and change degree extraction. By performing mutation feature collection and change degree extraction on feature graph data, abnormal changes in the transportation process can be effectively detected. Based on the mutation characteristics and degree of change data, it can accurately identify abnormal situations, such as severe vibration or speed changes during transportation, and provide early warning and take measures. Real-time monitoring and anomaly detection can help improve the safety of large-scale transportation and reduce accidents. It provides detailed mutation characteristics and degree of change data to support data-driven decision-making and optimize transportation routes and methods.

[0107] Optionally, S1 includes:

[0108] The position of the large-scale transport vehicle is collected through the virtual positioning system to obtain the position data of the large-scale transport vehicle, and the speed is calculated based on the position data of the large-scale transport vehicle to obtain the speed data of the first large-scale transport vehicle;

[0109] The acceleration data of the vehicle is collected by using an acceleration sensor preset on the large-scale transport vehicle to obtain the acceleration data of the large-scale transport vehicle, and the speed is calculated based on the acceleration data of the large-scale transport vehicle to obtain the speed data of the second large-scale transport vehicle;

[0110] Perform time series calibration and construct a speed graph based on the speed data of the first large-piece transport vehicle and the speed data of the second large-piece transport vehicle to obtain transport speed data;

[0111] The first vibration sensor preset on the large-cargo transport vehicle is used to collect vibration data of the large-cargo transport vehicle to obtain the first vibration data of the large-cargo transport vehicle, and the second vibration sensor preset on the large-cargo transport vehicle is used to collect vibration data of the large-cargo transport vehicle to obtain the second vibration data of the large-cargo transport vehicle;

[0112] A vibration graph is constructed based on the vibration data of the first large-piece transport vehicle and the vibration data of the second large-piece transport vehicle to obtain transport vibration data.

[0113] The present invention provides an embodiment, during the transportation process, the vehicle collects location data including longitude, latitude and time stamp at regular time intervals (such as every second) through a GPS device installed on the vehicle. Based on the continuous location information, the distance and time difference between adjacent location points are calculated, thereby calculating the speed.

[0114] Install a high-precision acceleration sensor on the vehicle. Calculate the speed by integrating the acceleration data. The acceleration sensor collects the acceleration data of the vehicle in real time and records the acceleration value at each time point. Calculate the speed by integrating the acceleration data.

[0115] Synchronize the two sets of speed data to eliminate the error caused by time asynchrony. Visualize the calibrated speed data and build a speed graph. Synchronize the two sets of speed data according to the timestamp to ensure that there are corresponding speed values ​​at the same time point. Visualize the synchronized speed data and build a speed graph.

[0116] Vibration sensors are installed at different locations on the vehicle to collect vibration data in real time. Vibration sensors installed at different locations on the vehicle collect vibration data during transportation in real time and record the vibration value at each time point.

[0117] Fuse the two sets of vibration data. Visualize the fused vibration data and construct a vibration map. Fuse the data collected by the two vibration sensors to obtain more comprehensive vibration data. Visualize the fused vibration data and construct a vibration map.

[0118] In the present invention, the transportation speed and vibration data are obtained in real time through sensors and virtual positioning systems to ensure the timeliness and accuracy of the data. Speed ​​and vibration data are obtained at the same time to provide a comprehensive information basis for subsequent analysis. High-precision sensors and positioning systems are used to ensure the accuracy of data, which helps to improve the accuracy of subsequent feature extraction and anomaly detection. Feature extraction is performed on speed and vibration data respectively to capture multi-dimensional information of the data, including time and frequency characteristics. The extracted data features include statistical features, frequency domain features and nonlinear features, which enrich the feature space and are conducive to a more comprehensive description of the transportation status. In the feature extraction process, noise is reduced by filtering and other methods to improve the quality of feature data. Constructing speed and vibration feature data into feature graphs is helpful for data visualization and intuitive analysis. The speed and vibration data are integrated through feature graphs, which is convenient for analysis under a unified framework. The feature graph integrates the multi-dimensional feature data of speed and vibration, improves the information density, and provides a good basis for mutation feature collection and change degree extraction. By performing mutation feature collection and change degree extraction on feature graph data, abnormal changes in the transportation process can be effectively detected. Based on the mutation characteristics and degree of change data, it can accurately identify abnormal situations, such as severe vibration or speed changes during transportation, and provide early warning and take measures. Real-time monitoring and anomaly detection can help improve the safety of large-scale transportation and reduce accidents. It provides detailed mutation characteristics and degree of change data to support data-driven decision-making and optimize transportation routes and methods.

[0119] Optionally, S2 includes:

[0120] Perform time series decomposition on the transport speed data to obtain transport speed time series decomposition data;

[0121] Perform dimension reduction processing on the time series decomposition data of transportation speed to obtain dimension reduction data of transportation speed;

[0122] Perform nonlinear feature extraction on the transport speed dimension reduction data to obtain transport speed feature data;

[0123] Extract time-frequency features of transportation vibration data to obtain time-frequency vibration feature data;

[0124] Extract the Mel-frequency cepstrum coefficients of the time-frequency vibration feature data to obtain the vibration frequency domain feature data;

[0125] The spectrum diagram is constructed according to the vibration frequency domain characteristic data to obtain the transportation vibration characteristic data.

[0126] The present invention provides an embodiment, which uses methods such as EMD (empirical mode decomposition) or SSA (singular spectrum analysis) to perform time series decomposition on transportation speed data. The transportation speed data is decomposed using EMD to obtain a number of intrinsic mode functions (IMFs). The transportation speed data is decomposed using SSA to extract main components and trends.

[0127] Use PCA (Principal Component Analysis) to reduce the dimension. Perform PCA dimensionality reduction on the IMF data after time series decomposition and retain the main characteristic components.

[0128] Use Kernel PCA (kernel principal component analysis) to extract nonlinear features. Perform Kernel PCA on the reduced-dimensional data to extract nonlinear features.

[0129] Use short-time Fourier transform (STFT) or continuous wavelet transform (CWT) to extract time-frequency features. Perform short-time Fourier transform on vibration data to extract time-frequency features. Use wavelet transform to extract time-frequency features.

[0130] Calculate the Mel-frequency cepstral coefficient (MFCC) to extract frequency domain features. Perform MFCC calculation on the time-frequency feature data to obtain frequency domain features.

[0131] Construct a spectrogram based on frequency domain feature data such as MFCC. Visualize MFCC data and construct a spectrogram.

[0132] In the present invention, a virtual positioning system (such as GPS, GNSS, etc.) is used for high-precision position acquisition to ensure the accuracy of position data. Real-time monitoring of the position of the transport vehicle is achieved, providing a continuous speed data calculation basis. The speed is calculated by the position data to obtain the speed data of the first large-scale transport vehicle, ensuring the continuity and accuracy of the speed data. The speed is calculated by the data collected by the acceleration sensor as the second speed data source, providing multi-source speed data support. Real-time acquisition of acceleration data reflects the dynamic behavior of the vehicle under different road conditions and improves the sensitivity to speed changes. The speed data calculated by acceleration can supplement and verify the speed data obtained by position calculation, and improve the reliability of the overall data. Through the timing calibration technology, the speed data from different sources are aligned to eliminate the errors caused by data asynchrony. Construct a speed map to comprehensively display the vehicle speed changes and provide a complete speed data basis for subsequent feature extraction. The two speed data are integrated to obtain more accurate and reliable transportation speed data, improving the accuracy of subsequent analysis. Vibration data is collected by multiple vibration sensors to comprehensively monitor the vibration of the vehicle during transportation. Vibration sensors at different positions can capture vibration information of different parts of the vehicle, improving the comprehensiveness of the data and the ability to capture details. Multi-sensor data collection provides redundant information to prevent the impact of a single sensor failure on overall monitoring. Vibration data from different locations are integrated to construct a complete vibration map to reflect the overall vibration of the vehicle. The construction of the vibration map makes the changes in vibration data more intuitive, providing an intuitive reference for subsequent feature extraction and anomaly detection. Vibration maps can be used to detect subtle vibration changes, improving sensitivity to abnormal conditions and detection accuracy.

[0133] Optionally, S3 includes:

[0134] Normalizing the transport speed characteristic data and the transport vibration characteristic data to obtain transport speed characteristic normalized data and transport vibration characteristic normalized data;

[0135] Performing two-dimensional spatial mapping according to the normalized data of the transport speed characteristics and the normalized data of the transport vibration characteristics to obtain the two-dimensional mapping data of the transport speed characteristics and the two-dimensional mapping data of the transport speed characteristics;

[0136] Perform feature fusion according to the two-dimensional mapping data of transport speed characteristics and the two-dimensional mapping data of transport speed characteristics to obtain a transport feature fusion point data set;

[0137] The image is constructed based on the transport feature fusion point data set to obtain a preliminary transport feature map;

[0138] The preliminary transport characteristic map is enhanced to obtain the transport speed vibration characteristic map data.

[0139] The present invention provides an embodiment, which uses standardization or min-max scaling to perform normalization processing. The data is standardized with a mean of 0 and a standard deviation of 1. The data is scaled with a minimum value of 0 and a maximum value of 1.

[0140] Use dimensionality reduction techniques such as t-SNE or UMAP for two-dimensional space mapping. Use t-SNE to map high-dimensional feature data to two-dimensional space. Use UMAP to map high-dimensional feature data to two-dimensional space.

[0141] The two-dimensional mapping results of the velocity features and vibration features are combined. The velocity features and vibration features after two-dimensional mapping are concatenated or weighted averaged to obtain a fused feature point data set.

[0142] Use point cloud map, heat map and other technologies to construct the image. Draw the fused feature point data set into a point cloud map. Draw the fused feature point data set into a heat map.

[0143] Use edge detection, contrast enhancement and other techniques to enhance the image. Apply the Canny edge detection algorithm to enhance the edge features in the image. Adjust the image contrast to improve the clarity and readability of the image. Use the Canny edge detection algorithm to extract edge features. Overlay the edge detection results with the original image to enhance the edges. Adjust the image contrast and brightness to further enhance the image clarity.

[0144] In the present invention, the normalization process converts different feature data to the same scale, eliminates dimensional differences, and improves the comparability of data. After normalization, the numerical range of the data is consistent, reducing the numerical instability in the process of feature fusion and image construction. The normalization process helps to optimize the subsequent machine learning and image processing algorithms and improve their convergence speed and accuracy. Through two-dimensional space mapping, high-dimensional feature data is converted into two-dimensional space for visualization and analysis. Two-dimensional mapping can help identify and distinguish different feature patterns and improve the intuitiveness of data analysis. Two-dimensional mapping realizes the dimensionality reduction of data, while reducing the computational complexity, retaining the main features of the data. The transport speed feature and the transport vibration feature are fused to comprehensively analyze the multi-dimensional feature information. The feature fusion point data set integrates the feature information of speed and vibration, providing a more comprehensive feature description. Feature fusion reduces the complexity of processing different features separately and improves the efficiency of data processing. Through image construction, the feature data is converted into image form to intuitively display the distribution and changes of the data. The image form is convenient for discovering and analyzing patterns and anomalies in the data, and improves the effect of data analysis. The feature data from different sources are uniformly represented as images, which is convenient for subsequent processing and analysis. Through image enhancement, the clarity and contrast of the preliminary transport feature map are improved, and the image readability is enhanced. Image enhancement technology can retain and highlight important details in the image, which is conducive to more detailed analysis and judgment. The enhanced feature map is easy to identify and detect abnormal points, and improve the accuracy and timeliness of transport status monitoring.

[0145] Optionally, S4 includes:

[0146] Perform edge processing according to the transport speed vibration characteristic diagram data to obtain edge characteristic diagram data;

[0147] Performing gradient calculation on edge feature map data to obtain gradient feature map data;

[0148] Perform mutation point detection based on the gradient characteristic map data to obtain transport mutation characteristic data;

[0149] Extract the degree of change according to the transport mutation characteristic data to obtain the transport change degree data;

[0150] Large-scale transportation anomaly detection operations are carried out based on transportation mutation feature data and transportation change degree data.

[0151] The present invention provides an embodiment, which uses the Canny edge detection algorithm to perform edge processing. The Canny edge detection algorithm is applied to process the transport speed vibration characteristic graph data to extract edge features in the image.

[0152] Use the Sobel operator to calculate the gradient. Apply the Sobel operator to perform gradient calculation on the edge feature map data and extract the gradient features in the image. The Sobel operator is used to calculate the gradient of an image to extract the edge features in the image. The Sobel operator obtains the horizontal and vertical edge information of the image by calculating the local gradient of the image grayscale value. # Use the Sobel operator to calculate the horizontal gradient grad_x = cv2.Sobel(blurred_image, cv2.CV_64F, 1, 0, ksize=3); # Use the Sobel operator to calculate the vertical gradient grad_y = cv2.Sobel(blurred_image, cv2.CV_64F, 0, 1, ksize=3). import numpy as np# Calculate the gradient amplitude # Convert to 8-bit image.

[0153] Use Pelt algorithm for mutation point detection. Apply Pelt algorithm to gradient feature map data to detect mutation points in the image. Pelt (Pruned Exact Linear Time) algorithm is an efficient mutation point detection algorithm used to detect mutation points in time series data. It optimizes the objective function to find the optimal mutation point set, and improves the efficiency of the algorithm through pruning operations during the processing. The cost function is used to measure the difference between the data before and after the segmentation point, and the mean squared error (MSE) is used as the cost function. Initialize the parameters of the Pelt algorithm, including the penalty coefficient and the initial point. Set the position of the first mutation point to 0, initialize the cost function array and the mutation point position array. For each possible segmentation point, calculate the cost from the starting point to the current point, and find the optimal segmentation point set. The Pelt algorithm performs pruning operations during the calculation process to reduce unnecessary calculations and improve the efficiency of the algorithm. The Pelt algorithm will output an optimal mutation point set, indicating the significant change points in the data.

[0154] Calculate the change range and change rate of the data before and after the mutation point. Calculate the difference of the data before and after each mutation point to get the change range. Calculate the ratio of the change range to get the change rate.

[0155] Combined analysis of mutation feature data and change degree data for anomaly detection. Based on mutation feature data and change degree data, set thresholds or rules for anomaly detection. After identifying anomalies, generate warning information.

[0156] The edge processing in the present invention can extract edge features in the image and highlight important structures and change areas in the feature map. Edge processing can filter out noise in the image, retain the main features of the image, and improve the clarity and availability of the data. Through edge processing, the detailed features in the image are enhanced, which is helpful for subsequent gradient calculation and mutation point detection. Gradient calculation can capture the change information in the image and identify the area with drastic changes in the image. Through gradient calculation, the contrast of the feature edge in the image is enhanced to highlight important feature information. The gradient map can simplify the data representation, making the subsequent mutation point detection more efficient and accurate. Mutation point detection can accurately identify the location where the mutation occurs in the image and capture abnormal changes in the transportation process. Through mutation point detection, possible problems in the transportation process can be discovered in advance and abnormal warnings can be provided. Mutation point detection can improve the sensitivity and detection accuracy of abnormal changes in data, which helps to take countermeasures in time. Change degree extraction can quantify the change amplitude and frequency of mutation points and provide detailed abnormal change information. Through the change degree data, the specific changes in the transportation process can be refined and analyzed to understand the degree and impact of abnormal changes. The change degree data provides quantitative data support for abnormal detection and analysis, which helps to make scientific decisions. By comprehensively utilizing mutation feature data and change degree data, we can conduct comprehensive large-scale transportation anomaly detection to ensure the accuracy and reliability of detection. We can achieve real-time monitoring and anomaly detection of large-scale transportation process to improve the efficiency and safety of transportation management. We can reduce transportation risks and ensure smooth transportation by timely discovering and preventing abnormal problems in the transportation process.

[0157] Optionally, the dimensionality reduction process includes:

[0158] Perform covariance calculation on the transport speed time series decomposition data to obtain the transport speed time series covariance matrix data;

[0159] Perform eigendecomposition on the transport speed time series covariance matrix data to obtain transport speed eigendecomposition data;

[0160] Perform principal component screening based on the transport speed characteristic decomposition data to obtain the transport speed principal component data;

[0161] According to the principal component data of transportation speed and the time series decomposition data of transportation speed, dimension reduction projection is performed to obtain the dimension reduction data of transportation speed.

[0162] The present invention provides an embodiment to calculate the covariance matrix of the transport speed time series decomposition data to reflect the linear relationship between the various features. The transport speed time series decomposition data is organized into a matrix form, each row is a feature value at a time. The covariance matrix is ​​calculated using a covariance function.

[0163] Perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and eigenvectors. Use the eigenvalue decomposition function to decompose the covariance matrix to obtain eigenvalues ​​and eigenvectors.

[0164] The main principal components are selected based on the size of the eigenvalues. Usually, the eigenvectors that can explain most of the variance are selected. The proportion of each eigenvalue in the total variance is calculated. The eigenvectors are sorted by the proportion of explained variance, and the eigenvectors whose cumulative explained variance exceeds a certain threshold (such as 95%) are selected as principal components.

[0165] The covariance matrix in the present invention can reflect the correlation between each transport speed feature, identify highly correlated or unrelated features, and lay the foundation for feature decomposition. Through the calculation of the covariance matrix, the internal structure and change trend of the data can be better understood, which is helpful for subsequent feature decomposition and principal component analysis. Covariance calculation can help identify and remove noise, extract the main change mode of the data, and improve the quality of the data. Through feature decomposition, high-dimensional data is simplified into a few eigenvectors to reduce data complexity. Feature decomposition can extract the most important characteristic components in the data, retain important information, and remove redundant features. The simplified data structure after feature decomposition is helpful for efficient subsequent calculation and analysis. Through principal component screening, the principal component that can explain the maximum variance is selected to ensure that as much information as possible of the original data is retained. Principal component screening can optimize the dimensionality reduction effect, remove noise and redundant information, and improve the signal-to-noise ratio of the data. The screened principal component data is simplified and has physical meaning, which is convenient for subsequent analysis and interpretation. Through dimensionality reduction projection, high-dimensional data is compressed into low-dimensional space to reduce storage and computing costs. Dimensionality reduction projection can retain the most important feature information while reducing data dimensions and improving computing efficiency. The data after dimensionality reduction removes redundant information and simplifies the data structure, making it easier for subsequent analysis and processing.

[0166] Optionally, the transport speed principal component data includes first transport speed principal component data and second transport speed principal component data, and the principal component screening includes:

[0167] Calculate the sum of eigenvalues ​​of the transport speed characteristic decomposition data to obtain the transport speed characteristic sum data;

[0168] The explained variance ratio is calculated based on the decomposition data of the transport speed characteristics and the sum of the transport speed characteristics to obtain the explained variance ratio data;

[0169] Sort the explained variance proportion data to obtain the explained variance sort proportion data;

[0170] The first principal component data of transportation speed is obtained by screening according to the explained variance sorting ratio data and the transportation speed characteristic decomposition data;

[0171] Sorting the transport speed characteristic decomposition data in descending order to obtain transport speed characteristic decomposition sorting data;

[0172] According to the transport speed characteristic decomposition and sorting data, the interval change rate is calculated to obtain the characteristic interval change rate data;

[0173] The maximum change rate is screened according to the characteristic interval change rate data and the transport speed characteristic decomposition sorting data to obtain the second transport speed principal component data.

[0174] The present invention provides an embodiment for calculating the sum of all eigenvalues ​​in the eigendecomposition data. The eigenvalues ​​in the eigendecomposition data are summed to obtain the eigenvalue sum.

[0175] Calculate the proportion of the total variance accounted for by each eigenvalue. Divide each eigenvalue by the sum of the eigenvalues ​​to get the proportion of variance explained.

[0176] Sort by explained variance ratio from large to small. Sort the explained variance ratio data in descending order to obtain sorted explained variance ratio data.

[0177] The main eigenvectors are selected as the first principal component data according to the explained variance ratio after sorting. The eigenvectors whose cumulative explained variance exceeds a certain threshold (such as 95%) are selected.

[0178] Sort the eigenvalues ​​in descending order to obtain sorted eigendecomposition data. Sort the eigenvalues ​​in the eigendecomposition data in descending order.

[0179] Compute the rate of change of eigenvalues ​​in adjacent intervals. Compute the rate of change between each eigenvalue and its previous eigenvalue.

[0180] According to the interval change rate, the eigenvector with the largest change is selected as the second principal component data. The eigenvector before the eigenvector with the largest interval change rate is selected as the second principal component.

[0181] The sum of eigenvalues ​​in the present invention reflects the overall variability of the entire data set, provides global information, and helps to understand the overall structure of the data. The sum of eigenvalues ​​serves as a benchmark to help determine the criteria and thresholds for subsequent feature selection. Calculating the sum of eigenvalues ​​can evaluate the contribution of each feature to the overall variation of the data, providing a basis for calculating the explained variance ratio. The explained variance ratio quantifies the contribution of each feature to the overall variation of the data, helping to identify the most important features. The explained variance ratio data provides a quantitative basis for feature screening, ensuring the scientificity and rationality of the screening process. By explaining the variance ratio, the degree of information retention after the feature is screened can be evaluated to ensure that the main information is not lost. The sorted explained variance ratio data clarifies the importance priority of each feature and provides clear guidance for subsequent feature screening. The sorting process organizes the data, which helps to analyze and understand the importance of each feature more intuitively. The sorted data makes the feature screening process more efficient and reduces the interference of irrelevant or minor features. Screening is performed according to the sorted explained variance ratio data to ensure that the selected principal component data has the largest explained variance. The screened first transport speed principal component data retains the main variability information of the data and reduces information loss. Through screening, the data structure is significantly simplified, and the efficiency of subsequent analysis and processing is improved. Descending sorting can quickly identify the most significant features in the data set for further analysis. Descending sorting arranges the data by importance, making the subsequent processing process clearer and more efficient. The sorted data clarifies the importance priority of each feature, providing a basis for subsequent change rate calculation and screening. Interval change rate calculation can capture the changing trend of feature values ​​in different intervals and identify features with significant changes. Through change rate data, the importance changes of features can be dynamically analyzed to provide more analysis dimensions. Change rate calculation helps optimize the feature screening process and ensures that the selected features have the largest information change rate. Maximum change rate screening ensures that the most significant features in different intervals are retained, enhancing the expressiveness of the data. The screened second transport speed principal component data provides a more diverse feature representation and improves the comprehensiveness of the analysis. Through maximum change rate screening, the dimensionality reduction effect of the data is further optimized, providing a higher quality data foundation for subsequent anomaly detection and analysis.

[0182] Optionally, the transport speed characteristic data includes first transport speed characteristic data and second transport speed characteristic data, and the nonlinear feature extraction includes:

[0183] Performing linear kernel matrix calculation on the first transport speed principal component data to obtain linear kernel matrix data;

[0184] Perform eigenvalue decomposition on the linear kernel matrix data to obtain linear kernel matrix decomposition data;

[0185] The linear kernel matrix decomposition data is screened by a preset characteristic threshold data and projected into the kernel principal component space to obtain the first transport speed characteristic data;

[0186] Performing Gaussian kernel matrix calculation on the first transport speed principal component data to obtain Gaussian kernel matrix data;

[0187] Perform eigenvalue decomposition on the Gaussian kernel matrix data to obtain Gaussian kernel matrix decomposition data;

[0188] Perform eigenvalue decomposition on the Gaussian kernel matrix decomposition data to obtain Gaussian kernel matrix decomposition data;

[0189] The Gaussian kernel matrix decomposition data is screened by a preset characteristic threshold data and projected into the kernel principal component space to obtain the second transport speed characteristic data.

[0190] The present invention provides an embodiment for calculating a linear kernel matrix of the principal component data of the first transport speed by calculating a dot product. The dot product of the principal component data of the first transport speed is calculated to obtain a linear kernel matrix.

[0191] Perform eigenvalue decomposition on the linear kernel matrix to obtain eigenvalues ​​and eigenvectors. Use the eigenvalue decomposition function to decompose the linear kernel matrix to obtain eigenvalues ​​and eigenvectors.

[0192] The feature vector is selected according to the preset feature threshold. The original data is projected into the feature vector space after the selection to obtain the first transport speed feature data. The feature vector is selected according to the size of the feature value and the feature vector greater than the preset threshold is selected. The original data is projected into the feature vector space after the selection.

[0193] The Gaussian kernel matrix of the principal component data of the first transport speed is calculated, such as by using the RBF kernel function. The RBF kernel function of the principal component data of the first transport speed is calculated to obtain the Gaussian kernel matrix.

[0194] Perform eigenvalue decomposition on the Gaussian kernel matrix to obtain eigenvalues ​​and eigenvectors. Use the eigenvalue decomposition function to decompose the Gaussian kernel matrix to obtain eigenvalues ​​and eigenvectors.

[0195] The feature vector is selected according to the preset feature threshold. The original data is projected into the feature vector space after the selection to obtain the second transport speed feature data. The feature vector is selected according to the size of the feature value and the feature vector greater than the preset threshold is selected. The original data is projected into the feature vector space after the selection.

[0196] The linear kernel calculation method in the present invention is simple and efficient, and is suitable for preliminary feature extraction. The linear kernel can effectively retain the linear relationship of the data and provide a basis for subsequent nonlinear analysis. The linear kernel matrix has low computational complexity and is suitable for rapid processing of large-scale data. Eigenvalue decomposition can extract the main characteristic components in the linear kernel matrix and simplify the data structure. Through eigenvalue decomposition, the dimension of the data is reduced and redundant information is reduced. The data representation after eigenvalue decomposition is more compact, which is convenient for subsequent processing and analysis. Through the preset feature threshold, the most important features are screened out to ensure that the main information of the data is retained. The screening and projection process further reduces the data dimension and improves the calculation efficiency. The first transport speed feature data after screening is more optimized, which is convenient for subsequent analysis and processing. The Gaussian kernel can effectively capture the nonlinear relationship in the data and make up for the shortcomings of the linear kernel. The Gaussian kernel parameters can mine the periodic laws in the principal component data of the first transport speed. Since the generation process of the principal component data of the first transport speed is similar to the elbow law, the data performance is relatively stable. The Gaussian kernel matrix can provide richer feature representation and improve the accuracy of data analysis. The eigenvalue decomposition can extract the main nonlinear characteristic components in the Gaussian kernel matrix and enhance the data representation capability. Through eigenvalue decomposition, the data structure is simplified and the main nonlinear features are retained. The data representation after eigenvalue decomposition is more compact, which is convenient for subsequent processing and analysis. The most important nonlinear features are screened out through the preset feature threshold to ensure that the main information of the data is retained. The screening and projection process enhances the diversity of feature representation and improves the comprehensiveness of data analysis. The screening and projection process further reduces the data dimension and improves computational efficiency.

[0197] Optionally, mutation point detection includes:

[0198] Generate an initial mutation feature set according to the transport mutation feature data to obtain initial mutation feature set data;

[0199] According to the transport mutation feature data and the initial mutation feature set data, a primary calculation is performed through the mutation feature calculation function to obtain the primary mutation feature set data; wherein the mutation feature calculation function is specifically:

[0200] ;

[0201] is the primary mutation signature set data, For transport mutation characteristic quantity data, is the transport mutation characteristic sequence data, To characterize mutation intensity data for transport mutations, For the Transport mutation feature data, is the transport mutation intensity error adjustment term, is the transport mutation intensity error adjustment term, is the weighted data of the transport mutation characteristic amplitude coefficient, is the characteristic amplitude coefficient data of transport mutation, is the weight data of transport mutation characteristic change, It is the periodic weight data of transport mutation characteristics;

[0202] Recursively split the primary mutation feature set data to obtain secondary mutation feature set data;

[0203] The recursive segmentation is as follows:

[0204] ;

[0205] is the secondary mutation feature set data, is the first location data to be placed, is the second location data to be placed, Data for the primary mutation signature set is at position The minimum cost data, Data for primary mutation signature collection from position To location The minimum cost data, is the penalty coefficient data;

[0206] The secondary mutation feature set data is pruned to obtain the transport mutation feature data; the pruning operation is specifically as follows:

[0207] ;

[0208] For transport mutation feature data, is the first location data to be placed, Data for the primary mutation signature set is at position The minimum cost data, Data for primary mutation signature collection from position To location The minimum cost data, is the penalty coefficient data, It is the secondary mutation feature set data.

[0209] The present invention provides an embodiment to extract an initial mutation feature set from transport mutation feature data. The transport mutation feature data is sorted into a sequence form, and each item is a feature value.

[0210] Calculate the primary mutation signature set data according to the formula. Apply the given mutation signature calculation function.

[0211] The primary mutation feature set data is segmented using a recursive segmentation algorithm. A dynamic programming algorithm is used for recursive segmentation to minimize the segmentation cost.

[0212] The secondary mutation feature set data is optimized through the pruning algorithm. Pruning is performed based on the segmentation cost and penalty coefficient to remove redundant segmentation points.

[0213] In the present invention, all transport mutation feature data are integrated through the generation of the initial feature set, laying the foundation for subsequent analysis. The generation of the initial feature set can be regarded as a preprocessing of the original data, removing some noise and improving the data quality. Through the reasonable generation of the feature set, the data organization method is optimized and the efficiency of subsequent processing is improved. Through the mutation feature calculation function, multiple features (such as amplitude, rate of change, and periodicity) are integrated to extract primary features to ensure the multi-dimensional representation of the data. The calculation results provide quantitative feature data, which provides an accurate data basis for subsequent recursive segmentation and pruning operations. Through recursive segmentation, the optimal segmentation point between different positions in the primary feature set is found to ensure the optimality of the segmentation result. The minimum cost is calculated during the recursive segmentation process to ensure that the cost of calculation and processing is reduced while ensuring the quality of the features. The segmentation process optimizes the structure of the feature set, making the data easier to analyze and process. Through the pruning operation, insignificant or redundant feature data is removed to improve the simplicity and effectiveness of the data set. The pruning operation reduces the amount of data, improves the efficiency of subsequent processing, and reduces the computational complexity. It ensures that important mutation features are retained, and improves the accuracy and reliability of anomaly detection.

[0214] Compared with the traditional cost function which usually only considers a single dimension or simple linear features, the mutation feature calculation function adopted in the present invention comprehensively considers multiple dimensions and nonlinear features. Taking into account the mutation intensity and its adjustment terms, it can more sensitively reflect the changes in mutation intensity. This item takes into account the logarithmic change of the amplitude, which can capture the impact of smaller amplitude changes and compress the change amplitude through a logarithmic function, making the mutation characteristics more stable. This item introduces the periodic change of mutation characteristics, and reflects the periodic change rate of mutation characteristics through derivatives, which can more accurately capture the periodic change characteristics of mutations. Compared with the traditional cost function, the calculation formula of the present invention can detect mutation characteristics more sensitively. The various processing in the formula can amplify small changes, and through the comprehensive consideration of multi-dimensional characteristics, the detection is more accurate and reliable. This highly sensitive mutation detection is of great significance for the early detection of potential problems and anomalies in the transportation process.

[0215] Optionally, the present application further provides a highway large-scale transport in-transit supervision system, which is used to execute the highway large-scale transport in-transit supervision method as described above, and the highway large-scale transport in-transit supervision system includes:

[0216] Large-scale transport vehicle supervision data collection module, used to obtain transport speed data and transport vibration data;

[0217] The transport supervision data feature extraction module is used to extract the transport speed feature of the transport speed data to obtain the transport speed feature data, and to extract the transport vibration feature of the transport vibration data to obtain the transport vibration feature data;

[0218] A transport speed vibration frequency map construction module is used to construct a feature map of the transport speed characteristic data and the transport vibration characteristic data to obtain transport speed vibration characteristic map data, and to perform frequency map conversion on the transport speed vibration characteristic map data to obtain transport speed vibration frequency map data;

[0219] The large-piece transport anomaly detection module is used to perform signal decomposition on the transport speed vibration frequency graph data to obtain frequency graph decomposition data, and to perform frequency feature extraction on the frequency graph decomposition data to obtain frequency graph feature data for large-piece transport anomaly detection operations.

[0220] The purpose of the present invention is to obtain the transport speed and vibration data in real time through a virtual positioning system and an acceleration sensor to ensure the comprehensiveness and timeliness of the data. This lays a solid foundation for subsequent feature extraction and analysis. Using a positioning system and a sensor to ensure the accuracy of the data is helpful to improve the accuracy of subsequent feature extraction and anomaly detection. Combining the data of the virtual positioning system and the acceleration sensor, the ability of multi-source data fusion is provided. By comprehensively analyzing data from different sources, more comprehensive and reliable transport status information can be obtained. Through time series decomposition, dimensionality reduction processing and nonlinear feature extraction, the characteristics of the transport speed and vibration data are fully extracted, including time characteristics, frequency characteristics and nonlinear characteristics. These multidimensional features provide rich information for the subsequent feature map construction. By using signal processing technology (such as short-time Fourier transform, Mel frequency cepstral coefficient extraction, etc.), the expression ability of the features is enhanced and the description ability of complex transport states is improved. In the feature extraction process, the noise in the data is removed by dimensionality reduction and screening, the main feature information is retained, and the data quality is improved. By constructing the speed and vibration feature data into a feature map, the visual display of the data is realized. The feature map can intuitively display the state changes during transportation and provide a clear image basis for anomaly detection. The feature graph combines speed and vibration data in a unified framework for analysis, providing a more comprehensive perspective and helping to more accurately identify and analyze abnormal situations. Feature graph construction simplifies the representation of complex data, making subsequent processing and analysis more efficient. The feature graph is fast to build and process, and can meet the needs of real-time monitoring. Through mutation feature collection and change degree extraction, mutation points in the transportation process can be accurately identified and significant changes in transportation status can be captured. This accurate identification capability is crucial for timely detection and handling of abnormal situations. Using algorithms (such as eigenvalue decomposition, change rate calculation, etc.), multi-dimensional change analysis of mutation features is performed to provide more detailed and accurate abnormal change information. Systematically combining mutation features and change degree data for real-time anomaly detection ensures timely warning and response when abnormal situations occur, improving the safety and reliability of the transportation process. Detailed mutation feature and change degree data provide strong data support for transportation managers, assisting in making scientific decisions, optimizing transportation routes and methods, and improving transportation efficiency and safety.

[0221] Therefore, from any point of view, the embodiments should be regarded as illustrative and non-restrictive, and the scope of the present invention is limited by the attached application documents rather than the above description, and it is intended that all changes falling within the meaning and scope of equivalent elements of the application documents are included in the present invention.

[0222] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for supervising the transportation of large items on highways, characterized in that: The method comprises: S1. Acquire transportation speed data and transportation vibration data; S2, extracting the transport speed feature from the transport speed data to obtain transport speed feature data, and extracting the transport vibration feature from the transport vibration data to obtain transport vibration feature data; S3, constructing a characteristic graph of the transport speed characteristic data and the transport vibration characteristic data to obtain transport speed vibration characteristic graph data; S4. Collect mutation characteristics and extract the degree of change according to the transport speed vibration characteristic diagram data to obtain transport mutation characteristic data and transport change degree data to perform large-scale transport anomaly detection operations; S4 includes: Perform edge processing according to the transport speed vibration characteristic diagram data to obtain edge characteristic diagram data; Performing gradient calculation on edge feature map data to obtain gradient feature map data; Perform mutation point detection based on the gradient characteristic map data to obtain transport mutation characteristic data; Extract the degree of change according to the transport mutation characteristic data to obtain the transport change degree data; Based on the transport mutation characteristic data and transport change degree data, anomaly detection is performed through set thresholds or rules to generate early warning information.

2. The method according to claim 1, characterized in that: S1 includes: The position of the large-scale transport vehicle is collected through the virtual positioning system to obtain the position data of the large-scale transport vehicle, and the speed is calculated based on the position data of the large-scale transport vehicle to obtain the speed data of the first large-scale transport vehicle; The acceleration data of the vehicle is collected by an acceleration sensor preset on the large-scale transport vehicle to obtain the acceleration data of the large-scale transport vehicle, and the speed is calculated according to the acceleration data of the large-scale transport vehicle to obtain the speed data of the second large-scale transport vehicle; Perform time series calibration and construct a speed graph based on the speed data of the first large-piece transport vehicle and the speed data of the second large-piece transport vehicle to obtain transport speed data; The first vibration sensor preset on the large-cargo transport vehicle is used to collect vibration data of the large-cargo transport vehicle to obtain the first vibration data of the large-cargo transport vehicle, and the second vibration sensor preset on the large-cargo transport vehicle is used to collect vibration data of the large-cargo transport vehicle to obtain the second vibration data of the large-cargo transport vehicle; A vibration graph is constructed based on the vibration data of the first large-piece transport vehicle and the vibration data of the second large-piece transport vehicle to obtain transport vibration data.

3. The method according to claim 2, characterized in that S2 include: Perform time series decomposition on the transport speed data to obtain transport speed time series decomposition data; Perform dimension reduction processing on the time series decomposition data of transportation speed to obtain dimension reduction data of transportation speed; Perform nonlinear feature extraction on the transport speed dimension reduction data to obtain transport speed feature data; Extract time-frequency features of transportation vibration data to obtain time-frequency vibration feature data; Extract the Mel-frequency cepstrum coefficients of the time-frequency vibration feature data to obtain the vibration frequency domain feature data; The spectrum diagram is constructed according to the vibration frequency domain characteristic data to obtain the transportation vibration characteristic data.

4. The method according to claim 1, characterized in that: S3 includes: Normalizing the transport speed characteristic data and the transport vibration characteristic data to obtain transport speed characteristic normalized data and transport vibration characteristic normalized data; Performing two-dimensional spatial mapping according to the normalized data of the transport speed characteristics and the normalized data of the transport vibration characteristics to obtain two-dimensional mapping data of the transport speed characteristics and two-dimensional mapping data of the transport vibration characteristics; Perform feature fusion according to the two-dimensional mapping data of transport speed characteristics and the two-dimensional mapping data of transport vibration characteristics to obtain a transport feature fusion point data set; The image is constructed based on the transport feature fusion point data set to obtain a preliminary transport feature map; The preliminary transport characteristic map is enhanced to obtain the transport speed vibration characteristic map data.

5. The method according to claim 3, characterized in that: Dimensionality reduction processing includes: Perform covariance calculation on the transport speed time series decomposition data to obtain the transport speed time series covariance matrix data; Perform eigendecomposition on the transport speed time series covariance matrix data to obtain transport speed eigendecomposition data; Perform principal component screening based on the transport speed characteristic decomposition data to obtain the transport speed principal component data; According to the principal component data of transportation speed and the time series decomposition data of transportation speed, dimension reduction projection is performed to obtain the dimension reduction data of transportation speed.

6. The method according to claim 5, characterized in that The transport speed principal component data includes the first transport speed principal component data and the second transport speed principal component data, and the principal component screening includes: Calculate the sum of eigenvalues ​​of the transport speed characteristic decomposition data to obtain the transport speed characteristic sum data; The explained variance ratio is calculated based on the decomposition data of the transport speed characteristics and the sum of the transport speed characteristics to obtain the explained variance ratio data; Sort the explained variance proportion data to obtain the explained variance sort proportion data; The first principal component data of transportation speed is obtained by screening according to the explained variance sorting ratio data and the transportation speed characteristic decomposition data; Sorting the transport speed characteristic decomposition data in descending order to obtain transport speed characteristic decomposition sorting data; According to the transport speed characteristic decomposition and sorting data, the interval change rate is calculated to obtain the characteristic interval change rate data; The maximum change rate is screened according to the characteristic interval change rate data and the transport speed characteristic decomposition sorting data to obtain the second transport speed principal component data.

7. The method according to claim 3, characterized in that The transport speed characteristic data includes first transport speed characteristic data and second transport speed characteristic data, and the nonlinear feature extraction includes: Performing linear kernel matrix calculation on the first transport speed principal component data to obtain linear kernel matrix data; Perform eigenvalue decomposition on the linear kernel matrix data to obtain linear kernel matrix decomposition data; The linear kernel matrix decomposition data is screened by using preset feature threshold data to obtain feature screening data, and a kernel principal component space projection is performed based on the feature screening data to obtain first transport speed feature data; Performing Gaussian kernel matrix calculation on the first transport speed principal component data to obtain Gaussian kernel matrix data; Perform eigenvalue decomposition on the Gaussian kernel matrix data to obtain Gaussian kernel matrix decomposition data; The Gaussian kernel matrix decomposition data is screened by using preset characteristic threshold data to obtain Gaussian kernel matrix screening data, and the kernel principal component space projection is performed based on the Gaussian kernel matrix screening data to obtain the second transport speed characteristic data.

8. The method according to claim 1, characterized in that Mutation point detection includes: Generate an initial mutation feature set according to the transport mutation feature data to obtain initial mutation feature set data; According to the transport mutation feature data and the initial mutation feature set data, a primary calculation is performed through the mutation feature calculation function to obtain the primary mutation feature set data; wherein the mutation feature calculation function is specifically: ； is the primary mutation signature set data, For transport mutation characteristic quantity data, is the transport mutation characteristic sequence data, To characterize mutation intensity data for transport mutations, For the Transport mutation feature data, is the transport mutation intensity error adjustment term, is the transport mutation intensity error adjustment term, is the weighted data of the transport mutation characteristic amplitude coefficient, is the characteristic amplitude coefficient data of transport mutation, is the weight data of transport mutation characteristic change, It is the periodic weight data of transport mutation characteristics; Recursively split the primary mutation feature set data to obtain secondary mutation feature set data; The recursive segmentation is as follows: ; is the secondary mutation feature set data, is the first location data to be placed, is the second location data to be placed, Data for the primary mutation signature set is at position The minimum cost data, Data for primary mutation signature collection from position To location The minimum cost data, is the penalty coefficient data; The secondary mutation feature set data is pruned to obtain the transport mutation feature data; the pruning operation is specifically as follows: ; For transport mutation feature data, is the first location data to be placed, Data for the primary mutation signature set is at position The minimum cost data, Data for primary mutation signature collection from position To location The minimum cost data, is the penalty coefficient data, It is the secondary mutation feature set data.

9. A highway large-scale transportation in-transit supervision system, characterized in that: Used to execute the highway large-scale transport in-transit supervision method as claimed in claim 1, the highway large-scale transport in-transit supervision system comprises: Large-scale transport vehicle supervision data collection module, used to obtain transport speed data and transport vibration data; The transport supervision data feature extraction module is used to extract the transport speed feature of the transport speed data to obtain the transport speed feature data, and to extract the transport vibration feature of the transport vibration data to obtain the transport vibration feature data; A transport speed vibration frequency map construction module is used to construct a feature map of the transport speed characteristic data and the transport vibration characteristic data to obtain transport speed vibration characteristic map data, and to perform frequency map conversion on the transport speed vibration characteristic map data to obtain transport speed vibration frequency map data; The large-piece transport anomaly detection module is used to perform signal decomposition on the transport speed vibration frequency graph data to obtain frequency graph decomposition data, and to perform frequency feature extraction on the frequency graph decomposition data to obtain frequency graph feature data, and to collect mutation features and extract the degree of change based on the frequency graph feature data to obtain mutation feature data and degree of change data, and to perform anomaly detection based on the mutation feature data and degree of change data through set thresholds or rules to generate early warning information.

Images