Road surface slippery state detection method, system and platform based on machine learning

By combining video images, near-infrared spectroscopy and lidar data, using the support vector machine algorithm to build a pavement state prediction model and extract multi-dimensional features, solving the problems of low accuracy and poor robustness of pavement slippery state detection in the existing technology, and realizing the Efficient detection in complex environments.

CN118965135APending Publication Date: 2024-11-15GUANGZHOU NORTH SECOND RING TRANSPORT TECH CO LTD
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202411019099.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-29
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

The existing road surface slippery state detection methods have low accuracy and poor robustness, especially in complex road sections and different environmental conditions, making it difficult to effectively detect slippery states of ultra-high gradient sections and small synthetic longitudinal slope sections.

Method used

Using a machine learning-based method, combining video image data, near-infrared spectral data and lidar data, a pavement state prediction model is constructed through a support vector machine classification learning algorithm, multi-dimensional features are extracted and feature engineering is carried out, and detection accuracy and robustness are improved.

Benefits of technology

It significantly improves the accuracy and robustness of wet and slippery road surface detection, and can effectively detect the slippery state of complex road sections under various environmental conditions, enhancing the system's adaptability and detection effect.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118965135A_ABST
    Figure CN118965135A_ABST
Patent Text Reader

Abstract

The invention discloses a road surface slippery state detection method, system and platform based on machine learning, and a storage medium, and the method comprises the steps: obtaining first road surface data corresponding to a to-be-detected slippery state road surface through collection, and carrying out the processing of the first road surface data, and generating corresponding second road surface data; wherein the first road surface data comprises video image data, near infrared spectrum data and laser radar data; the second road surface data are characteristic data related to the to-be-detected wet and slippery road surface; on the basis of a support vector machine classification learning algorithm, constructing a prediction model corresponding to the road surface state information in real time, and evaluating, training and processing the prediction model in combination with test reference data; generating third road surface data corresponding to the road surface in real time according to the second road surface data in combination with a prediction model subjected to evaluation training processing; the third road surface data is road surface slippery state prediction data, and the corresponding system, platform and storage medium can improve the precision and robustness of slippery road surface detection.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of road slippery state detection and processing, and specifically relates to a road slippery state detection method, system and platform based on machine learning. Background Art

[0002] At present, the drainage path of multi-lane highway pavement is long, especially the super-high gradient section, the small synthetic longitudinal slope and other difficult drainage sections, which are very likely to cause poor pavement drainage. When the road surface is covered with water film, due to the lubrication effect of the water film at the tire-pavement interface, the adhesion coefficient between the tire and the road is significantly reduced, resulting in an increase in the vehicle's braking distance and a decrease in driving stability. It may even cause hydroplaning, which seriously affects driving safety. Based on the above reasons, it is particularly important to detect the slippery state of the road surface.

[0003] Traditional methods for detecting slippery roads mainly rely on data from a single sensor (such as a camera or lidar), which has a single source and results in low detection accuracy. They are also unstable and have poor robustness under different environmental conditions (such as light changes and weather changes). They also have poor detection effects on complex road sections (such as super-elevated gradient sections and sections with small synthetic longitudinal slopes).

[0004] Therefore, in view of the technical defects of the above traditional methods in detecting slippery road conditions, such as low accuracy and poor robustness, it is urgent to design and develop a method, system and platform for detecting slippery road conditions based on machine learning. Summary of the invention

[0005] In order to overcome the shortcomings and difficulties of the above-mentioned prior art, the purpose of the present invention is to provide a road slippery state detection method, system, platform and storage medium based on machine learning to improve the accuracy and robustness of slippery road detection.

[0006] The first purpose of the present invention is to provide a method for detecting a slippery road condition based on machine learning; the second purpose of the present invention is to provide a system for detecting a slippery road condition based on machine learning; the third purpose of the present invention is to provide a platform for detecting a slippery road condition based on machine learning; the fourth purpose of the present invention is to provide a computer-readable storage medium.

[0007] The first object of the present invention is achieved in that the method comprises the following steps:

[0008] Acquire first road surface data corresponding to the road surface in a slippery state to be detected, and generate corresponding second road surface data according to the first road surface data; wherein the first road surface data includes video image data, near infrared spectrum data and laser radar data; and the second road surface data is characteristic data related to the road surface in a slippery state to be detected;

[0009] Based on the support vector machine classification learning algorithm, a prediction model corresponding to the road surface state information is constructed in real time, and the prediction model is evaluated and trained in combination with the test reference data;

[0010] According to the second road surface data and in combination with the prediction model after evaluation and training, third road surface data corresponding to the road surface is generated in real time; wherein the third road surface data is road surface slippery state prediction data.

[0011] Furthermore, the collecting and acquiring first road surface data corresponding to the road surface in the wet and slippery state to be detected, and generating corresponding second road surface data according to the first road surface data, further includes:

[0012] Preprocess the first road surface data in real time, and extract and generate second road surface data corresponding to the first road surface data; wherein the preprocessing includes missing value processing, data standardization processing and feature engineering processing; the second road surface data includes color feature data, texture feature data, shape feature data, water molecule absorption peak feature data, road surface material absorption peak feature data, road surface scattering feature data, road surface reflectivity feature data and surface height feature data.

[0013] Furthermore, the collecting and acquiring first road surface data corresponding to the road surface in the wet and slippery state to be detected, and generating corresponding second road surface data according to the first road surface data, further includes:

[0014] Determine and generate missing value status data of samples corresponding to the first road surface data, and perform real-time removal processing or central interpolation processing on the first road surface data according to the missing value status data;

[0015] Standardizing the first road surface data based on a minimum-maximum scaling method, and mapping the processed data to at least one range;

[0016] The first road surface data is processed according to the feature selection of the LASSO regression method in the embedding method to remove the corresponding redundant feature data, and the high-dimensional feature data corresponding to the first road surface data is processed by dimensionality reduction combined with the principal component analysis method.

[0017] Furthermore, the dimensionality reduction processing of the high-dimensional feature data corresponding to the first road surface data by combining the principal component analysis method also includes:

[0018] Performing zero-mean processing on all feature data corresponding to the first road surface data;

[0019] Calculate and generate a corresponding characteristic covariance matrix; and perform eigenvalue decomposition on the covariance matrix to generate eigenvalue data and eigenvector data corresponding to the covariance matrix; wherein the formula for calculating and generating the corresponding characteristic covariance matrix is:

[0020]

[0021] Among them, x i For each feature in the data, is the mean, T is the transpose of the centered eigenvector, and n is the number of samples.

[0022] Furthermore, the prediction model corresponding to the road surface state information is constructed in real time based on the support vector machine classification learning algorithm, and the prediction model is evaluated and trained in combination with the test reference data, and further includes:

[0023] Based on a support vector machine classification learning algorithm, dividing the second road surface data into at least three corresponding feature data types;

[0024] According to the characteristic data type, establishing at least three corresponding support vector machine models, and training and processing the support vector machine models;

[0025] Optimizing the parameter data corresponding to the support vector machine model by using a grid search method, and generating corresponding parameter combination data based on a cross-validation method; wherein the parameter data includes penalty parameter data and kernel function parameter data; and the parameter combination data is the optimal parameter combination data;

[0026] Based on the accuracy index data, precision index data and recall index data in the test reference data, the performance of the prediction model trained and processed is evaluated in real time.

[0027] Furthermore, the grid search method is combined to optimize the parameter data corresponding to the support vector machine model, and the corresponding parameter combination data is generated based on the cross-validation method, and further includes:

[0028] Using the logarithmic scale as the search range, determine and generate the corresponding search range of the penalty parameter C and the kernel function parameter γ; where the range of C is: 2 -5 , 2 -3 , 2 -1 , 2 1 , 2 3 , 2 5 ; Range of γ: 2 -5 , 2 -3 , 2 -1 , 21 , 2 3 , 2 5 ;

[0029] Within the search range, all possible C and γ combinations are traversed point by point, and the performance corresponding to each combination is evaluated in combination with the cross-validation method; this includes dividing the training set data into K parts, and for each parameter combination, dividing the data into a training set and a validation set, performing K iterations, each time selecting a different part as the validation set, and the remaining K-1 parts as the training set.

[0030] The second object of the present invention is achieved by: the system is applied to the method for detecting a slippery road surface based on machine learning, and the system comprises:

[0031] a data acquisition processing unit, configured to collect and acquire first road surface data corresponding to the road surface in a slippery state to be detected, and to generate corresponding second road surface data according to the first road surface data; wherein the first road surface data includes video image data, near infrared spectrum data and laser radar data; and the second road surface data is characteristic data related to the road surface in a slippery state to be detected;

[0032] A prediction model building unit, used to build a prediction model corresponding to the road surface state information in real time based on a support vector machine classification learning algorithm, and to evaluate and train the prediction model in combination with test reference data;

[0033] The prediction data generating unit is used to generate third road surface data corresponding to the road surface in real time according to the second road surface data and in combination with the prediction model after evaluation and training processing; wherein the third road surface data is the road surface slippery state prediction data.

[0034] Furthermore, the data acquisition processing unit further includes:

[0035] A first data generation module is used to preprocess the first road surface data in real time, and extract and generate second road surface data corresponding to the first road surface data; wherein the preprocessing includes missing value processing, data standardization processing and feature engineering processing; the second road surface data includes color feature data, texture feature data, shape feature data, water molecule absorption peak feature data, road surface material absorption peak feature data, road surface scattering feature data, road surface reflectivity feature data and surface height feature data;

[0036] And / or, the data acquisition processing unit further includes:

[0037] A first data determination module, used for determining and generating missing value status data of samples corresponding to the first road surface data, and performing real-time removal processing or center interpolation processing on the first road surface data according to the missing value status data;

[0038] A first data processing module, configured to standardize the first road surface data based on a minimum-maximum scaling method, and map the processed data to at least one range;

[0039] A second data processing module is used to process the first road surface data according to the feature selection of the LASSO regression method in the embedding method, remove the corresponding redundant feature data, and combine the principal component analysis method to reduce the dimension of the high-dimensional feature data corresponding to the first road surface data;

[0040] And / or, the second data processing module further includes:

[0041] A third data processing module, used for performing zero mean processing on all feature data corresponding to the first road surface data;

[0042] The first calculation generation module is used to calculate and generate a corresponding characteristic covariance matrix; and to process the covariance matrix by eigenvalue decomposition to generate eigenvalue data and eigenvector data corresponding to the covariance matrix; wherein the formula for calculating and generating the corresponding characteristic covariance matrix is:

[0043]

[0044] Among them, x i For each feature in the data, is the mean, T is the transpose of the centered eigenvector; n is the number of samples;

[0045] And / or, the prediction model building unit further includes:

[0046] A second data generating module, configured to divide the second road surface data into at least three corresponding feature data types based on a support vector machine classification learning algorithm;

[0047] A fourth data processing module, used to establish at least three corresponding support vector machine models according to the characteristic data type, and train and process the support vector machine models;

[0048] A third data generation module is used to optimize and process the parameter data corresponding to the support vector machine model in combination with a grid search method, and generate corresponding parameter combination data based on a cross-validation method; wherein the parameter data includes penalty parameter data and kernel function parameter data; and the parameter combination data is the optimal parameter combination data;

[0049] A fifth data processing module, used for evaluating the performance of the training and processing of the prediction model in real time based on the accuracy index data, the precision index data and the recall index data in the test reference data;

[0050] And / or, the third data generating module further includes:

[0051] The fourth data generation module is used to determine and generate the corresponding search range of the penalty parameter C and the kernel function parameter γ with the logarithmic scale as the search range; wherein the range of C is: 2 -5 , 2 -3 , 2 -1 , 2 1 , 2 3 , 2 5 ; Range of γ: 2 -5 , 2 -3 , 2 -1 , 2 1 , 2 3 , 2 5 ;

[0052] The sixth data processing module is used to traverse all possible C and γ combinations within the search range point by point, and use the cross-validation method to evaluate the performance corresponding to each combination; which includes dividing the training set data into K parts, and for each parameter combination, dividing the data into a training set and a validation set, performing K iterations, each time selecting a different part as the validation set, and the remaining K-1 parts as the training set.

[0053] The third object of the present invention is achieved as follows: it includes a processor, a memory and a road surface slippery state detection platform control program based on machine learning; wherein the road surface slippery state detection platform control program based on machine learning is executed by the processor, and the road surface slippery state detection platform control program based on machine learning is stored in the memory, and the road surface slippery state detection platform control program based on machine learning implements the road surface slippery state detection method based on machine learning.

[0054] The fourth object of the present invention is achieved in this way: the computer-readable storage medium stores a road slippery state detection platform control program based on machine learning, and the road slippery state detection platform control program based on machine learning implements the road slippery state detection method based on machine learning.

[0055] The present invention collects and acquires first road surface data corresponding to a road surface in a slippery state to be detected through a method, and generates corresponding second road surface data based on the first road surface data; wherein, the first road surface data includes video image data, near-infrared spectrum data and lidar data; the second road surface data is feature data related to the road surface in a slippery state to be detected; based on a support vector machine classification learning algorithm, a prediction model corresponding to the road surface state information is constructed in real time, and the prediction model is evaluated and trained in combination with test reference data; based on the second road surface data, and in combination with the prediction model after evaluation and training, third road surface data corresponding to the road surface is generated in real time; wherein, the third road surface data is road surface slippery state prediction data, as well as a system, platform and storage medium corresponding to the method, which can improve the accuracy and robustness of slippery road surface detection.

[0056] The solution of the present invention combines multi-source data from cameras, near-infrared spectrometers and lidars, and extracts multi-dimensional features (color, texture, shape, water molecule absorption peaks, road material absorption peaks, road reflectivity, etc.) through a multi-view learning method, greatly improving the detection accuracy; by using near-infrared spectra and lidars, these sensors are insensitive to light changes and weather conditions, and combined with the SVM classification algorithm, the system has high robustness under various environmental conditions; by extracting road geometric features (such as surface height and curvature) through lidar data, combined with multi-source data analysis, the system's adaptability and detection effect in complex road sections are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0058] Figure 1 This is a slippery road surface detection flow chart of an embodiment of a method for detecting a slippery road surface based on machine learning of the present invention;

[0059] Figure 2 A schematic diagram of a slippery road surface detection system according to an embodiment of a method for detecting a slippery road surface based on machine learning of the present invention;

[0060] Figure 3 This is a schematic diagram of the SVM learning process of an embodiment of a method for detecting a slippery road surface based on machine learning in the present invention;

[0061] Figure 4 A schematic diagram of a process of detecting a slippery road surface based on machine learning according to the present invention;

[0062] Figure 5 This is a schematic diagram of the architecture of a road slippery state detection system based on machine learning in the present invention;

[0063] Figure 6 This is a schematic diagram of the architecture of a road slippery state detection platform based on machine learning in the present invention;

[0064] Figure 7 A schematic diagram of a computer-readable storage medium architecture in an embodiment of the present invention;

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

[0066] In order to better understand the purpose, technical solutions and advantages of the present invention, the present invention is further described below in conjunction with the accompanying drawings and specific implementation methods. Those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification.

[0067] The present invention may also be implemented or applied through other different specific examples, and the details in this specification may also be modified and changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention.

[0068] It should be noted that if the embodiments of the present invention involve directional indications (such as up, down, left, right, front, back, etc.), the directional indications are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0069] In addition, if there are descriptions involving "first", "second", etc. in the embodiments of the present invention, the descriptions of "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or suggesting their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. Secondly, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in the field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0070] Preferably, the present invention's method for detecting slippery road conditions based on machine learning is applied in one or more terminals or servers. The terminal is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0071] The terminal can be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The terminal can interact with the client through a keyboard, a mouse, a remote control, a touch pad, or a voice control device.

[0072] The present invention is to realize a method, system, platform and storage medium for detecting the slippery state of a road surface based on machine learning.

[0073] like Figure 4 , which is a flow chart of a method for detecting a slippery road surface based on machine learning provided in an embodiment of the present invention.

[0074] In this embodiment, the road slippery state detection method based on machine learning can be applied to a terminal with a display function or a fixed terminal. The terminal is not limited to a personal computer, a smart phone, a tablet computer, a desktop computer or an all-in-one computer equipped with a camera, etc.

[0075] The method for detecting a slippery road surface based on machine learning can also be applied to a hardware environment consisting of a terminal and a server connected to the terminal via a network. The network includes but is not limited to: a wide area network, a metropolitan area network, or a local area network. The method for detecting a slippery road surface based on machine learning in the embodiment of the present invention can be executed by a server, or by a terminal, or by both a server and a terminal.

[0076] For example, for a terminal that needs to perform a machine learning-based road slippery state detection function, the machine learning-based road slippery state detection function provided by the method of the present invention can be directly integrated on the terminal, or a client for implementing the method of the present invention can be installed. For another example, the method provided by the present invention can also be run on a server or other device in the form of a software development kit (SDK), and an interface for a machine learning-based road slippery state detection function is provided in the form of SDK. The terminal or other device can implement the machine learning-based road slippery state detection function through the provided interface. The present invention is further described below in conjunction with the accompanying drawings.

[0077] like Figure 1-Figure 4 As shown, the present invention provides a method for detecting a slippery road surface based on machine learning, wherein the method comprises the following steps:

[0078] S1. Acquire first road surface data corresponding to a road surface in a slippery state to be detected, and generate corresponding second road surface data according to the first road surface data; wherein the first road surface data includes video image data, near infrared spectrum data and laser radar data; and the second road surface data is characteristic data related to the road surface in a slippery state to be detected;

[0079] S2. Based on the support vector machine classification learning algorithm, a prediction model corresponding to the road surface state information is constructed in real time, and the prediction model is evaluated and trained in combination with the test reference data;

[0080] S3. Based on the second road surface data and in combination with the prediction model after evaluation and training, generate in real time third road surface data corresponding to the road surface; wherein the third road surface data is road surface slippery state prediction data.

[0081] The collecting and acquiring first road surface data corresponding to the road surface in the wet and slippery state to be detected, and generating corresponding second road surface data according to the first road surface data, further includes:

[0082] S11. Preprocess the first road surface data in real time, and extract and generate second road surface data corresponding to the first road surface data; wherein the preprocessing includes missing value processing, data standardization processing and feature engineering processing; the second road surface data includes color feature data, texture feature data, shape feature data, water molecule absorption peak feature data, road surface material absorption peak feature data, road surface scattering feature data, road surface reflectivity feature data and surface height feature data.

[0083] The collecting and acquiring first road surface data corresponding to the road surface in the wet and slippery state to be detected, and generating corresponding second road surface data according to the first road surface data, further includes:

[0084] S12, determining and generating missing value status data of samples corresponding to the first road surface data, and performing real-time removal processing or center interpolation processing on the first road surface data according to the missing value status data;

[0085] S13, standardizing the first road surface data based on a minimum-maximum scaling method, and mapping the processed data to at least one range;

[0086] S14, processing the first road surface data according to the feature selection of the LASSO regression method in the embedding method, removing the corresponding redundant feature data, and combining the principal component analysis method to reduce the dimension of the high-dimensional feature data corresponding to the first road surface data.

[0087] The step of combining the principal component analysis method with the high-dimensional feature data corresponding to the first road surface data for dimensionality reduction processing also includes:

[0088] S141, performing zero-mean processing on all feature data corresponding to the first road surface data;

[0089] S142, calculate and generate a corresponding characteristic covariance matrix; and perform eigenvalue decomposition on the covariance matrix to generate eigenvalue data and eigenvector data corresponding to the covariance matrix; wherein the formula for calculating and generating the corresponding characteristic covariance matrix is:

[0090]

[0091] Among them, x i For each feature in the data, is the mean, T is the transpose of the centered eigenvector, and n is the number of samples.

[0092] The method of constructing a prediction model corresponding to the road surface status information in real time based on the support vector machine classification learning algorithm and evaluating and training the prediction model in combination with the test reference data also includes:

[0093] S21, dividing the second road surface data into at least three corresponding feature data types based on a support vector machine classification learning algorithm;

[0094] S22, establishing at least three corresponding support vector machine models according to the characteristic data types, and training and processing the support vector machine models;

[0095] S23, optimizing and processing the parameter data corresponding to the support vector machine model by using a grid search method, and generating corresponding parameter combination data based on a cross-validation method; wherein the parameter data includes penalty parameter data and kernel function parameter data; and the parameter combination data is the optimal parameter combination data;

[0096] S24. Based on the accuracy index data, precision index data and recall index data in the test reference data, the performance of the training and processing of the prediction model is evaluated in real time.

[0097] The method of optimizing the parameter data corresponding to the support vector machine model by combining the grid search method and generating corresponding parameter combination data based on the cross-validation method also includes:

[0098] S231, using the logarithmic scale as the search range, determine and generate the corresponding search range of the penalty parameter C and the kernel function parameter γ; wherein the range of C is: 2 -5 , 2 -3 , 2 -1 , 2 1 , 2 3 , 2 5 ; Range of γ: 2 -5 , 2 -3 , 2 -1 , 2 1 , 2 3 , 2 5 ;

[0099] S232. Within the search range, traverse all possible C and γ combinations point by point, and use the cross-validation method to evaluate the performance corresponding to each combination; including dividing the training set data into K parts, and for each parameter combination, dividing the data into a training set and a validation set, performing K iterations, each time selecting a different part as the validation set, and the remaining K-1 parts as the training set.

[0100] Specifically, in an embodiment of the present invention, the present invention proposes a method for detecting slippery roads based on machine learning, which can effectively integrate information from video images, near-infrared spectra, and lidar data to improve the accuracy and robustness of slippery road detection. Figure 1 The method shown comprises the following steps:

[0101] Data collection: Use cameras, near-infrared spectrometers and lidars to collect video images, near-infrared spectrum data and lidar data respectively. Data preprocessing: Preprocess the raw data, including missing value processing, data standardization and feature engineering. Feature extraction: Extract features related to slippery road detection from video images, near-infrared spectra and lidar data, including color features, texture features, shape features, water molecule absorption peak features, pavement material absorption peak features, pavement reflectivity features, pavement texture features and pavement geometry features. Model training and evaluation: Use the support vector machine (SVM) classification learning algorithm to train the pavement state information prediction model, and use test data to evaluate the performance of the trained SVM model. Pavement state prediction: Input the new collected data into the trained SVM model to predict whether the pavement state is slippery.

[0102] In the slippery road detection method, the data preprocessing step includes: missing value processing: using methods such as removal and interpolation to process missing values. Data standardization: using the minimum-maximum scaling method to standardize data. Feature engineering: using methods such as feature selection and dimensionality reduction to perform feature engineering.

[0103] The slippery road surface detection method, wherein the feature extraction step includes extracting the following features: color features: such as RGB values, HSV values, etc. Texture features: such as gray level co-occurrence matrix (GLCM), etc. Shape features: such as shape context (SC), etc. Features of water molecule absorption peaks: such as peak position, peak intensity, peak width, etc. Features of road surface material absorption peaks: such as peak position, peak intensity, etc. Road surface scattering features: scattering index. Road surface reflectivity features: such as average reflectivity, standard deviation, etc. Surface height features: such as road surface feature height.

[0104] The wet and slippery road detection method, wherein the model training and evaluation steps include: respectively classifying color features, texture features, shape features, water molecule absorption peak features, road material absorption peak features, road scattering features, road reflectivity features, and surface height features into three categories, and standardizing them respectively. The three categories of extracted features are used as the input of the model using a multi-view learning method, three models are established using SVM, and the road state information (such as wet, dry, etc.) is used as the separate output of the model. The performance of the trained SVM model is evaluated using test data, the prediction results of the three models are fused, and the overall model output is completed using a weighted method.

[0105] like Figure 2The figure shows a block diagram of the slippery road surface detection system of the present invention, which includes the following parts: Data acquisition unit: This unit is composed of a camera, a near-infrared spectrometer and a laser radar, which are responsible for collecting video images, near-infrared spectrum data and laser radar data respectively. Data preprocessing unit: This unit is responsible for preprocessing the video images, near-infrared spectrum and laser radar data from the data acquisition unit. The preprocessing steps include missing value processing, data standardization and feature engineering. Feature extraction unit: This unit is responsible for extracting features related to slippery road detection from the preprocessed data, including: Camera features: extracting color features (mean, variance and histogram of HSV color space), texture features (contrast, dissimilarity, homogeneity, energy and entropy of gray-level co-occurrence matrix) and shape features (shape context descriptor); Near infrared spectrum features: extracting features of water molecule absorption peak (center position, maximum value, peak width), features of pavement material absorption peak (peak position, peak intensity, peak width) and pavement scattering features (scattering index); LiDAR features: extracting pavement reflectivity features (average reflectivity and standard deviation) and surface height features (average height, standard deviation, curvature). Classification learning unit: This unit uses the support vector machine (SVM) classification learning algorithm to train the slippery road detection model. The specific steps include: feature division, model training and parameter optimization. Decision unit: This unit predicts new data based on the trained slippery road detection model to determine whether the road is slippery. The specific steps include: data input, feature extraction, model prediction, and result judgment.

[0106] The preprocessing steps in the data preprocessing unit include missing value processing, data standardization and feature engineering, specifically:

[0107] Missing value processing: Use the removal and interpolation methods to remove samples with more missing values, and use the central interpolation method to fill in samples with fewer missing values.

[0108] Data normalization: Map the data to the specified range [0,1] using the min-max scaling method.

[0109] Feature Engineering: This includes feature selection and dimensionality reduction. LASSO regression is used to select important features, and principal component analysis (PCA) is used for dimensionality reduction.

[0110] The feature extraction unit specifically extracts each feature as follows: camera feature extraction: color feature extraction, using the HSV color space to convert the RGB image into an HSV image, and extracting statistical features of the three channels of hue (H), saturation (S) and brightness (V) from the HSV image, including mean, variance and histogram.

[0111] Texture feature extraction: Use the gray-level co-occurrence matrix to extract image texture features. The specific steps are as follows: Convert the color image to a gray-scale image. Calculate the gray-level co-occurrence matrix for different directions (0°, 45°, 90°, 135°) and different distances (1, 2, 3 pixels). Extract texture features from the gray-level co-occurrence matrix, including contrast, dissimilarity, homogeneity, energy, and entropy.

[0112] Shape feature extraction: Use the shape context method to extract shape features in the image. The specific steps are as follows: Use the edge detection algorithm (Canny algorithm) to extract the edges of the image. For each edge point, calculate its distance and angle distribution to other edge points to form a shape context descriptor. Perform statistical analysis on the shape context descriptor to extract shape features.

[0113] Near-infrared spectrum feature extraction: Water molecule absorption peak feature extraction, identify the characteristic bands of water molecule absorption peaks in near-infrared spectrum data, and calculate the center position, maximum value, peak width and other features of the absorption peak.

[0114] Pavement material absorption peak feature extraction: Identify the characteristic absorption peak bands of pavement materials (such as asphalt, concrete, etc.) and calculate the peak position, peak intensity, peak width and other characteristics of the absorption peak.

[0115] Road surface scattering feature extraction: Calculate the scattering index of spectral data to describe the scattering characteristics of the spectrum. The scattering index is defined as the rate of change of reflectivity in different bands.

[0116] LiDAR feature extraction: Road surface reflectivity feature extraction, extract reflectivity information from LiDAR point cloud data, calculate the average reflectivity and standard deviation of the road surface area as road surface reflectivity features.

[0117] Surface height feature extraction: Surface height information is extracted from the LiDAR point cloud data, and the average height, standard deviation, and curvature of the road surface area are calculated as surface height features.

[0118] The classification learning unit has the following specific steps: Feature division: The extracted features are divided into three categories: camera features, near-infrared spectral features, and lidar features. Model training: Three SVM models are trained separately: Model 1: Input camera features, including color features, texture features, and shape features, with a total of 3 dimensions; Model 2: Input near-infrared spectral features, including water molecule absorption peak features, pavement material absorption peak features, and pavement scattering features, with a total of 6 dimensions; Model 3: Input lidar features, including pavement reflectivity features and surface height features, with a total of 5 dimensions. Model parameter optimization: The grid search method is used to optimize the hyperparameters of SVM, including penalty parameters and kernel function parameters, and the cross-validation method is used to evaluate the model performance and select the best parameter combination.

[0119] The decision-making unit has the following operating steps: Data input: input new video images, near infrared spectrum data and lidar data. Feature extraction: extract relevant features from the new data. Model prediction: input the extracted features into the trained SVM model and output the road surface state prediction results. Result judgment: judge whether the road surface is slippery based on the prediction results of the model.

[0120] The model training specifically divides the features of every two days into a training set and a test set, where the training set accounts for 80% of the total data and the test set accounts for 20%;

[0121] The model parameter optimization is specifically to optimize the hyperparameters of the SVM model to improve the prediction accuracy and generalization ability of the model. The key hyperparameters include the penalty parameter C and the kernel function parameter γ, and the grid search method and the cross-validation method are used.

[0122] Define the search range: Use the logarithmic scale as the search range and determine the search range of the penalty parameter C and the kernel function parameter γ. The range of C: 2 -5 , 2 -3 , 2 -1 , 2 1 , 2 3 , 2 5 . Range of γ: 2 -5 , 2 -3 , 2 -1 , 2 1 , 2 3 , 2 5 .

[0123] Grid search process: Within the search range, traverse all possible C and γ combinations point by point, and use the cross-validation method to evaluate the performance of each combination. The specific steps are: divide the training set data into K parts (let K = 10), for each parameter combination, divide the data into a training set and a validation set, perform K iterations, and each time select a different 1 part as the validation set, and the remaining K-1 parts as the training set.

[0124] Model evaluation: Model evaluation trains the SVM model and evaluates the model performance on the validation set, recording the accuracy of each validation. Calculate the average accuracy of K iterations as the performance indicator of this parameter combination.

[0125] Example 1: This example introduces a slippery road surface detection method based on machine learning, which uses a camera, a near-infrared spectrometer and a lidar to collect data, and uses a support vector machine (SVM) classification learning algorithm to predict the road surface condition.

[0126] Data collection: Use cameras, near-infrared spectrometers and lidars to collect video images, near-infrared spectral data and lidar data respectively. The camera used is a high-definition color camera installed on the front of the vehicle, with a collection frequency of 30 frames per second and a resolution of 1920x1080 pixels. The near-infrared spectrometer covers a wavelength range of 900-1700nm and is installed at the bottom of the vehicle to collect near-infrared spectral data reflected from the road surface. The lidar model is XYZ-LiDAR 3000, with a collection frequency of 10Hz, installed on the front of the vehicle to collect 3D point cloud data of the road surface.

[0127] Camera: Collects color video images to extract color features and texture features.

[0128] Near-infrared spectrometer: collects near-infrared spectrum data to extract the characteristics of water molecule absorption peaks and pavement material absorption peaks.

[0129] LiDAR: Collect LiDAR data to extract road surface scattering characteristics, road surface reflectivity characteristics and surface height characteristics.

[0130] Data preprocessing: missing value processing, using methods such as removal and central interpolation to process missing values.

[0131] Removal: For samples with more missing values, directly remove them from the data set. Samples with more missing values ​​are defined as those with more than 20% of the total number of features.

[0132] Interpolation: For samples with fewer missing values, central interpolation is used for filling. For continuous data, linear interpolation is used to fill missing values, that is, interpolation is calculated through linear functions based on the known values ​​before and after the missing value; for categorical data, nearest neighbor interpolation is used to fill missing values, that is, the closest known value is selected for filling based on the known values ​​before and after the missing value.

[0133] Data standardization: Use the minimum-maximum scaling method to standardize the data and map the data to a specified range [0,1], specifically for each eigenvalue x i , calculate its minimum value x min and the maximum value x max , scale each eigenvalue, and the calculation formula is:

[0134]

[0135] Feature engineering: feature selection, using the LASSO regression method in the embedding method to select features and remove redundant features.

[0136] Dimensionality reduction: Use the principal component analysis (PCA) method to reduce high-dimensional features to low-dimensional features. The specific steps are as follows:

[0137] All features are zero-meaned, that is, for each feature x i , minus its mean

[0138] Calculate the feature covariance matrix ∑:

[0139]

[0140] Among them, x i For each feature in the data, is the mean, T is the transpose of the centered eigenvector, and n is the number of samples.

[0141] Perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and eigenvectors.

[0142] The first k eigenvectors with cumulative contribution rates reaching 95% are selected to construct the feature space after dimensionality reduction.

[0143] Feature extraction: Extract features related to slippery road detection from video images, near-infrared spectra and lidar data. The features extracted from video images include color features, texture features, and shape features. The features extracted from near-infrared spectra include the features of water molecule absorption peaks, road material absorption peaks, and road scattering features. The features extracted from lidar include road reflectivity features and surface height features.

[0144] The color feature uses HSV value, namely hue, saturation and brightness value, the texture feature uses gray level co-occurrence matrix (GLCM), and the shape feature uses shape context (SC); the characteristics of water molecule absorption peak use center position, maximum value and width, the characteristics of pavement material absorption peak use peak position and peak intensity, and the pavement scattering feature uses scattering index; the pavement reflectivity feature uses average reflectivity and standard deviation, and the surface height feature uses average height, standard deviation and curvature.

[0145] Model training: A multi-view learning method was adopted, and part of the features extracted from the three categories were used as the training set as the model input. Three models were established using SVM. The first model included 1 dimension of color feature, 1 dimension of texture feature, and 1 dimension of shape feature, totaling 3 dimensions; the second model included 3 dimensions of water molecule absorption peak, 2 dimensions of pavement material absorption peak, and 1 dimension of pavement scattering feature, totaling 6 dimensions; the third model included 2 dimensions of pavement reflectivity and 3 dimensions of surface height, totaling 5 dimensions.

[0146] The road surface condition information (such as wet, dry, etc.) is used as a separate output of the model.

[0147] SVM has good results under small sample, nonlinear and high dimensional conditions. Its learning process is as follows: Figure 3As shown. The following takes the first model as an example to illustrate its principle. The color feature, texture feature, and shape feature are captured using pythonOpenCV, pythonscikit-image, and pythonOpenCV libraries respectively to form color feature A, texture feature B, and shape feature C.

[0148] The color feature A, texture feature B, and shape feature C are fused to create a three-dimensional tuple, and a training set DataSet is constructed, that is, DataSet = [(A1, B1, C1),..., (A i ,B i ,C i ),...,(A l ,B l ,C l )].

[0149] The mathematical expression of the generative model of SVM is The minimum value of. Among them, [(A1,B1,C1),...,(A i ,B i ,C i ),...,(A i ,B l ,C l )] is the training set, is the weight vector, and C is a hyperparameter to control the loss value ε i The Gaussian kernel function RBF is selected for nonlinear classification, and the convex quadratic programming (SMO) is used to solve the support vector. Finally, the decision function can be constructed in the form of where α i is the Langrangian operator of the support vector, y i is the category label of the support vector, is the Gaussian kernel function RBF.

[0150] SVM: Model evaluation: Accuracy, precision, and recall are used to evaluate model performance.

[0151] Accuracy: the ratio of correctly predicted samples to the total number of samples. Precision: the ratio of correctly predicted positive samples to the number of predicted positive samples. Recall: the ratio of correctly predicted positive samples to the number of actual positive samples. The three SVM models were evaluated using the test set respectively, and their accuracy, precision, recall and F1 values ​​were calculated.

[0152] The evaluation indicators of the three models are compared to analyze the contribution of different features to slippery road detection. The overall model output weight is determined according to the evaluation indicators of each model, the performance of the trained SVM model is evaluated using test data, the prediction results of the three models are fused, and the overall model output is completed in a weighted manner.

[0153] To achieve the above object, the present invention also provides a road slippery state detection system based on machine learning, wherein the system is applied to the road slippery state detection method based on machine learning, such as Figure 5 As shown, the system specifically includes:

[0154] a data acquisition processing unit, configured to collect and acquire first road surface data corresponding to the road surface in a slippery state to be detected, and to generate corresponding second road surface data according to the first road surface data; wherein the first road surface data includes video image data, near infrared spectrum data and laser radar data; and the second road surface data is characteristic data related to the road surface in a slippery state to be detected;

[0155] A prediction model building unit, used to build a prediction model corresponding to the road surface state information in real time based on a support vector machine classification learning algorithm, and to evaluate and train the prediction model in combination with test reference data;

[0156] The prediction data generating unit is used to generate third road surface data corresponding to the road surface in real time according to the second road surface data and in combination with the prediction model after evaluation and training processing; wherein the third road surface data is the road surface slippery state prediction data.

[0157] The data acquisition processing unit further includes:

[0158] A first data generation module is used to preprocess the first road surface data in real time, and extract and generate second road surface data corresponding to the first road surface data; wherein the preprocessing includes missing value processing, data standardization processing and feature engineering processing; the second road surface data includes color feature data, texture feature data, shape feature data, water molecule absorption peak feature data, road surface material absorption peak feature data, road surface scattering feature data, road surface reflectivity feature data and surface height feature data;

[0159] And / or, the data acquisition processing unit further includes:

[0160] A first data determination module, used for determining and generating missing value status data of samples corresponding to the first road surface data, and performing real-time removal processing or center interpolation processing on the first road surface data according to the missing value status data;

[0161] A first data processing module, configured to standardize the first road surface data based on a minimum-maximum scaling method, and map the processed data to at least one range;

[0162] A second data processing module is used to process the first road surface data according to the feature selection of the LASSO regression method in the embedding method, remove the corresponding redundant feature data, and combine the principal component analysis method to reduce the dimension of the high-dimensional feature data corresponding to the first road surface data;

[0163] And / or, the second data processing module further includes:

[0164] A third data processing module, used for performing zero mean processing on all feature data corresponding to the first road surface data;

[0165] The first calculation generation module is used to calculate and generate a corresponding characteristic covariance matrix; and to process the covariance matrix by eigenvalue decomposition to generate eigenvalue data and eigenvector data corresponding to the covariance matrix; wherein the formula for calculating and generating the corresponding characteristic covariance matrix is:

[0166]

[0167] Among them, x i For each feature in the data, is the mean, T is the transpose of the centered eigenvector; n is the number of samples;

[0168] And / or, the prediction model building unit further includes:

[0169] A second data generating module, configured to divide the second road surface data into at least three corresponding feature data types based on a support vector machine classification learning algorithm;

[0170] A fourth data processing module, used to establish at least three corresponding support vector machine models according to the characteristic data type, and train and process the support vector machine models;

[0171] A third data generation module is used to optimize and process the parameter data corresponding to the support vector machine model in combination with a grid search method, and generate corresponding parameter combination data based on a cross-validation method; wherein the parameter data includes penalty parameter data and kernel function parameter data; and the parameter combination data is the optimal parameter combination data;

[0172] A fifth data processing module, used for evaluating the performance of the training and processing of the prediction model in real time based on the accuracy index data, the precision index data and the recall index data in the test reference data;

[0173] And / or, the third data generating module further includes:

[0174] The fourth data generation module is used to determine and generate the corresponding search range of the penalty parameter C and the kernel function parameter γ with the logarithmic scale as the search range; wherein the range of C is: 2 -5 , 2 -3 , 2 -1 , 2 1 , 2 3 , 2 5 ; Range of γ: 2 -5 , 2 -3 , 2 -1 , 2 1 , 2 3 , 2 5 ;

[0175] The sixth data processing module is used to traverse all possible C and γ combinations within the search range point by point, and use the cross-validation method to evaluate the performance corresponding to each combination; which includes dividing the training set data into K parts, and for each parameter combination, dividing the data into a training set and a validation set, performing K iterations, each time selecting a different part as the validation set, and the remaining K-1 parts as the training set.

[0176] In the system solution embodiment of the present invention, the method steps involved in the detection of slippery road conditions based on machine learning have been described above in detail. That is to say, the functional modules in the system are used to implement the steps or sub-steps in the above method embodiment, which will not be repeated here.

[0177] To achieve the above objectives, the present invention also provides a road slippery state detection platform based on machine learning, such as Figure 6 As shown, it includes a processor, a memory, and a machine learning-based road slippery state detection platform control program; wherein, the machine learning-based road slippery state detection platform control program is executed by the processor, the machine learning-based road slippery state detection platform control program is stored in the memory, and the machine learning-based road slippery state detection platform control program implements the machine learning-based road slippery state detection method steps. For example:

[0178] S1. Acquire first road surface data corresponding to a road surface in a slippery state to be detected, and generate corresponding second road surface data according to the first road surface data; wherein the first road surface data includes video image data, near infrared spectrum data and laser radar data; and the second road surface data is characteristic data related to the road surface in a slippery state to be detected;

[0179] S2. Based on the support vector machine classification learning algorithm, a prediction model corresponding to the road surface state information is constructed in real time, and the prediction model is evaluated and trained in combination with the test reference data;

[0180] S3. Based on the second road surface data and in combination with the prediction model after evaluation and training, generate in real time third road surface data corresponding to the road surface; wherein the third road surface data is road surface slippery state prediction data.

[0181] The specific details of the steps have been explained above and will not be repeated here.

[0182] In the embodiment of the present invention, the built-in processor of the road slippery state detection platform based on machine learning can be composed of an integrated circuit, for example, a single packaged integrated circuit, or a plurality of integrated circuits with the same or different functions, including one or more central processing units (CPU), microprocessors, digital processing chips, graphics processors, and a combination of various control chips. The processor uses various interfaces and lines to connect various components, and executes or executes programs or units stored in the memory, and calls data stored in the memory to perform various functions of road slippery state detection based on machine learning and process data;

[0183] The memory is used to store program codes and various data. It is installed in the road slippery condition detection platform based on machine learning, and realizes high-speed and automatic access to programs or data during operation.

[0184] The memory includes read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electronically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0185] To achieve the above object, the present invention also provides a computer readable storage medium, such as Figure 7 As shown, the computer-readable storage medium stores a road slippery state detection platform control program based on machine learning, and the road slippery state detection platform control program based on machine learning implements the road slippery state detection method steps based on machine learning, for example:

[0186] S1. Acquire first road surface data corresponding to a road surface in a slippery state to be detected, and generate corresponding second road surface data according to the first road surface data; wherein the first road surface data includes video image data, near infrared spectrum data and laser radar data; and the second road surface data is characteristic data related to the road surface in a slippery state to be detected;

[0187] S2. Based on the support vector machine classification learning algorithm, a prediction model corresponding to the road surface state information is constructed in real time, and the prediction model is evaluated and trained in combination with the test reference data;

[0188] S3. Based on the second road surface data and in combination with the prediction model after evaluation and training, generate in real time third road surface data corresponding to the road surface; wherein the third road surface data is road surface slippery state prediction data.

[0189] The specific details of the steps have been explained above and will not be repeated here.

[0190] In the description of the embodiments of the present invention, it should be noted that any process or method description in the flowchart or otherwise described herein may be understood as representing a module, fragment or portion of a code comprising one or more executable instructions for implementing steps of a specific logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations, in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present invention belong.

[0191] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processing module, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM).

[0192] In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways if necessary, and then stored in a computer memory.

[0193] In an embodiment of the present invention, in order to achieve the above-mentioned purpose, the present invention further provides a chip system, wherein the chip system includes at least one processor, and when the program instructions are executed in the at least one processor, the chip system executes the steps of the method for detecting a slippery road surface based on machine learning, for example:

[0194] S1. Acquire first road surface data corresponding to a road surface in a slippery state to be detected, and generate corresponding second road surface data according to the first road surface data; wherein the first road surface data includes video image data, near infrared spectrum data and laser radar data; and the second road surface data is characteristic data related to the road surface in a slippery state to be detected;

[0195] S2. Based on the support vector machine classification learning algorithm, a prediction model corresponding to the road surface state information is constructed in real time, and the prediction model is evaluated and trained in combination with the test reference data;

[0196] S3. Based on the second road surface data and in combination with the prediction model after evaluation and training, generate in real time third road surface data corresponding to the road surface; wherein the third road surface data is road surface slippery state prediction data.

[0197] The specific details of the steps have been explained above and will not be repeated here.

[0198] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application. Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working process of the above-described system, device and unit can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0199] The present invention collects and acquires first road surface data corresponding to a road surface in a slippery state to be detected through a method, and generates corresponding second road surface data based on the first road surface data; wherein, the first road surface data includes video image data, near-infrared spectrum data and lidar data; the second road surface data is feature data related to the road surface in a slippery state to be detected; based on a support vector machine classification learning algorithm, a prediction model corresponding to the road surface state information is constructed in real time, and the prediction model is evaluated and trained in combination with test reference data; based on the second road surface data, and in combination with the prediction model after evaluation and training, third road surface data corresponding to the road surface is generated in real time; wherein, the third road surface data is road surface slippery state prediction data, as well as a system, platform and storage medium corresponding to the method, which can improve the accuracy and robustness of slippery road surface detection.

[0200] In other words, in a specific embodiment of the scheme of the present invention, multi-source data of cameras, near-infrared spectrometers and lidars are combined, and multi-dimensional features (color, texture, shape, water molecule absorption peaks, pavement material absorption peaks, pavement reflectivity, etc.) are extracted through a multi-view learning method, thereby greatly improving the detection accuracy; by using near-infrared spectra and lidars, these sensors are insensitive to light changes and weather conditions, and combined with the SVM classification algorithm, the system has high robustness under various environmental conditions; by extracting pavement geometric features (such as surface height and curvature) through lidar data, combined with multi-source data analysis, the system's adaptability and detection effect in complex road sections are improved.

[0201] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A method for detecting slippery road conditions based on machine learning, characterized in that: The method comprises the steps of: Acquire first road surface data corresponding to the road surface in a slippery state to be detected, and generate corresponding second road surface data according to the first road surface data; wherein the first road surface data includes video image data, near infrared spectrum data and laser radar data; and the second road surface data is characteristic data related to the road surface in a slippery state to be detected; Based on the support vector machine classification learning algorithm, a prediction model corresponding to the road surface state information is constructed in real time, and the prediction model is evaluated and trained in combination with the test reference data; According to the second road surface data and in combination with the prediction model after evaluation and training, third road surface data corresponding to the road surface is generated in real time; wherein the third road surface data is road surface slippery state prediction data.

2. The method for detecting slippery road conditions based on machine learning according to claim 1, characterized in that: The collecting and acquiring first road surface data corresponding to the road surface in the wet and slippery state to be detected, and generating corresponding second road surface data according to the first road surface data, further includes: Preprocess the first road surface data in real time, and extract and generate second road surface data corresponding to the first road surface data; wherein the preprocessing includes missing value processing, data standardization processing and feature engineering processing; the second road surface data includes color feature data, texture feature data, shape feature data, water molecule absorption peak feature data, road surface material absorption peak feature data, road surface scattering feature data, road surface reflectivity feature data and surface height feature data.

3. A method for detecting slippery road conditions based on machine learning according to claim 1 or 2, characterized in that: The collecting and acquiring first road surface data corresponding to the road surface in the wet and slippery state to be detected, and generating corresponding second road surface data according to the first road surface data, further includes: Determine and generate missing value status data of samples corresponding to the first road surface data, and perform real-time removal processing or central interpolation processing on the first road surface data according to the missing value status data; Standardizing the first road surface data based on a minimum-maximum scaling method, and mapping the processed data to at least one range; The first road surface data is processed according to the feature selection of the LASSO regression method in the embedding method to remove the corresponding redundant feature data, and the high-dimensional feature data corresponding to the first road surface data is processed by dimensionality reduction combined with the principal component analysis method.

4. The method for detecting slippery road conditions based on machine learning according to claim 3, characterized in that: The step of combining the principal component analysis method with the high-dimensional feature data corresponding to the first road surface data for dimensionality reduction processing also includes: Performing zero-mean processing on all feature data corresponding to the first road surface data; Calculate and generate a corresponding characteristic covariance matrix; and perform eigenvalue decomposition on the covariance matrix to generate eigenvalue data and eigenvector data corresponding to the covariance matrix; wherein the formula for calculating and generating the corresponding characteristic covariance matrix is: Among them, x i For each feature in the data, is the mean, T is the transpose of the centered eigenvector, and n is the number of samples.

5. The method for detecting slippery road conditions based on machine learning according to claim 1, characterized in that: The method of constructing a prediction model corresponding to the road surface status information in real time based on the support vector machine classification learning algorithm and evaluating and training the prediction model in combination with the test reference data also includes: Based on a support vector machine classification learning algorithm, dividing the second road surface data into at least three corresponding feature data types; According to the characteristic data type, establishing at least three corresponding support vector machine models, and training and processing the support vector machine models; Optimizing the parameter data corresponding to the support vector machine model by using a grid search method, and generating corresponding parameter combination data based on a cross-validation method; wherein the parameter data includes penalty parameter data and kernel function parameter data; and the parameter combination data is the optimal parameter combination data; Based on the accuracy index data, precision index data and recall index data in the test reference data, the performance of the prediction model trained and processed is evaluated in real time.

6. The method for detecting slippery road conditions based on machine learning according to claim 5, characterized in that: The method of optimizing the parameter data corresponding to the support vector machine model by combining the grid search method and generating corresponding parameter combination data based on the cross-validation method also includes: Using the logarithmic scale as the search range, determine and generate the corresponding search range of the penalty parameter C and the kernel function parameter γ; where the range of C is: 2 -5 , 2 -3 , 2 -1 , 2 1 , 2 3 , 2 5 ; Range of γ: 2 -5 , 2 -3 , 2 -1 , 2 1 , 2 3 , 2 5 ; Within the search range, all possible C and γ combinations are traversed point by point, and the performance corresponding to each combination is evaluated in combination with the cross-validation method; this includes dividing the training set data into K parts, and for each parameter combination, dividing the data into a training set and a validation set, performing K iterations, each time selecting a different part as the validation set, and the remaining K-1 parts as the training set.

7. A road slippery state detection system based on machine learning, characterized in that: The system is applied to the method for detecting a slippery road surface based on machine learning as claimed in any one of claims 1 to 6, and the system comprises: a data acquisition processing unit, configured to collect and acquire first road surface data corresponding to the road surface in a slippery state to be detected, and to generate corresponding second road surface data according to the first road surface data; wherein the first road surface data includes video image data, near infrared spectrum data and laser radar data; and the second road surface data is characteristic data related to the road surface in a slippery state to be detected; A prediction model building unit, used to build a prediction model corresponding to the road surface state information in real time based on a support vector machine classification learning algorithm, and to evaluate and train the prediction model in combination with test reference data; The prediction data generating unit is used to generate third road surface data corresponding to the road surface in real time according to the second road surface data and in combination with the prediction model after evaluation and training processing; wherein the third road surface data is the road surface slippery state prediction data.

8. The machine learning-based road slippery state detection system according to claim 7, characterized in that: The data acquisition processing unit further includes: A first data generation module is used to preprocess the first road surface data in real time, and extract and generate second road surface data corresponding to the first road surface data; wherein the preprocessing includes missing value processing, data standardization processing and feature engineering processing; the second road surface data includes color feature data, texture feature data, shape feature data, water molecule absorption peak feature data, road surface material absorption peak feature data, road surface scattering feature data, road surface reflectivity feature data and surface height feature data; And / or, the data acquisition processing unit further includes: A first data determination module, used for determining and generating missing value status data of samples corresponding to the first road surface data, and performing real-time removal processing or center interpolation processing on the first road surface data according to the missing value status data; A first data processing module, configured to standardize the first road surface data based on a minimum-maximum scaling method, and map the processed data to at least one range; A second data processing module is used to process the first road surface data according to the feature selection of the LASSO regression method in the embedding method, remove the corresponding redundant feature data, and combine the principal component analysis method to reduce the dimension of the high-dimensional feature data corresponding to the first road surface data; And / or, the second data processing module further includes: A third data processing module, used for performing zero mean processing on all feature data corresponding to the first road surface data; The first calculation generation module is used to calculate and generate a corresponding characteristic covariance matrix; and to process the covariance matrix by eigenvalue decomposition to generate eigenvalue data and eigenvector data corresponding to the covariance matrix; wherein the formula for calculating and generating the corresponding characteristic covariance matrix is: Among them, x i For each feature in the data, is the mean, T is the transpose of the centered eigenvector; n is the number of samples; And / or, the prediction model building unit further includes: A second data generating module, configured to divide the second road surface data into at least three corresponding feature data types based on a support vector machine classification learning algorithm; A fourth data processing module, used to establish at least three corresponding support vector machine models according to the characteristic data type, and train and process the support vector machine models; A third data generation module is used to optimize and process the parameter data corresponding to the support vector machine model in combination with a grid search method, and generate corresponding parameter combination data based on a cross-validation method; wherein the parameter data includes penalty parameter data and kernel function parameter data; and the parameter combination data is the optimal parameter combination data; A fifth data processing module, used for evaluating the performance of the training and processing of the prediction model in real time based on the accuracy index data, the precision index data and the recall index data in the test reference data; And / or, the third data generating module further includes: The fourth data generation module is used to determine and generate the corresponding search range of the penalty parameter C and the kernel function parameter γ with the logarithmic scale as the search range; wherein the range of C is: 2 -5 , 2 -3 , 2 -1 , 2 1 , 2 3 , 2 5 ; Range of γ: 2 -5 , 2 -3 , 2 -1 , 2 1 , 2 3 , 2 5 ; The sixth data processing module is used to traverse all possible C and γ combinations within the search range point by point, and use the cross-validation method to evaluate the performance corresponding to each combination; which includes dividing the training set data into K parts, and for each parameter combination, dividing the data into a training set and a validation set, performing K iterations, each time selecting a different part as the validation set, and the remaining K-1 parts as the training set.

9. A road slippery state detection platform based on machine learning, characterized in that: It includes a processor, a memory, and a road surface slippery state detection platform control program based on machine learning; wherein, the road surface slippery state detection platform control program based on machine learning is executed by the processor, and the road surface slippery state detection platform control program based on machine learning is stored in the memory, and the road surface slippery state detection platform control program based on machine learning implements the road surface slippery state detection method based on machine learning as described in any one of claims 1 to 6.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a road surface slippery state detection platform control program based on machine learning, and the road surface slippery state detection platform control program based on machine learning implements the road surface slippery state detection method based on machine learning as described in any one of claims 1 to 6.

Citation Information

Cited By

  • Structured road type identification method and system

    CN121121683A