Tunnel portal dark ice monitoring and early warning method, early warning system and electronic device

Abstract

The present application relates to the field of tunnel entrance road monitoring, in particular to a tunnel entrance dark ice monitoring and early warning method, an early warning system and electronic equipment, the early warning system comprising a data acquisition module, a data preprocessing module, a detection module, an analysis module and an early warning module, the tunnel road surface condition is acquired by using a hyperspectral camera, spatial information and spectral information are combined, real-time monitoring and early warning of the tunnel entrance dark ice are realized, and the safety of the tunnel entrance road traffic accident-prone section is greatly improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of tunnel portal road monitoring, and particularly relates to a tunnel portal dark ice monitoring and early warning method, an early warning system and an electronic device. BACKGROUND

[0002] As a key component in the traffic network, the tunnel carries important social and economic significance. Due to the characteristics of the tunnel, such as the closed nature, strong restriction and narrow space, the complexity and difficulty of rescue and evacuation in the tunnel are far more than those on the open road once an accident occurs. Therefore, the safe operation of the tunnel is crucial to the traffic safety. However, in the cold winter, the tunnel is often a disaster area of traffic accidents. Due to the large temperature difference between the inside and outside of the tunnel and the relatively low solar radiation inside the tunnel, the ground water vapor at the entrance and exit of the tunnel is easily condensed into ice, forming "dark ice". The dark ice is a colorless and transparent deadly threat, which is almost the same color as the road surface and is difficult to be detected by the driver. If the vehicle encounters dark ice obstacles during high-speed driving, the vehicle will lose control. In addition, dark ice will also affect the braking of the vehicle, so that the driver cannot normally react to cause the vehicle to lose control. Therefore, dark ice will pose a great threat to driving and easily cause traffic accidents. The dark ice at the tunnel portal is an important problem affecting the safety of the tunnel. Therefore, further research on the early warning of the dark ice at the tunnel portal has important practical significance. Dark ice also appears in the following three scenarios: toll station, the vehicles passing through the toll station portal stop and go, the temperature of the vehicles plus the temperature of the tires causes the dark ice to be easily formed after snowfall; bridge and culvert, the temperature is relatively high during the day after snowfall, plus the ventilation under the bridge or culvert, and a thin layer of dark ice is formed on the bridge at night. On the slope, due to the long sunshine time of the slope section, the snow is easily melted, which causes the thin ice to be formed after sunset. The current dark ice early warning method uses the same way to monitor the dark ice in various scenarios, and cannot give the most appropriate judgment information and road conditions to the driving vehicles for the dark ice information at the tunnel portal. SUMMARY

[0003] In order to overcome the above problems or at least partially solve the above problems, the present application provides a tunnel portal dark ice monitoring and early warning method, an early warning system and an electronic device, which can effectively overcome the inaccuracy in the current tunnel portal dark ice early warning, and effectively improve the accuracy of data extraction and the early warning effect.

[0004] In a first aspect, the present application provides a tunnel portal dark ice monitoring and early warning method, comprising the following steps:

[0005] I. Collecting hyperspectral images of the dark ice road and the safe road at the tunnel portal to establish an image database;

[0006] II. Extracting the spectral reflectance of different roads in the image database and performing preprocessing to establish a spectral database; the reflectance in the hyperspectral image is Where θ represents a pixel point of the hyperspectral image, I m (θ) is the original hyperspectral image, I d (θ) is the dark background image, I w (θ) is the standard whiteboard image; the fitting curve y=a o +a1x+a2x 2 +...+a m x m As a baseline, the corrected spectral curve is obtained by subtracting the baseline from the original spectral curve;

[0007] III. Real-time acquisition of hyperspectral images of tunnel entrance roads, and dimension reduction operation on image information, as follows:

[0008] H M×N×j is the hyperspectral data of each sample, where M, N represent the length of a single band image, j represents the total number of bands, X i =(X1,X2…,Xj), represents different band images, and all bands are averaged to obtain a new set of features:

[0009]

[0010] The covariance matrix is obtained:

[0011]

[0012] Where each covariance is calculated as:

[0013]

[0014] The eigenvalues and eigenvectors of C are:

[0015] Cu-λu=0

[0016] The largest k eigenvalues and their corresponding k eigenvectors are selected, and the k eigenvectors are used as new projection axes to obtain the k-dimensional features after dimension reduction. For each sample, the new features are:

[0017]

[0018] IV. Real-time monitoring of abnormal road surface data using a target detection neural network on road surface image information; comparing abnormal road surface data with spectral information in the spectral database, as follows:

[0019] C 1…j is the continuous spectral data of the detected abnormal area, S i is the spectral data of different road surface conditions in the database, then C 1…j is calculated with each S 1-iThe distance of the curve is the closest road surface condition, and the distance formula used is the Euclidean distance, as follows:

[0020]

[0021] Where i represents different road conditions, and j represents the number of wave bands.

[0022] The obtained D i is sorted from small to large, and the minimum D i The corresponding road type is the final analysis result.

[0023] Five, when monitoring the existence of dark ice on the road surface, send a warning message.

[0024] The second aspect of the present application provides a tunnel dark ice monitoring and warning system, comprising:

[0025] The data acquisition module is used for respectively shooting hyperspectral images of the tunnel dark ice road and the safety road, establishing an image database for positive and negative sample images, extracting the spectral reflectivity of different regional road surfaces in the spectral dimension, and preprocessing the spectral data to establish a spectral database.

[0026] The data preprocessing module is used for dimensionality reduction operation on the image dimension information of the tunnel dark ice road collected in real time.

[0027] The detection module uses a target detection neural network to detect the road surface condition in real time, and transmits the detected abnormal road surface data to the analysis module, so that the analysis module further judges whether the road surface is dark ice.

[0028] The analysis module extracts the spectral information of the detected abnormal road surface area, compares it with the spectral information in the database, optimizes the output result of the processing module network, and transmits the analyzed result to the warning module.

[0029] The warning module is used for receiving the analysis result transmitted by the analysis module, and sending a warning message when the analysis module detects the existence of dark ice on the road surface.

[0030] Further, the data acquisition module is composed of a hyperspectral camera and a light source mounted above the tunnel, and the preprocessing method of the spectral curve in step one is as follows:

[0031] The reflectivity in the hyperspectral image is extracted as an evaluation index for classification of different substances, and the specific formula is as follows:

[0032]

[0033] Where θ represents a certain pixel point in the hyperspectral image data, I m(I) is an original hyperspectral image, I d (I) is a dark background image, I w (I) is a standard white plate image;

[0034] A baseline correction method based on polynomial fitting is adopted, and a group of curves is constructed to fit the spectral curve with background noise:

[0035] y=a o +a1x+a2x 2 +…+a m x m

[0036] A series of points (x1, x2,...x i ) on the non-peak value of the known spectral curve are selected, and the square sum of the distance of each point on the spectral curve to the curve is:

[0037]

[0038] The partial derivative of a i is obtained:

[0039]

[0040] In matrix form, we have:

[0041]

[0042] Let XA=Y, then A=X -1 Y

[0043] y=a o +a1x+a2x 2 +…+a m x m is the obtained fitting curve;

[0044] The obtained fitting curve is used as the baseline, and the corrected spectral curve is obtained by subtracting the baseline from the original spectral curve.

[0045] Further, the hyperspectral camera is matched with a light source for light compensation, the spectral range of the light source is between 400-1000 nanometers, and the camera and the light source are carried on the tunnel entrance, for real-time monitoring of the tunnel entrance.

[0046] Further, the dimensionality reduction operation is as follows:

[0047] Assume that there are M samples, H M×N×j is the hyperspectral data of each sample, where M, N represent the length of a single band image, j represents the total number of bands, and X i =(X1,X2…,X j(), representing different band images. First, the mean is removed from all bands to obtain a new set of features:

[0048]

[0049] Find its covariance matrix:

[0050]

[0051] Each covariance is calculated according to the following formula:

[0052]

[0053] The eigenvalues ​​and eigenvectors of C are obtained using the following formula:

[0054] Cu-λu=0

[0055] We select the k largest eigenvalues ​​and their corresponding k eigenvectors, and use these k eigenvectors as the new projection axes to obtain the new k-dimensional features after dimensionality reduction. For each sample, the new features are calculated by the following formula:

[0056]

[0057] Furthermore, the network is an improved YOLOv8 target detection model used to monitor road surface conditions in real time. When an anomaly is detected, the location information of the abnormal area is automatically detected and the location information is output to the analysis module.

[0058] Furthermore, the analysis module classifies the spectral information of the output location to analyze the road surface condition, as follows:

[0059] C 1…j For continuous spectral data of the detected anomalous region, S i Given the spectral data of different road surface conditions in the database, calculate C respectively. 1…j With each S 1-i The distance to the curve is the predicted road surface condition, and the distance formula used is the Euclidean distance, as follows:

[0060]

[0061] Where i represents different road surface conditions and j represents the number of bands.

[0062] The obtained D i Sort the data from smallest to largest, then the smallest D is... i The corresponding road surface type is the result obtained from the final analysis.

[0063] In a third aspect, an electronic device is provided, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the tunnel dark ice monitoring and early warning method according to the first aspect.

[0064] In a fourth aspect, a non-transitory computer-readable storage medium is provided, which stores computer instructions, and the computer instructions are executed by a computer to implement the steps of the tunnel dark ice monitoring and early warning method according to the first aspect.

[0065] The tunnel dark ice monitoring and early warning method provided by the present application uses a hyperspectral camera to collect the road surface conditions of the tunnel, combines spatial information and spectral information, realizes real-time monitoring and early warning of the tunnel dark ice, and has a high promotion effect on the safety of the high-accident section of the tunnel road. BRIEF DESCRIPTION OF DRAWINGS

[0066] Figure 1 FIG. 1 is a flowchart of the tunnel dark ice monitoring and early warning method.

[0067] Figure 2 FIG. 2 is a schematic diagram of the hyperspectral camera mounting method in the tunnel dark ice monitoring and early warning method.

[0068] The figure mark is: 1-mounting device, 2-hyperspectral camera, 3-light source, 4-tunnel asphalt pavement, 5-dark ice.

[0069] Figure 3 FIG. 3 is a schematic diagram of the improved network structure of the feature extraction of the model. DETAILED DESCRIPTION

[0070] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme of the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the embodiments of the present application. In order to solve the defects of the prior art.

[0071] In the face of the problem that the warning without distinction of simply considering road icing and dark ice at the tunnel entrance seriously affects the warning effect, the embodiment of the present application effectively combines the tunnel road surface condition collected by the hyperspectral camera and the joint spatial information and spectral information, and optimizes the corresponding model and method, so as to effectively carry out the dark ice warning at the tunnel entrance, and effectively overcome the inaccuracy in the dark ice warning at the present tunnel entrance, thereby effectively improving the warning effect. The method of the present application will be specifically described and introduced through the following embodiments.

[0072] Figure 1 The flowchart of the hyperspectral-based dark ice monitoring and warning method at the tunnel entrance provided by the embodiment of the present application comprises the following steps:

[0073] S1, constructing an image database and a spectral database;

[0074] The data acquisition module respectively shoots 1000 positive and negative sample images of the road covering the dark ice at the tunnel entrance and the safe road, establishes an image database, divides the data set according to the training quantity: test quantity: verification quantity ratio of 8:1:1, and is used for the subsequent detection of the detection module to detect the road surface anomaly, extracts the spectral reflectivity of the road surface in different areas for the spectral dimension, and pre-processes the spectral data to establish a spectral database;

[0075] The data acquisition module is composed of a hyperspectral camera and a light source mounted on the top of the tunnel, and the pre-processing method of the spectral curve is specifically:

[0076] S1.1, extracting reflectivity;

[0077] Since the DN value does not have a quantitative index in actual application, the reflectivity in the hyperspectral image is extracted as an evaluation index for classification of different substances, and the specific formula is as follows:

[0078]

[0079] Where θ represents a pixel point in the hyperspectral image data, I m (θ) is the original hyperspectral image, I d (θ) is the dark background image, I w (θ) is the standard whiteboard image.

[0080] S1.2, polynomial fitting;

[0081] In order to eliminate the noise interference caused by the light source and the background in the collected dark ice spectral data, a baseline correction method based on polynomial fitting is adopted. First, a group of curves is constructed to fit the spectral curve with background noise:

[0082] y=a o +a1x+a2x 2 +...+am x m

[0083] Select a series of points (x1, x2, … x i ) on the non-peak value of the known spectrum curve, then the square sum of the distance of each point on the spectrum curve to the curve is:

[0084]

[0085] The partial derivative of a i is obtained:

[0086]

[0087] Using matrix representation, we get:

[0088]

[0089] Let XA=Y, then A=X -1 Y

[0090] The obtained y=a o +a1x+a2x 2 +…+a m x m is the fitting curve.

[0091] S1.3, baseline correction;

[0092] Use the obtained fitting curve as the baseline, and subtract the baseline from the original spectrum curve to obtain the corrected spectrum curve.

[0093] S2, collect tunnel entrance road image;

[0094] The hyperspectral image of the road surface state of the tunnel entrance is collected in real time by the data acquisition module, and the data is transmitted to the data preprocessing module in real time;

[0095] As Figure 2 shown, the hyperspectral camera is equipped with a light source for light compensation, and the spectral range of the light source is between 400 and 1000 nanometers. The hyperspectral camera and the light source are carried on the top of the tunnel entrance to monitor the road surface of the tunnel entrance in real time.

[0096] S3, pre-process the image information;

[0097] The pre-processing module receives the image dimension information transmitted by the data acquisition module, which is used for the positioning of the abnormal road surface area by the detection module. Considering the processing capacity of the device, the pre-processing module performs dimension reduction operation on the image information; the dimension reduction method is as follows:

[0098] Assume that there are M samples, H M×N×jFor each sample hyperspectral data, where M, N represent the length of a single band image, j represents the total number of bands, X i = (X1, X2…, X j ), represents different band images. First, the mean of all bands is removed, and a new set of features is obtained:

[0099]

[0100] The covariance matrix is obtained:

[0101]

[0102] where each covariance is calculated according to the following formula:

[0103]

[0104] The eigenvalues and eigenvectors of C are obtained according to the following formula:

[0105] Cu-λu=0

[0106] The largest k eigenvalues and their corresponding k eigenvectors are selected, and the k eigenvectors are used as new projection axes to obtain new k-dimensional features after dimension reduction. The new features for each sample can be calculated according to the following formula:

[0107]

[0108] S4, a target detection neural network is built to detect the condition of the road surface in real time, and the detected abnormal road surface data is transmitted to the analysis module, and the analysis module further judges whether the road surface is caused by the existence of dark ice;

[0109] The built network is an improved yolov8 target detection model, which is used to monitor the road surface condition in real time. When an anomaly is detected, the position information of the abnormal area is automatically detected and output to the analysis module. In order to meet the lightweight deployment requirements of the road monitoring system, the feature extraction backbone network of the model uses MobileNetv1, and the improved network structure is as shown in Figure 3

[0110] The analysis module uses KNN method to classify and process the output position spectrum information, analyzes the condition of the road surface, and specifically as follows:

[0111] C is the continuous spectrum data of the detected abnormal area, S i is the spectrum data of different road conditions in the database, then the distance between C and each S i curve is calculated, and the nearest road condition is the predicted road condition. The distance formula used is the Euclidean distance, and the formula is as follows: ​

[0112]

[0113] Wherein, i represents different road conditions, and j represents the number of wave bands.

[0114] The obtained D i is sorted in ascending order, and the minimum D i The corresponding road type is the final analysis result.

[0115] S5, extracting the spectral information of the detected abnormal road area, comparing the spectral information with the spectral information in the database, optimizing the output result of the processing module network, and transmitting the analyzed result to the warning module;

[0116] S6, receiving the analysis result transmitted by the analysis module, when the analysis module detects that the road surface exists dark ice, sending the warning information to the relevant department or vehicle control system through the warning module, and displaying on the connected road traffic facilities such as road side variable information board, realizing real-time warning of icy road conditions.

[0117] Based on the same concept, the embodiment of the application provides a tunnel portal dark ice monitoring and warning system, which is used for monitoring and warning the tunnel portal dark ice in the above-mentioned embodiments.

[0118] Therefore, the description and definition in the tunnel portal dark ice monitoring and warning method of the above-mentioned embodiments can be used for the understanding of each execution module in the embodiment of the application, and the specific can refer to the above-mentioned embodiments, which will not be repeated here.

[0119] It comprises a data acquisition module for respectively shooting hyperspectral images of the tunnel portal dark ice road and the safe road, establishing an image database for positive and negative sample images, extracting the spectral reflectivity of different area road surfaces in the spectral dimension, and preprocessing the spectral data to establish a spectral database;

[0120] A data preprocessing module is used for dimension reduction operation on the information of the image dimension of the tunnel portal dark ice road collected in real time.

[0121] A detection module uses a target detection neural network to detect the condition of the road surface in real time, and transmits the detected abnormal road surface data to an analysis module, which further judges whether the road surface is caused by the existence of dark ice.

[0122] An analysis module extracts the spectral information of the detected abnormal road area, compares the spectral information with the spectral information in the database, optimizes the output result of the processing module network, and transmits the analyzed result to a warning module.

[0123] A warning module is used for receiving the analysis result transmitted by the analysis module, and sending the warning information when the analysis module detects that the road surface exists dark ice.

[0124] As still another aspect of the embodiments of the present application, the embodiments provide an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the tunnel dark ice monitoring and early warning method according to the above embodiments when executing the computer program.

[0125] The embodiments of the present application also provide a non-transitory computer readable storage medium having computer instructions stored thereon, wherein the computer instructions implement the steps of the tunnel dark ice monitoring and early warning method according to the above embodiments when executed by a computer.

[0126] It can be understood that the above-described embodiments of the apparatus, electronic device and storage medium are merely illustrative, wherein the units described as separate parts may or may not be physically separate, and may be located in one place or distributed on different network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the embodiments. Those skilled in the art can understand and implement without creative labor.

[0127] From the above description of the embodiments, those skilled in the art can clearly understand that the embodiments can be realized by means of software and necessary universal hardware platforms, and of course can also be realized by hardware. Based on such understanding, the above technical solutions can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a U disk, a mobile hard disk, a ROM, a RAM, a magnetic disk or an optical disk, and includes a plurality of instructions to make a computer device (such as a personal computer, a server or a network device) execute the method described in the above method embodiments or some parts of the method embodiments.

[0128] In addition, those skilled in the art should understand that in the application file of the embodiments of the present application, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or device. Without more limitations, the element defined by the statement "including a" does not exclude the presence of other identical elements in the process, method, article or device including the element.

[0129] In the description of embodiments of the application, numerous specific details are recited. However, it should be understood that embodiments of the application can be practiced without these specific details. In some instances, well-known methods, structures and techniques have not been described in detail in order to not obscure the understanding of this description. Similarly, it is to be understood that the various features of the application described herein are sometimes grouped together in a single embodiment, figure or description of a related art for the purpose of streamlining the disclosure and aiding in the understanding of one or more of the various aspects and embodiments of the application. The disclosure of these features in a specific close proximity should not be interpreted as an intention that the inventive aspect is limited to these specifically recited features nor to only one of the specifically recited features.

[0130] However, this disclosure of a method should not be interpreted, nor should the claims be construed, to reflect an intention that the claimed application requires more features than are explicitly recited in each claim. Rather, as is reflected in the claims themselves, the inventive aspects lie in fewer than all of the features of the previously disclosed single embodiments. Accordingly, the claims appended to this detailed description are expressly incorporated herein by reference, wherein each claim by itself is a separate embodiment of the claimed application.

[0131] Finally, it should be noted that the above-mentioned embodiments are merely used to illustrate the technical solutions of the present application, but not limit the present application; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced by equivalent features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for monitoring and early warning of dark ice at a tunnel portal, characterized in that Comprise the following steps: I. Collecting hyperspectral images of tunnel portal dark ice road and safe road to establish an image database; II. Extracting spectral reflectance of different roads from the image database and preprocessing to establish spectral database; the reflectance in the hyperspectral image is where θ represents a pixel point of the hyperspectral image, I m (θ) is the original hyperspectral image, I d (θ) is the dark background image, I w (θ) is the standard whiteboard image; a fitting curve y=a o +a1x+a2x 2 +...+a m x m is used as a baseline, and the corrected spectral curve is obtained by subtracting the baseline from the original spectral curve; III. Real-time collection of hyperspectral images of tunnel portal road, and dimensionality reduction operation on image information, as follows: H M×N×j Hyperspectral data for each sample, where M, N represent the length of the individual band images, j represents the total number of bands, X i = (X1, X2…, X j ), represents different band images, and all bands are de-meaned to obtain a new set of features: Solve its covariance matrix: Wherein, each covariance is calculated according to the formula: The eigenvalues and eigenvectors of C are: Cu-λu=0 Select the largest k eigenvalues and the corresponding k eigenvectors, take the k eigenvectors as new projection axes, and get the k-dimensional features after dimensionality reduction, for each sample, the new features are: IV. Real-time monitoring of abnormal road surface data using target detection neural network; compare the abnormal road surface data with the spectral information in the spectral database, as follows: C 1…j For continuous spectral data of the detected anomalous region, S i Given the spectral data of different road surface conditions in the database, calculate C respectively. 1…j With each S 1-i The distance to the curve is the predicted road surface condition, and the distance formula used is the Euclidean distance, as follows: Wherein, i represents different road conditions, and j represents the number of bands; The obtained D i Sort in ascending order, the minimum D i The corresponding road type is the final analysis result; V. When monitoring that the road surface has dark ice, send warning information to the relevant departments or vehicles.

2. A tunnel portal dark ice monitoring and warning system, characterized in that, Based on the tunnel portal dark ice monitoring and warning method of claim 1, comprising: A data acquisition module for shooting hyperspectral images of tunnel portal dark ice road and safe road, respectively, establishing an image database of positive and negative sample images, extracting the spectral reflectance of different area road surfaces, and preprocessing the spectral data to establish a spectral database; A data preprocessing module for dimensionality reduction operation on the image dimension information of the real-time collected tunnel portal dark ice road; A detection module using a target detection neural network to detect the road surface in real time and transmitting the detected abnormal road surface data to an analysis module for further judgment of whether the road surface has dark ice; An analysis module for extracting the spectral information of the detected abnormal road surface area and comparing it with the spectral information in the database, optimizing the output results of the processing module network, and transmitting the analyzed results to a warning module; A warning module for receiving the analysis results transmitted by the analysis module and sending warning information when the analysis module detects that the road surface has dark ice.

3. The tunnel portal dark ice monitoring and warning system of claim 2, wherein, The data acquisition module is composed of a hyperspectral camera and a light source mounted above the tunnel.

4. The tunnel portal dark ice monitoring and warning system of claim 2, wherein, The preprocessing method of the spectral curve is as follows: The reflectivity is taken as an evaluation index for classification of different substances, and the specific formula is as follows: where θ represents a certain pixel point in the hyperspectral image data, I m (θ) is the original hyperspectral image, I d (θ) is the dark background image, I w (θ) is the standard whiteboard image; A baseline correction method based on polynomial fitting is adopted to construct a group of curves to fit the spectral curve with background noise: y = a o + a1x + a2x 2 +... + a m x m A series of points (x1, x2, … x i ) on the non-peak value of the known spectral curve are selected, and the sum of squares of the distances from the known points on the spectral curve to the curve is: Respectively, a i The partial derivative is obtained: The matrix is represented as: Let XA=Y, then A=X -1 Y y = a o + a1x + a2x 2 +... + a m x m is the resulting fitted curve; The obtained fitting curve is used as the baseline, and the corrected spectral curve is obtained by subtracting the baseline from the original spectral curve.

5. The tunnel portal dark ice monitoring and warning system of claim 2, wherein, The hyperspectral camera is matched with a light source for light supplement, the spectral range of the light source is 400-1000 nanometers, and the camera and the light source are mounted above the tunnel portal to monitor the road surface in real time.

6. The tunnel portal dark ice monitoring and warning system of claim 2, wherein, The dimensionality reduction operation is as follows: Suppose there are M samples, H M×N×j is the hyperspectral data for each sample, where M, N represent the length of the single band image, j represents the total number of bands, X i = (X1, X2…, X j ) represents different band images; first, the mean of all bands is removed, and a new set of features is obtained: Solve its covariance matrix: Wherein, each covariance is calculated according to the formula: The eigenvalues and eigenvectors of C are: Cu-λu=0 Select the largest k eigenvalues and the corresponding k eigenvectors, take the k eigenvectors as new projection axes, and get the k-dimensional features after dimensionality reduction; for each sample, the new features are calculated by the following formula:

7. The tunnel portal dark ice monitoring and warning system of claim 2, wherein, The network is a target detection model of yolov8, which is used for real-time monitoring of the road surface condition, automatically detects the position information of the abnormal area when detecting the abnormality, and outputs the position information to an analysis module.

8. The tunnel portal dark ice monitoring and warning system of claim 2, wherein, The analysis module classifies the spectral information of the output position and specifically as follows: C 1…j For the continuous spectral data of the detected abnormal area, S i For the spectral data of different road surface conditions known in the database, C 1…j The distance of each S 1-i Curve is calculated, and the closest road surface condition is the predicted road surface condition. The distance formula used is the Euclidean distance, as follows: Wherein, i represents different road surface conditions, and j represents the number of wave bands. The obtained D i Sort in ascending order, the minimum D i The corresponding road type is the final analysis result.

9. An electronic device, comprising: The computer program stored on the memory and executable on the processor, when the processor executes the computer program, realizes the steps of the tunnel portal dark ice monitoring and early warning method in claim 1.

10. A computer-readable storage medium, characterized in that, The computer program stored on the memory and executable on the processor, when the processor executes the computer program, realizes the steps of the tunnel portal dark ice monitoring and early warning method in claim 1.

Images