Ultra-long subsea tunnel road section vehicle positioning and traffic state prompting method and system

Abstract

The invention provides a super-long subsea tunnel road section vehicle positioning and traffic state prompting method and system, and relates to the technical field of road state perception, and the method comprises the steps: obtaining a road infrared image and a hyperspectral image, employing a strategy of first target detection and then local fusion, carrying out the hyperspectral registration and cross-modal fusion of an infrared detection structure abnormal region, and obtaining a hyperspectral image; fine matching with a database standard target is carried out, and high-precision and high-robustness tunnel road section positioning is realized in combination with front and back frame verification and an anomaly detection mechanism; and according to the road section numbers and the current timestamps corresponding to the positioning results, historical traffic state data of the current road section and the previous several set road sections and vehicle real-time speeds are obtained, the historical traffic state data of the current road section and the vehicle real-time speeds are jointly input into a state judgment model, and a state judgment result is output. According to the invention, through combination of time-space consistency of front and back frames and a multi-mode interference elimination mechanism, stable, accurate and deployable road section positioning capability in a complex tunnel environment is realized.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of road state perception, in particular to a method and system for vehicle positioning and traffic state prompting in an ultra-long submarine tunnel section. BACKGROUND

[0002] The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute the prior art.

[0003] With the development of intelligent transportation and vehicle perception technology, the accuracy requirements of vehicle positioning and traffic state perception on the road are constantly improving. In the conventional road environment, the positioning system based on GNSS (such as GPS, Beidou), inertial navigation system (INS) and roadside camera and other perception means have been relatively mature. However, in the scenarios of ultra-long submarine tunnels, closed tunnels and elevated bridge spaces, due to factors such as satellite signal shielding, complex electromagnetic environment, poor lighting conditions, etc., the above traditional means often face problems such as low positioning accuracy, signal interruption, visual blur failure, etc., and it is difficult to meet the needs of safe travel and real-time traffic information prompting.

[0004] Currently, some research attempts to introduce vehicle vision aided positioning methods, such as using RGB cameras to obtain front road images and realizing road section matching through image recognition. However, due to the single material of the tunnel road surface, unstable lighting, weak road section differences, RGB images have significant limitations in fine road section division and accurate recognition, especially in night or low visibility environments, the stability and robustness of recognition are poor.

[0005] In addition, in the existing image recognition system, the possible foreign interference factors (such as falling objects, oil stains, construction obstructions, etc.) in the road environment are generally ignored. Such interference will directly affect the image feature extraction and positioning accuracy, increase the risk of misjudgment and omission, and reduce the engineering applicability of the system. Although some positioning enhancement schemes (such as laying markers, laying sensors, etc.) can improve the recognition accuracy to some extent, they usually involve modification of the road structure or additional devices, which have high deployment costs and poor universality, and are not suitable for most existing tunnels and complex environments.

[0006] With the development of deep learning, the existing traffic prediction method and device are provided, which trains a prediction model according to historical traffic data of a road section, first period traffic data and first period traffic data of adjacent road sections, and realizes prediction of the traffic state of the road section, but the method is realized under the condition that the position of the road section is known, and does not involve real-time positioning of the road section. Secondly, an existing method provides a road vehicle auxiliary positioning method based on road surface relief identification, which realizes vehicle positioning by laying relief identification on the road surface according to the regular up and down of the vehicle during driving. But this method is realized by laying relief identification on the road section to realize vehicle positioning, which cannot realize nondestructive testing and has no function of predicting traffic flow. There is also a method of fusing hyperspectral images and infrared images based on a graph Laplacian model, which uses local kernel ridge regression to construct the non-linear mapping relationship between images, and realizes the generation of fused images through manifold regularization and energy minimization. But this method only focuses on the image fusion algorithm itself and does not involve key steps such as image alignment and registration; at the same time, this method is a theoretical construction and does not face specific application scenarios. SUMMARY

[0007] To solve the above problems, the present disclosure provides a long subsea tunnel road section vehicle positioning and traffic state prompting method and system, which constructs a standard road section database, screens candidate regions with structural differences based on echo response characteristics, then performs target detection on the continuous infrared image of the candidate region based on a target detection algorithm of structural saliency rules, extracts the ROI region with road structure characteristics through the thermal radiation continuity and local gradient change of the infrared image, and matches it with the hyperspectral image characteristics to realize high-precision current road section recognition and positioning; fuse vehicle wheel speed information and historical traffic data, realize comprehensive evaluation of the current and front road section traffic state based on a space-time analysis model, and realize driving assistance feedback through image prompt and voice broadcast mode.

[0008] According to some embodiments, the present disclosure adopts the following technical solutions: The long subsea tunnel road section vehicle positioning and traffic state prompting method comprises: A predetermined road section image-traffic state data index library is constructed; Real-time road infrared images and hyperspectral images are acquired, and preprocessed, the preprocessed infrared texture image and spectral feature image are spliced, and the overlapping area registration is performed using a feature point-based splicing RANSAC algorithm to obtain a continuous infrared image and a continuous spectral image; The candidate regions with structural differences in the road surface structure are screened based on an echo response mechanism, the continuous infrared image of the candidate region is detected based on a target detection algorithm of structural saliency rules, and the ROI region with road structure characteristics is extracted through the thermal radiation continuity and local gradient change of the infrared image; Based on the ROI area and the spatial mapping relationship between infrared and hyperspectral cameras, registration is performed with the corresponding area of ​​the hyperspectral image as the benchmark to obtain infrared-hyperspectral target area pairs. Cross-modal feature fusion is performed on the infrared-hyperspectral target area pairs to obtain local target feature vectors. The fusion value is evaluated using the quantitative index of cross-modal feature complementarity to obtain key fusion features. Perform target-level feature matching between the key fusion features and the pre-generated local target feature vectors in the data index library, calculate the comprehensive matching score, and select the database section number with the highest score as the preliminary positioning result; According to the road section number and current timestamp corresponding to the positioning result, the historical traffic status data of the current road section and the previous set road sections, as well as the real-time speed of the vehicle, are obtained. The historical traffic status data of the current road section and the real-time speed of the vehicle are jointly input into the state judgment model, the state judgment result is output, and a state prompt is given.

[0009] According to some embodiments, the present disclosure adopts the following technical solutions: The vehicle positioning and traffic status prompt system for ultra-long submarine tunnel sections includes: Initialization module, used to pre-build a set road section image-traffic status data index library; The image acquisition and preprocessing module is used to acquire real-time road infrared images and hyperspectral images, and preprocess them. The preprocessed infrared texture map and spectral feature image are spliced ​​together, and the overlapping area is registered using the feature point-based splicing RANSAC algorithm to obtain continuous infrared images and continuous spectral images. The positioning module is used to screen candidate areas with structural differences in the pavement structure based on the echo response mechanism, perform target detection on the continuous infrared images of the candidate areas using a target detection algorithm based on structural significance rules, and extract the ROI area with road structure characteristics through the thermal radiation continuity and local gradient changes of the infrared image; based on the ROI area and the spatial mapping relationship between the infrared and hyperspectral cameras, alignment is performed with the corresponding area of ​​the hyperspectral image as a benchmark to obtain infrared-hyperspectral target area pairs, and cross-modal feature fusion is performed on the infrared-hyperspectral target area pairs to obtain local target feature vectors. The fusion value is evaluated using a quantitative indicator of cross-modal feature complementarity to obtain key fusion features; target-level feature matching is performed on the key fusion features with the pre-generated local target feature vectors in the data index library, and a comprehensive matching score is calculated. The database section number with the highest score is selected as the preliminary positioning result; The state judgment module is configured to obtain historical traffic state data of the current road section and the previous set road sections and a real-time speed of the vehicle according to the road section number corresponding to the positioning result and the current timestamp, input the historical traffic state data of the current road section and the real-time speed of the vehicle into a state judgment model, output a state judgment result, and perform state prompting.

[0010] According to some embodiments, the present disclosure adopts the technical solutions as follows: A computer program product comprises a computer program, which, when executed by a processor, implements the vehicle positioning and traffic state prompting method for an ultra-long submarine tunnel road section.

[0011] According to some embodiments, the present disclosure adopts the technical solutions as follows: A non-transitory computer-readable storage medium is configured to store computer instructions, which, when executed by a processor, implement the vehicle positioning and traffic state prompting method for an ultra-long submarine tunnel road section.

[0012] According to some embodiments, the present disclosure adopts the technical solutions as follows: An electronic device comprises a processor, a memory, and a computer program; the processor is connected with the memory, and the computer program is stored in the memory; when the electronic device is running, the processor executes the computer program stored in the memory, so that the electronic device implements the vehicle positioning and traffic state prompting method for an ultra-long submarine tunnel road section.

[0013] Compared with the prior art, the present disclosure has the following beneficial effects: The super-long submarine tunnel road section vehicle positioning and traffic state prompting method of the present disclosure fuses ultrasonic, infrared and hyperspectral multi-modal information, and has the comprehensive advantages of high precision, low redundancy and strong robustness. Infrared and hyperspectral images are synchronously collected, road surface structure and spectral features are extracted, and multiple frames of images are spliced to enhance expression. The road surface structure difference is detected in real time by ultrasonic waves, and the candidate regions with significant features are preferentially selected as the trigger condition for the image fusion process. The "target detection first, then local fusion" strategy is innovatively adopted, and only the hyperspectral registration and cross-modal fusion are performed on the structure abnormal regions to avoid the interference of the whole image redundancy. The fusion result is matched with the standard target in the database at the structure level, and the front and rear frame verification and abnormal detection mechanism are combined to realize high-precision and strong-robustness tunnel road section positioning. Based on the identification of the road section number and the current time, the historical traffic state is queried, and the real-time vehicle speed calculated by the wheel speed encoder is combined to judge whether there is congestion, high risk or critical fluctuation state. According to the judgment result, driving prompts are generated, including visual identification and voice broadcast. Each prompt event records the road section number, vehicle speed, prompt type and user feedback for subsequent system optimization. The system also regularly performs data resampling and accuracy evaluation to ensure that the road section recognition and traffic judgment accuracy is stable above 80%, and if necessary, the database can be rebuilt by resampling.

[0014] The super-long submarine tunnel road section vehicle positioning and traffic state prompting method of the present disclosure is mainly applied to closed road environments where positioning signals such as GPS are weak, lighting and environmental perception are limited, and is especially suitable for special road conditions such as super-long tunnels and submarine tunnels where traditional navigation methods cannot work effectively, to realize accurate road section positioning of vehicles, real-time traffic state prompting and intelligent warning of abnormal conditions. The road surface is collected by the vehicle-mounted infrared thermal imager and hyperspectral camera in a long-distance and non-contact manner, without the need to lay physical markers or sensors on the road surface, and without the need to rely on road reconstruction or modification, to realize efficient perception and positioning support for road section features, with good engineering applicability and deployment flexibility.

[0015] The super-long submarine tunnel road section vehicle positioning and traffic state prompting method of the present disclosure uses the temperature distribution characteristics of infrared images and the material reflection characteristics of hyperspectral images to construct a multi-modal fusion model, extracts weak but stable "invisible" physical differences (such as thermal inertia and spectral response) between road sections, effectively overcomes the recognition difficulty of RGB images in low-light and high-similarity road surface environments, and significantly improves the positioning robustness and accuracy in complex scenes such as tunnels.

[0016] The super-long submarine tunnel road section vehicle positioning and traffic state prompting method of the present disclosure proposes a "target-guided dynamic local fusion" mechanism, breaking through the limitations of traditional whole-image fusion methods in terms of computing efficiency and anti-interference ability. The system detects abnormal target areas in the infrared image and guides the hyperspectral image to perform sub-pixel level registration and cross-modal feature fusion in the key area, significantly improving the relevance and effectiveness of the fusion. At the same time, the introduction of thermal-spectral joint confidence score and feature complementarity evaluation ensures that the fusion area has actual discrimination ability and avoids invalid information interference. This method combines spatiotemporal consistency matching and multi-modal interference elimination mechanism to achieve more stable, accurate and deployable road section positioning capability in complex tunnel environment.

[0017] The super-long submarine tunnel road section vehicle positioning and traffic state prompting method of the present disclosure realizes an accurate synchronous sampling mechanism based on wheel speed encoder. The wheel speed sensor is used to realize equidistant interval sampling of infrared and spectral images instead of equal time interval sampling, solving the influence of vehicle speed change on image sampling uniformity. The image spacing is calculated by formula to ensure a 30% overlap rate, and the encoder pulse is used to control the accurate triggering of infrared and spectral device sampling, realizing accurate hardware-level synchronization at the sampling level and providing a high-precision data basis for subsequent image stitching and comparison.

[0018] The super-long submarine tunnel road section vehicle positioning and traffic state prompting method of the present disclosure proposes a dynamic database self-learning and updating mechanism. The system does not rely on a static template library, but uses a "candidate update pool" to automatically replace the database image when the image quality is higher than the template. An image quality evaluation and historical version management fusion mechanism is designed to realize the self-evolution and self-optimization capability of the database. The dynamically iterative image template library enables the system to adapt to "road condition changes / aging / pollution".

[0019] The super-long submarine tunnel road section vehicle positioning and traffic state prompting method of the present disclosure identifies and locates abnormal areas in the image through joint analysis of temperature features in infrared images and spectral features in hyperspectral images, and uses a mask removal and image repair unit to mask and repair abnormal areas in subsequent image processing to improve the overall recognition robustness and system adaptability. BRIEF DESCRIPTION OF DRAWINGS

[0020] The accompanying drawings, which form a part of this disclosure, are included to provide a further understanding of the disclosure, the illustrative embodiments of the present disclosure, and explanations of the present disclosure, and are not intended to limit the present disclosure.

[0021] Figure 1 The image-traffic state data index library construction flowchart of the present disclosure embodiment is shown in the following figure: Figure 2This is a flow chart of the fusion of infrared images and hyperspectral images according to an embodiment of the present disclosure; Figure 3 A flowchart of road segment positioning according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0022] The present disclosure will be further described below with reference to the accompanying drawings and embodiments.

[0023] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of the present disclosure. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present disclosure belongs.

[0024] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present disclosure. As used herein, unless the context clearly indicates otherwise, the singular form is intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or combinations thereof.

[0025] Example 1 In one embodiment of the present disclosure, a method for positioning a vehicle in an ultra-long submarine tunnel section and providing a traffic status prompt is provided, comprising the following steps: Step 1: Pre-build a set road section image-traffic status data index library; Step 2: Acquire real-time road infrared images and hyperspectral images, preprocess them, stitch the preprocessed infrared texture map and spectral feature image, and use the feature point-based stitching RANSAC algorithm to align the overlapping areas to obtain continuous infrared images and continuous spectral images; Step 3: Based on the echo response mechanism, candidate areas with structural differences in the pavement structure are screened. Target detection is performed on the continuous infrared images of the candidate areas using a target detection algorithm based on structural saliency rules. The ROI region with road structural characteristics is extracted based on the thermal radiation continuity and local gradient changes of the infrared images. Step 4: Based on the ROI area and the spatial mapping relationship between the infrared and hyperspectral cameras, registration is performed with the corresponding area of ​​the hyperspectral image as the benchmark to obtain the infrared-hyperspectral target area pair. Cross-modal feature fusion is performed on the infrared-hyperspectral target area pair to obtain the local target feature vector. The fusion value is evaluated using the cross-modal feature complementarity quantitative index to obtain the key fusion feature.

[0026] Step five: Perform target-level feature matching between the key fusion features and the pre-generated local target feature vectors in the data index library, and calculate the comprehensive matching score. The road section number with the highest score is selected as the preliminary positioning result. Step six: According to the road section number corresponding to the positioning result and the current timestamp, obtain the historical traffic state data of the current road section and the previous several set road sections, and the real-time speed of the vehicle. The historical traffic state data of the current road section and the real-time speed of the vehicle are jointly input into the state judgment model, and the state judgment result is output, and a state prompt is given.

[0027] As an embodiment, the super-long submarine tunnel road section vehicle positioning and traffic state prompting method of the present disclosure filters out candidate regions with structural differences in the road surface structure based on the echo response mechanism. The ultrasonic detection mechanism of structural reflection characteristics is used to preliminarily judge whether there is a structural abnormal region in the current road section, which serves as the trigger condition for the subsequent image fusion process. On this basis, a target detection algorithm based on structural saliency rules is further used to perform target detection on the continuous infrared image of the candidate region. Through the thermal radiation continuity and local gradient change of the infrared image, the ROI region with road structure characteristics is extracted. The "target detection first, then local fusion" strategy is innovatively adopted, and only the structural abnormal region is executed for hyperspectral registration and cross-modal fusion, avoiding the interference of the whole image redundancy. The fusion result is matched with the standard target in the database, and compared with the pre-constructed standard road section database, to realize high-precision current road section recognition and positioning. According to the positioning result, the current speed and the historical traffic data, the current and front road section traffic states are comprehensively evaluated, and the driving assistance feedback is given through image prompt and voice broadcast. The specific implementation process is as follows: Step 1: Pre-construct a set road section image-traffic state data index library, including: Divide the set road section into road section units, number each road section unit, obtain road images, road spectrum data and historical traffic data, and the historical traffic data includes vehicle speed and congestion probability. Bind each image sequence with road section number, infrared image sequence, hyperspectral image sequence, acquisition time, corresponding road stake position signal and historical traffic data, and construct an image-traffic state data index library corresponding to road section number, image spectrum data and historical traffic data.

[0028] Specifically, step 1-1: The tunnel is equally divided according to a fixed length, and each 50 meters is set as a road section unit, which is numbered in turn. The numbering format is Tunnel-XXX, where XXX is an integer number (such as Tunnel-001, Tunnel-002…). Each numbered section is bound with its start and end mileage coordinate value as the main index of the image and traffic information of the section.

[0029] Step 1-2: Set up the infrared image acquisition unit and the spectral image acquisition unit behind the image acquisition vehicle body. Among them, the thermal infrared imager is used to capture the road surface temperature distribution and structural texture, and the hyperspectral camera is used to capture the spectral reflection characteristics of the road surface. The two devices are coaxially installed, the field of view is calibrated to overlap, the unified perspective ensures the consistency of the sampling area.

[0030] Step 1-3: The image acquisition device is installed at a height of , and the inclination angle is , then the ground projection length (coverage length) D of a single frame image is:

[0031] In order to ensure the accuracy of subsequent image continuous splicing, it is necessary to ensure that there is a certain overlap rate between the front and rear images. Assuming that the image overlap rate is 30%, the continuous sampling interval is:

[0032] That is, the vehicle advances X meters to complete the acquisition of infrared images and hyperspectral images, ensuring continuous splicing without breakpoints.

[0033] Step 1-4: To achieve consistent image spacing and synchronous sampling, the device is equipped with an optical wheel speed encoder installed on the non-driven wheel, which outputs a pulse signal. The encoder has the following characteristics: it outputs N pulses per revolution; the pulse frequency and cumulative pulse number can be read by the microcontroller; combined with the vehicle tire radius R, the vehicle travel distance can be accurately calculated. The vehicle advances a distance L, the encoder outputs pulses.

[0034] According to the pre-set sampling interval X, the current cumulative pulse total in the controller is ΣP, and every time: that is, a sampling event is triggered, and then the controller counter is reset to zero to start the next sampling period.

[0035] Step 1-5: The controller outputs a 5ms high-level pulse through the GPIO interface, connected to the Trigger IN of the infrared thermal imager and the Trigger IN of the hyperspectral camera. Both devices are set to "external trigger mode" and sample immediately after receiving the signal, achieving millisecond-level synchronization and avoiding misalignment or spatial mismatch caused by time deviation.

[0036] Step 1-6: Bind the following data items for each collected image sequence: road section number Tunnel-XXX, infrared image sequence, hyperspectral image sequence, image acquisition time and corresponding road stake position information, historical traffic data (average passing vehicle speed; congestion probability; historical accident or abnormal event frequency). Among them, the time label is composed of [week + holiday / weekday + time period], and the flat peak or valley period is divided by 15 minutes, and the peak period is divided by 5 minutes.

[0037] Thus, the process of pre-constructing the image-traffic state data index library of the designated road section is completed.

[0038] Step 2: Obtain real-time road infrared images and hyperspectral images, and pre-process them. The pre-processed infrared texture map and spectral feature image are spliced, and the overlapping area is registered using the feature point-based splicing RANSAC algorithm to obtain continuous infrared images and continuous spectral images. Specifically, considering the actual tunnel environment, the visual differences of different road sections are very limited - the road surface material is uniform, the lighting conditions are similar, and traditional RGB visual information is almost impossible to distinguish each 50-meter road section. However, there are still differences in the road surface at the fine-grained level, such as: local structural texture changes (construction joints, repair layers); material aging degree, roughness, pollutant residue; asphalt oxidation, dust deposition, water stain distribution, etc.; "invisible" spectral reflection differences or thermal inertia differences on the road surface. Such differences show certain regularity in infrared images and hyperspectral images, with distinguishing degree. Therefore, infrared image acquisition units and spectral image acquisition units are installed at the rear of the vehicle to collect infrared images and hyperspectral images, and the method described in step 1 is used to ensure consistent image spacing and field of view, facilitating subsequent image splicing. Through fusion algorithm, the two kinds of image information are deeply fused, thereby completing the identification of the road section.

[0039] First, step 2-1: Pre-process the obtained infrared image, use a high-pass filter (such as Laplacian) for edge enhancement, highlight texture joints, cracks and other structures, and use a temperature gradient enhancement algorithm (difference temperature map), which calculates the temperature difference of each pixel with its neighborhood pixels to generate a local thermal heterogeneity map. Perform histogram equalization on the local thermal heterogeneity map (Histogram Equalization) to unify the brightness difference at different acquisition times and enhance the contrast, and finally output a 1-channel normalized infrared texture map .

[0040] Further, the hyperspectral image is pre-processed, and principal component analysis (PCA) is used to reduce the dimension of the image: the original hyperspectral image is denoted as , Wavelength; using PCA to decompose the spatial covariance of all bands, extract the first three principal components Corresponding to the maximum variation direction respectively, these three principal components form a pseudo-RGB image, which is used as the final spectral feature image .

[0041] Step 2-2: Image stitching, since one frame of image (covering an area of only a few meters) is not enough to express the characteristics of the entire section, image stitching is used to enhance the recognition effect. In order to avoid too much calculation overhead caused by 50m full section stitching, the number of stitching frames is set to N, according to the single frame image coverage length D and the sampling interval in step 1, the total coverage length of the stitched image is:

[0042] That is, the stitched image is about times the single frame coverage length.

[0043] As an embodiment, the specific steps of image stitching include: using vehicle wheel speed encoder to record the driving distance, obtaining the current frame and the previous N-1 frames of images according to the image sampling order, using feature point-based stitching RANSAC algorithm for overlapping area registration, performing vertical direction stitching according to the ground driving direction, and outputting continuous infrared image , continuous spectral image .

[0044] Step 2-3: Screening out candidate areas with structural differences in the road surface structure based on echo response mechanism.

[0045] Specifically, a candidate area screening mechanism driven by sound waves is used, which uses a simple ultrasonic sensor installed at the bottom of the vehicle to emit short pulse high frequency sound wave signals (such as narrow band pulses of about 40 kHz) vertically to the road surface of the current driving path. By measuring the time delay distribution, amplitude change and frequency spectrum characteristics of the echo signal, the system can identify the subtle differences in road material structure. For example, when the sound wave encounters areas with different density, material or surface roughness (such as joints, repairs, slight cracks, embedded objects, etc.), the echo signal will show characteristics such as energy reflection enhancement, abnormal attenuation or frequency spectrum distortion. The system uses a lightweight anomaly detection algorithm (such as statistical change rate threshold judgment in a sliding window) to quickly analyze the sound wave echo signal. Once it is determined that there is a difference in sound wave response in the current area, it is determined that the location has structural heterogeneity.

[0046] Such structural heterogeneity is not equivalent to serious diseases, but part of the normal variation in conventional pavement structure, but it is significantly different from the acoustic response characteristics of adjacent areas, that is, it can be used as an "image processing trigger point". The system then starts the fusion processing flow of the infrared image and hyperspectral image in this area, skips the redundant areas without change, significantly reduces invalid processing, and improves the efficiency of sensing resource scheduling and the response speed of the processing chain.

[0047] By real-time detection of pavement structure differences by ultrasonic waves, candidate areas with significant features are preferentially selected as the trigger condition for image fusion processing.

[0048] Step 2-4: Target area extraction and dynamic ROI area generation based on structural saliency rules. In the position preliminarily determined by the ultrasonic module to have structural differences, the system further detects the target in the infrared image of the corresponding candidate area driven by physical priori. The present disclosure proposes a "structural saliency guided" target detection method without deep model, completely based on the physical priori of infrared image, to realize low-cost and high-robustness target extraction in complex tunnel scenes. The core idea of this method is: through the continuity of thermal radiation and local gradient change of infrared image, the areas with road structure characteristics (such as cracks, joints, oil stains, potholes, strong light reflection, etc.) are excavated, and a set of dynamic and high-confidence ROI areas are constructed for subsequent fusion processing. The specific steps are as follows: (1) Preliminary saliency region extraction: input the continuous infrared image corresponding to the candidate area , calculate its temperature gradient image to depict the local thermal change intensity of the pixel points. Set a threshold , extract the pixel set with temperature gradient higher than the threshold as the initial mask of the candidate target area. This process does not require learning parameters and relies entirely on thermal structural saliency.

[0049] (2) Structure candidate region clustering: perform connected region extraction on the temperature gradient salient pixel set obtained in step (1) , use the connected component analysis method based on four-neighborhood or eight-neighborhood rules to identify the spatially connected thermal salient regions, and define each connected region as a candidate target region . Then calculate the statistical properties (such as area, temperature mean, temperature standard deviation, boundary aspect ratio, main direction angle, etc.) of each region, and can combine edge operators (such as Laplacian or Canny) to enhance the continuity and stability of the boundary profile.

[0050] (3) Multi-index target saliency scoring: for each candidate target region , define a multi-factor confidence score function, which considers the thermal gradient mean , regional temperature distribution concentration ( ), boundary smoothness, spatial size ratio and other information. The specific scoring function is as follows:

[0051] in, Score the saliency of the 𝑖th target region; is the mean temperature gradient of the target area; is the ratio of the regional temperature standard deviation to the mean, which is used to characterize the concentration of heat distribution; is the boundary smoothness factor, which reflects the continuity and regularity of the target contour; It is the aspect ratio or area ratio of the region, which is used to constrain the spatial geometry; is the experience weighting factor, satisfying The scoring function does not require training, and the threshold and weighting factors are determined by empirical rules and domain knowledge, which has good interpretability.

[0052] (4) Dynamic ROI generation: based on confidence score and set threshold , filter out a set of high-confidence target areas Each region is generated based on structural significance and has stable spatial positioning and physical meaning.

[0053] The method disclosed above achieves the precise extraction of high-confidence target regions within the structural change regions initially screened by ultrasound using infrared structural saliency rules, effectively avoiding the ineffective processing of redundant background regions. It also achieves step-by-step convergence from acoustic anomaly perception to infrared image structural analysis, building a low-cost, highly robust local perception chain and laying the foundation for subsequent cross-modal fusion and tunnel section identification. Its output can serve as the sole input for subsequent hyperspectral registration and cross-modal fusion, fundamentally improving fusion effectiveness and reducing ineffective computational overhead.

[0054] Step 2-5: Local area extraction and fine registration. Based on the target area set detected in step 2-4 , further perform local area mapping and sub-pixel registration, the specific process is: (1) For each infrared target area ,According to the spatial mapping relationship between infrared and hyperspectral cameras obtained in the previous calibration, the corresponding target area position in the hyperspectral image is determined; (2) Based on the corresponding area of ​​the hyperspectral image, the mutual information maximization registration algorithm is performed in a small scale to accurately search for the local optimal displacement vector so that the target area of ​​the infrared image is strictly aligned with the corresponding area of ​​the hyperspectral image; (3) After registration, a precisely aligned infrared-hyperspectral target area pair is formed:

[0055] Step 2-6: Dynamic local fusion mechanism of target guidance. Cross-modal feature fusion is performed on the registered target region pairs in step 2-5, which significantly improves the fusion efficiency and regional relevance. The specific steps are as follows: (1) In the target region of the infrared image, extract feature information based on thermal radiation characteristics, including: temperature distribution statistics (such as regional average temperature, temperature standard deviation, range), thermal distribution concentration (such as entropy or energy statistics), and thermal edge structure features obtained by edge detection algorithms (such as Canny or Sobel) (such as edge length, main direction angle, curvature distribution, shape complexity, etc.). The above features are used to describe the thermal and geometric properties of the thermal anomaly region; (2) In the region corresponding to the infrared target region in the hyperspectral image, the principal component analysis (PCA) method is used to extract the statistical features (such as mean, standard deviation) of the first several principal component images (for example, the first 3-5 principal component images), and the reflectance features of the representative key bands (such as typical material absorption peaks) are extracted from the original spectral data, and the spectral angle (Spectral Angle Mapper, SAM) between the standard material spectral library is calculated as the material identification feature, which comprehensively describes the spectral information features in the region; (3) The feature information of the two kinds of data is spliced to form the fused local target feature vector:

[0056] Wherein, represents the temperature features of the target region in the infrared image, including regional average temperature, temperature standard deviation, range, and thermal distribution concentration, etc. is the edge structure feature of the target region, such as edge length, main direction angle, average curvature, and complexity, which reflects its geometric shape; represents the spectral features of the corresponding region in the hyperspectral image, including the mean statistical value of the principal component image, the reflectance of the specific key band, and the spectral angle (SAM) between the standard material spectrum; is the geometric structure information of the target, including the area, aspect ratio, and barycenter coordinates, etc.

[0057] Further, the above local target feature vector set is denoted as:

[0058] Further, to further improve the effectiveness of the fusion region, the present disclosure designs a cross-modal feature complementarity quantification index after constructing the local fusion features, which is used to dynamically evaluate the fusion value of each target region, avoiding information redundancy or invalid fusion. Specifically, a thermal-spectral coupling coefficient As a complementary evaluation function, it is defined as follows:

[0059] Wherein, represents the temperature gradient map of the infrared target region, represents the intensity gradient map of the hyperspectral image principal component, is the Pearson correlation coefficient. Approaching 0 indicates that the infrared and hyperspectral modalities have large information difference and strong complementarity in the region; and if is too high, it indicates that the two modalities tend to be the same, and the fusion value is limited. The system can set a complementarity threshold , when , the region can be determined as a fusion redundant region, which can be selectively down-weighted or removed, to ensure that the subsequent matching and positioning link only relies on the key region features with cross-modal gain, improving the overall recognition stability and anti-interference ability.

[0060] Unlike traditional full-image-level fusion strategies, the present disclosure proposes a "target-guided dynamic local fusion" mechanism. Through pre-sequence infrared image target detection, a high-confidence ROI set is generated in real time, and only it is subjected to hyperspectral registration and fusion operation, and a thermal-spectral joint confidence score mechanism is introduced to determine the fusion weight and reliability of each region, reducing redundant calculation and improving the effectiveness and anti-interference ability of fusion.

[0061] Step 2-7: Target matching and positioning based on regional fusion features. Based on the key region feature set obtained in step 2-6, the pre-generated target region fusion features of each road segment in the standard database are matched and identified: (1) The standard database pre-constructs the fusion feature set of each road segment in advance ; (2) Perform target-level feature matching to determine the feature distance (such as cosine distance or Euclidean distance) of each target in the current frame to the most similar target in the database; (3) Construct the target graph (with target regions as nodes and spatial position relationships between nodes as edges) from the feature matching results; (4) Considering the feature similarity and spatial structure consistency between targets, calculate the comprehensive matching score:

[0062] Wherein, is the feature similarity score, which represents the inverse distance between each fusion feature vector in the current image frame and the most similar target in the database, which can be calculated by weighted average of cosine similarity or Euclidean distance, reflecting the consistency of the fusion feature itself; For the spatial structure consistency score, it represents the similarity between the current frame and the target region in the standard road segment in terms of spatial arrangement (such as node distance, relative angle, topological structure, etc.), which can be realized by a structure similarity measurement function between node graphs. For the matching completeness score, it represents the matching coverage degree between the target set in the current frame and the standard template, for example, the ratio of the number of successfully matched targets to the total number of targets; α, β, γ are the weight coefficients of the three score components respectively, used to balance the influence of feature similarity, structure consistency and matching coverage in the total score, satisfying α+β+γ=1.

[0063] Finally, the database road segment number with the highest score is selected as the preliminary positioning result.

[0064] Step 2-8: To further improve the positioning robustness and accuracy, if the preliminary positioning result has uncertainty or the score is critical, the candidate region expansion verification is performed: (1) Starting from the region where the preliminary matching target is located, the search for other structural feature targets is expanded to the adjacent region; (2) The expanded region is subjected to target detection→cross-modal fusion→feature matching process again; (3) The spatial structure consistency between the expanded target regions is verified to confirm the stability and consistency of the feature matching before and after; (4) If the verification is passed, the preliminary matching result is confirmed; otherwise, the matching process is re-executed.

[0065] Step 2-9: Joint comparison mechanism of front and rear frames. After the positioning of the current frame image, the positioning result of the previous frame (the previous sampling point) is obtained; if the current segment and the previous segment simultaneously satisfy the following conditions: the similarity with the corresponding numbered segment image in the database is higher than the set threshold (such as 85%); the numbering of the previous segment and the current segment is continuous in the numbering order; the time stamp and the driving distance are consistent (judged by the wheel speed encoder); the comparison result of the continuous splicing image is higher than the unified threshold; it is considered that the positioning result is reliable, and it is confirmed as the tunnel road segment number where the current vehicle is located. If only one of the current segment or the previous segment matches, one more frame of data needs to be involved in the judgment (sliding window mechanism).

[0066] Step 2-10: Interference elimination and abnormal image processing mechanism. In actual tunnel road surface images, there are often interference factors such as dropped objects, construction equipment and vehicle projections, which may destroy the local structure or spectral features and affect the positioning accuracy. To enhance the anti-interference ability of the system, the present disclosure designs the following multi-stage cooperative processing process: Interference detection method: In the infrared image, the temperature difference threshold method is used to detect local abnormal heat sources (such as patches with a temperature difference of more than 3°C), morphological processing (opening and closing operations) is used to remove small noise points, and only suspicious foreign object areas are retained; In the hyperspectral image, based on spectral angle mapping (SAM) or anomaly detection algorithm RXDetector, abnormal spectral area is identified, and cosine similarity comparison is performed with the normal road spectral library, and if it is lower than the threshold (such as 0.85), it is abnormal; If the abnormal areas detected by the infrared image and the spectral image overlap in space, they are marked as "effective foreign object areas".

[0067] Mask removal and image repair: Generate a mask for the detected foreign object area; In the image comparison stage, ignore the mask area, and only perform image matching on the unobstructed part; If the area of the abnormal area accounts for more than a threshold (such as 15%) of the image area, the current frame image is invalid and the identification is skipped; If the area is small, the system can use edge interpolation based on image context (such as Navier-Stokes repair) or texture filling algorithm (such as PatchMatch) to complete the occluded area, and improve the image integrity.

[0068] S2-11: Database dynamic update mechanism. After successful positioning (satisfying double-segment continuous comparison, no occlusion), the current image pair enters the "candidate update pool"; If the clarity and texture intensity are higher than the original template, it is automatically updated; The new image is marked with a timestamp to form a traceable evolution history; A four-dimensional data structure of "time-road segment-image-traffic" is constructed for subsequent trend analysis or learning.

[0069] Step 3: According to the road segment number corresponding to the positioning result and the current timestamp, the historical traffic state data of the current road segment and the previous several set road segments are obtained, and the real-time speed of the vehicle is obtained, the historical traffic state data of the current road segment and the real-time speed of the vehicle are combined and input into the state judgment model, and the state judgment result is output., and state prompt.

[0070] Specifically, step 3-1: Obtain the current positioning result as the road segment number Tunnel-XXX, and obtain the current timestamp T_now, and parse the time label, which includes the week label, such as Monday to Sunday; Or holiday label: such as weekday / official holiday; And time period label: extract the current segment (such as 08:30~08:45) according to the set granularity (such as 15 minute segment).

[0071] In the traffic database, take {Tunnel-XXX + date label + time period} as the joint index, extract the historical traffic state information of the current segment in this period, including but not limited to: average speed, congestion probability, average passing time, abnormal event frequency, and lane effective traffic capacity estimate. Query the data of the next 2-3 segments to predict the traffic trend.

[0072] Step 3-2: Based on the pulse signals output by the photoelectric wheel speed encoder installed at the bottom of the vehicle for each wheel rotation, combined with the tire radius R and the number of pulses per revolution N, the distance traveled by the vehicle per unit time can be calculated, thus estimating the vehicle speed in real time:

[0073] where, is the number of pulses in the time window; is the sampling period (e.g., 1s); is the instantaneous speed of the vehicle.

[0074] Step 3-3: Input the average vehicle speed , congestion probability , and real-time vehicle speed of the current section into the state judgment model, which sets the judgment logic to determine whether there is an abnormal state or a prompt event, including: (1) Congestion or high-risk state judgment: If one of the following conditions is met, it is determined that the current state is congested or high-risk: 1) and the duration is ≥10s; 2) The congestion probability of the current section ; 3) Accident frequency in the current section or downstream section ≥3 times / week; 4) The measured passing time is more than 30% higher than the historical average.

[0075] (2) Smooth traffic judgment: 1) and the historical vehicle speed of the two consecutive rear sections is ≥30km / h; 2) Congestion probability ≤20%, accident frequency <1 / week; (3) Critical fluctuation state: The current section is normal but the congestion probability of the downstream section increases.

[0076] Step 3-4: Generate multi-modal prompt content based on the comprehensive judgment result: (1) Visual prompt (center display screen): Display the current state with road section number and recognition image, red for congestion or high-risk state, yellow for critical state, and green for smooth traffic state.

[0077] (2) Voice prompt: For example, "You are currently located in Tunnel-026 section, the historical congestion probability of this period is 68%, the current vehicle speed is low, it is recommended to slow down and observe the situation in front of you." Or "The next two sections are expected to be smooth, and the speed is expected to be maintained above 40km / h." As an embodiment, each prompt event is cached locally, recording the following information items: section number, prompt category (congestion / abnormality / suggestion), timestamp, current speed, user feedback (such as "ignore prompt", "prompt accurate"). This feedback information can be used for later model optimization and background data training. Support offline caching at the vehicle end + regular return fusion.

[0078] The detection vehicle (collection vehicle) periodically performs complete driving in the target tunnel according to a fixed period (such as once a month), collects a complete set of image and actual positioning, traffic judgment data, and uses it as a self-detection sample of the system. The system compares the current section recognition result of the detection vehicle with the actual known section; at the same time, it compares the traffic state prediction with the real vehicle speed / traffic time, and calculates: section positioning accuracy; traffic state judgment accuracy. If any index <80%, it is marked as "system degradation", and step 1 is entered to re-sample and update the database.

[0079] Embodiment 2 In an embodiment of the present disclosure, a vehicle positioning and traffic state prompting system for an ultra-long submarine tunnel section is provided, comprising: The vehicle positioning and traffic state prompting system for an ultra-long submarine tunnel section comprises: An initialization module for pre-constructing a fixed section image-traffic state data index library; An image acquisition and preprocessing module for acquiring real-time road infrared images and hyperspectral images, and preprocessing them, splicing the preprocessed infrared texture map and spectral feature image, and using a feature point-based splicing RANSAC algorithm for overlapping area registration to obtain continuous infrared images and continuous spectral images; The positioning module is used for screening out candidate regions with structural differences in the road surface structure based on an echo response mechanism, performing target detection on a continuous infrared image of the candidate region based on a target detection algorithm of structural saliency rules, extracting an ROI region with road structure characteristics through thermal radiation continuity and local gradient change of the infrared image; based on the ROI region and based on a spatial mapping relationship of the infrared-hyperspectral camera, registering a corresponding region of a hyperspectral image as a benchmark to obtain an infrared-hyperspectral target region pair, performing cross-modal feature fusion on the infrared-hyperspectral target region pair to obtain a local target feature vector, evaluating a fusion value by using a cross-modal feature complementarity quantitative index to obtain a key fusion feature; performing target level feature matching on the key fusion feature and a pre-generated local target feature vector in a data index library, calculating a comprehensive matching score, and selecting a database road section number with the highest score as a preliminary positioning result. The state judgment module is used for obtaining historical traffic state data of a current road section and previous set road sections and a real-time speed of the vehicle according to the road section number corresponding to the positioning result and a current timestamp, inputting the historical traffic state data of the current road section and the real-time speed of the vehicle into a state judgment model, outputting a state judgment result, and performing state prompting.

[0080] Embodiment 3 In an embodiment of the present disclosure, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the vehicle positioning and traffic state prompting method for an ultra-long submarine tunnel road section.

[0081] Embodiment 4 In an embodiment of the present disclosure, a non-transitory computer readable storage medium is provided, which is used for storing computer instructions, and the computer instructions, when executed by a processor, implement the vehicle positioning and traffic state prompting method for an ultra-long submarine tunnel road section.

[0082] Embodiment 5 In an embodiment of the present disclosure, an electronic device is provided, comprising a processor, a memory and a computer program; wherein the processor is connected with the memory, and the computer program is stored in the memory; when the electronic device is running, the processor executes the computer program stored in the memory, so that the electronic device performs the vehicle positioning and traffic state prompting method for an ultra-long submarine tunnel road section.

[0083] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flow or blocks. Figure 1 one or more flow or blocks.

[0084] These computer program instructions can also be loaded into a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flow or blocks. Figure 1 one or more flow or blocks.

[0085] Although the present disclosure has been described with reference to the embodiments thereof, it is apparent that a variety of modifications or changes can be made thereto without departing from the scope of the present disclosure.

Claims

1. A method for positioning vehicles and providing traffic status notification in an ultra-long submarine tunnel section, characterized in that: include: Pre-build a set road section image-traffic status data index library; Real-time road infrared images and hyperspectral images are acquired and preprocessed to produce continuous infrared images and continuous spectral images. Candidate areas with structural differences in the road surface structure are screened based on the echo response mechanism. Target detection is performed on the continuous infrared images of the candidate areas using a target detection algorithm based on structural saliency rules. ROI areas with road structural characteristics are extracted based on the continuity of thermal radiation and local gradient changes in the infrared images. Based on the ROI region and the spatial mapping relationship between the infrared and hyperspectral cameras, registration is performed with the corresponding area of ​​the hyperspectral image as the benchmark to obtain infrared-hyperspectral target region pairs. Cross-modal feature fusion is performed on the infrared-hyperspectral target region pairs to obtain local target feature vectors. The fusion value is evaluated using a quantitative index of cross-modal feature complementarity to obtain key fusion features. Target-level feature matching is performed on the key fusion features with the pre-generated local target feature vectors in the data index library, and a comprehensive matching score is calculated. The database section number with the highest score is selected as the preliminary positioning result. According to the road section number and current timestamp corresponding to the positioning result, the historical traffic status data of the current road section and the previous set road sections, as well as the real-time speed of the vehicle, are obtained. The historical traffic status data of the current road section and the real-time speed of the vehicle are jointly input into the state judgment model, the state judgment result is output, and a state prompt is given.

2. The method for vehicle positioning and traffic status prompting in an ultra-long submarine tunnel section according to claim 1, characterized in that: The set road section is divided into sections, and each section unit is numbered in sequence. Road images, road surface spectral data, and historical traffic data are obtained. The historical traffic data includes vehicle speed and congestion probability. Each image sequence is bound to the section number, infrared image sequence, hyperspectral image sequence, acquisition time, corresponding road pile position signal, and historical traffic data to construct an image-traffic status data index library corresponding to the section number, image spectral data, and historical traffic data.

3. The method for vehicle positioning and traffic status prompting in an ultra-long submarine tunnel section according to claim 1, characterized in that: Real-time road infrared images and hyperspectral images are acquired, and the infrared images are preprocessed. A high-pass filter is used for edge enhancement to highlight texture seams and crack structures. A temperature gradient enhancement algorithm is used to calculate the temperature difference between each pixel and its neighboring pixels to generate a local thermal heterogeneity map. Histogram equalization is used to unify the brightness differences at different acquisition times, enhance contrast, and finally output a one-channel normalized infrared image. The hyperspectral image is preprocessed and principal component analysis is used to decompose the spatial covariance of all bands. The first three principal components are extracted and combined into a pseudo-RGB image as the final hyperspectral image.

4. The method for vehicle positioning and traffic status prompting in an ultra-long submarine tunnel section according to claim 1, characterized in that: Ultrasonic sensors installed under the vehicle detect the reflective properties of the road surface structure and screen candidate areas with structural differences in real time. A target detection algorithm based on structural saliency rules performs target detection on continuous infrared images. The continuous infrared image of the candidate area is input, its temperature gradient map is calculated, a threshold is set, and a set of pixels with temperature gradients above the threshold is extracted as the initial mask for the candidate target area. Connected region clustering is performed on the pixel set to identify a series of spatially continuous target regions and calculate their statistical properties. For each target region, a multi-factor confidence scoring function is defined, which comprehensively considers its thermal gradient mean, regional temperature distribution concentration, boundary smoothness, and spatial size ratio information. Based on the confidence score and the set threshold, a set of high-confidence target regions is screened out.

5. The method for vehicle positioning and traffic status prompting in an ultra-long submarine tunnel section according to claim 1, characterized in that: The specific process of obtaining the local target feature vector is as follows: (1) for each infrared target area, the corresponding target area position in the hyperspectral image is determined based on the infrared-hyperspectral camera spatial mapping relationship obtained in the previous calibration; (2) based on the corresponding area of ​​the hyperspectral image, the mutual information maximization registration algorithm is performed in a small scale to accurately search for the local optimal displacement vector so that the target area of ​​the infrared image is strictly aligned with the corresponding area of ​​the hyperspectral image; (3) after registration, a precisely aligned infrared-hyperspectral target area pair is formed; (4) cross-modal feature fusion is performed on the registered target area pair to extract the feature information in the infrared target area, including temperature distribution statistics and thermal edge structure features; feature information in the corresponding hyperspectral area is extracted, and the feature information of the two data is spliced ​​to form the fused local target feature vector.

6. The method for vehicle positioning and traffic status prompting in an ultra-long submarine tunnel section according to claim 1, characterized in that: The historical traffic status data of the current road section and the real-time vehicle speed are jointly input into the state judgment model. The specific execution logic of the state judgment model includes: If any of the following conditions is met, the vehicle is considered to be in a congested or high-risk state: 1) , and the duration is ≥10s; 2) Congestion probability of the current segment ; 3) The frequency of accidents in the current section or downstream section is ≥ 3 times / week; 4) The measured passing time is more than 30% higher than the historical average; Smooth traffic judgment: 1) The historical speed of the two consecutive rear sections is ≥30km / h; 2) Congestion probability ≤ 20%, accident frequency < 1 / week; Critical fluctuation state: The current segment is normal but the congestion probability of the downstream segment increases.

7. The vehicle positioning and traffic status prompt system for ultra-long submarine tunnel sections is characterized by: include: Initialization module, used to pre-build a set road section image-traffic status data index library; Image acquisition and preprocessing module, used to acquire real-time road infrared images and hyperspectral images, and preprocess them to obtain continuous infrared images and continuous spectral images; The positioning module is used to screen candidate areas with structural differences in the road surface structure based on the echo response mechanism. The target detection algorithm based on the structural significance rule performs target detection on the continuous infrared images of the candidate areas. The ROI area with road structure characteristics is extracted by analyzing the thermal radiation continuity and local gradient changes of the infrared image. Based on the ROI region and the spatial mapping relationship between the infrared and hyperspectral cameras, registration is performed with the corresponding area of ​​the hyperspectral image as the benchmark to obtain infrared-hyperspectral target region pairs. Cross-modal feature fusion is performed on the infrared-hyperspectral target region pairs to obtain local target feature vectors. The fusion value is evaluated using a quantitative index of cross-modal feature complementarity to obtain key fusion features. Target-level feature matching is performed on the key fusion features with the pre-generated local target feature vectors in the data index library, and a comprehensive matching score is calculated. The database section number with the highest score is selected as the preliminary positioning result. The status judgment module is used to obtain the historical traffic status data of the current section and the previous set sections, as well as the real-time speed of the vehicle based on the section number corresponding to the positioning result and the current timestamp. The historical traffic status data of the current section and the real-time speed of the vehicle are jointly input into the status judgment model, the status judgment result is output, and the status prompt is provided.

8. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the method for positioning a vehicle in an ultra-long submarine tunnel section and providing a traffic status prompt is implemented as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium, characterized in that The non-transitory computer-readable storage medium is used to store computer instructions. When the computer instructions are executed by the processor, the method for vehicle positioning and traffic status prompting in an ultra-long submarine tunnel section as described in any one of claims 1 to 6 is implemented.

10. An electronic device, characterized in that: include: A processor, a memory, and a computer program; wherein the processor is connected to the memory, and the computer program is stored in the memory. When the electronic device is running, the processor executes the computer program stored in the memory to enable the electronic device to implement the method for vehicle positioning and traffic status prompting in an ultra-long undersea tunnel section as described in any one of claims 1 to 6.

Images