Track image recognition method, system, and storage medium for plateau construction tunnels

Abstract

The present invention relates to the fields of rail transit and image processing technology, and in particular to a method and system for image recognition of track in plateau construction tunnels, as well as a storage medium. The method aims to cope with extreme environments characterized by high temperature, high humidity, and inadequate sensors and lighting facilities. The method comprises installing ultrasonic, short-wave infrared, and hyperspectral imaging equipment at the end of a train transporting materials for multimodal data acquisition, and extracting geometric and spectral features through multi-scale convolution and spectral fusion. The system utilizes spectral-geometric feature collaborative optimization and active learning modules for data processing and model updates, ensuring the model's adaptability and recognition accuracy in changing environments. Finally, path optimization is achieved through real-time prediction and reinforcement learning modules to ensure track recognition accuracy and train operation safety. The present invention significantly improves detection accuracy and system responsiveness in complex plateau environments.

Description

Technical Field

[0001] The present invention relates to the field of rail transportation and image processing technology, and in particular to a method and system for recognizing images of tracks in plateau construction tunnels, and a storage medium. Background Art

[0002] Track image recognition in plateau construction tunnels is crucial for ensuring construction safety and the operational stability of trains transporting supplies. Construction tunnels in my country's Qinghai-Tibet Plateau (over 5,000 meters above sea level) often experience extreme conditions, such as high temperatures and humidity. In the early stages of construction, adequate sensors and lighting are often lacking. This makes accurate identification of tracks and surrounding obstacles extremely difficult, directly impacting train safety and construction progress.

[0003] Existing track obstacle detection technologies (Chinese invention patent, publication number: CN117325918A, title: Method and system for detecting obstacles in tunnels) mostly use trackside sensing units or onboard sensors, such as visual sensors and lidar. While these technologies can provide effective detection in relatively stable environments, they expose numerous drawbacks in the extreme conditions of plateau tunnel construction:

[0004] High temperature and high humidity conditions can affect sensor reliability and data acquisition accuracy. Existing technologies, such as video image and point cloud analysis algorithms, struggle to balance real-time performance and accuracy, resulting in insufficient recognition speed and accuracy. The fusion of video imagery and radar point cloud data often relies on simple synchronization and fusion algorithms, failing to fully optimize and enhance the feature fusion of multimodal data, thus impacting overall detection effectiveness. Summary of the Invention

[0005] To address the numerous issues with the aforementioned existing technologies, the present invention provides a method, system, and storage medium for image recognition of track in high-altitude tunnel construction. This method collects multimodal data using ultrasonic imaging, shortwave infrared, and hyperspectral imaging equipment, and extracts and fuses features using a multiscale convolutional network and spectral fusion. Utilizing an adaptive modeling module and spectral-geometric collaborative optimization, the system dynamically updates model parameters and implements path simulation and optimization through reinforcement learning and variational autoencoders. This invention significantly improves track and obstacle recognition capabilities and response speed in extreme environments.

[0006] A method for recognizing track images in a plateau construction tunnel comprises the following steps:

[0007] Ultrasonic imaging sensors, short-wave infrared sensors, and hyperspectral imaging equipment are installed at the end of the train transporting materials. During the train operation, ultrasonic depth data, optical image data, and spectral feature data in the tunnel are collected, and data synchronization and pre-processing are performed to generate synchronized cleaning data.

[0008] The synchronized cleaned data is input into the multimodal data fusion module, and geometric feature data and spectral enhancement feature data are extracted through multi-scale convolution and hyperspectral fusion, and feature fusion is performed to generate fused feature data;

[0009] The fused feature data is input into the adaptive modeling module, the fused feature data is optimized using the spectral-geometric feature collaborative optimization algorithm, and the model parameters of the adaptive modeling module are updated through adaptive training to generate optimized feature data;

[0010] The optimized feature data is input into the real-time prediction and multi-objective decision-making system, and the latent space distribution is generated through the variational autoencoder. The reinforcement learning module is combined to perform path simulation and path optimization to generate path prediction data and path optimization data.

[0011] The active learning module collects execution feedback of path optimization data and performs active learning and model updating on the adaptive modeling module.

[0012] Preferably, in the data synchronization and preprocessing steps, the ultrasonic depth data, optical image data and spectral feature data are time aligned and noise removed by a time synchronization module.

[0013] Preferably, the multimodal data fusion module adopts a multi-scale convolutional network to extract geometric feature data layer by layer through convolution kernels with different receptive fields, and embeds spectral enhancement feature data in each convolution layer to generate fused feature data.

[0014] Preferably, the spectral-geometric feature collaborative optimization algorithm optimizes the geometric feature data and the spectral feature data using the following calculation formula:

[0015] W opt =W geo +α·W spec

[0016] Among them, the optimized weight combination is W opt ; The geometric feature weight is W geo ; The spectral feature weight is W spec ; The weight adjustment coefficient is α.

[0017] Preferably, during the adaptive training process, the adaptive modeling module introduces a meta-learning mechanism so that the adaptive modeling module automatically adjusts the learning rate and weight parameters according to feedback data.

[0018] Preferably, the real-time prediction and multi-objective decision-making system generates a latent space distribution through a variational autoencoder, and the latent space distribution is described by the following formula:

[0019] p(Z|X)=N(μ(X),σ(X))

[0020] Among them, the latent space variable is Z; the optimized feature data is X; the mean of the feature data is μ(X); and the standard deviation of the feature data is σ(X).

[0021] Preferably, the reinforcement learning module includes a policy network and a value network, the policy network performs path simulation based on the generated latent space distribution, and the value network is used to evaluate the safety and efficiency of the path simulation to generate path prediction data and path optimization data.

[0022] Preferably, when collecting execution feedback of the path optimization data, the active learning module generates labeled feedback data through an automatic labeling algorithm and performs multi-level data cleaning to remove outliers and redundant information.

[0023] A system for implementing the plateau construction tunnel track image recognition method, the system comprising:

[0024] Ultrasonic imaging sensors, short-wave infrared sensors, and hyperspectral imaging equipment installed at the end of the train transporting materials are used to collect ultrasonic depth data, optical image data, and spectral feature data in the tunnel while the train is running;

[0025] A time synchronization module is used to time align and preprocess the collected ultrasonic depth data, optical image data and spectral feature data to generate synchronized cleaning data;

[0026] A multimodal data fusion module, including a multi-scale convolution and hyperspectral fusion unit, is used to extract geometric feature data and spectral enhancement feature data from the synchronous cleaned data, and perform feature fusion to generate fused feature data;

[0027] An adaptive modeling module is configured to receive the fused feature data and perform optimization processing using a spectral-geometric feature collaborative optimization algorithm, and to update model parameters of the adaptive modeling module through adaptive training to generate optimized feature data;

[0028] A real-time prediction and multi-objective decision-making system, including a variational autoencoder and a reinforcement learning module, is used to generate a latent space distribution based on optimized feature data, perform path simulation and path optimization to generate path prediction data and path optimization data;

[0029] The active learning module is used to collect execution feedback of path optimization data and perform active learning and model updates on the adaptive modeling module to improve the recognition accuracy and adaptability of the system in high temperature and high humidity environments.

[0030] A storage medium stores a computer program, which, when executed by a processor, implements the steps of the plateau construction tunnel track image recognition method.

[0031] Compared with the prior art, the advantages and beneficial effects of the present invention are:

[0032] The present invention achieves efficient track and obstacle recognition in extreme environments of high temperature and high humidity through multimodal data fusion and adaptive modeling modules;

[0033] Through the spectral-geometric feature collaborative optimization algorithm and active learning mechanism, the system can adaptively update and improve the recognition performance of the model, ensuring accurate analysis and real-time response of orbital images;

[0034] The present invention realizes path prediction and optimization by introducing a reinforcement learning module and a variational autoencoder, making the system more adaptable and robust in complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 Schematic diagram of the process of the present invention;

[0036] Figure 2 This is a relationship diagram of the multimodal data fusion module in the present invention;

[0037] Figure 3 This is a relationship diagram of the spectrum-geometric feature collaborative optimization module in the present invention;

[0038] Figure 4 This is a diagram showing the relationship between the reinforcement learning module and path optimization in the present invention;

[0039] Figure 5 This is a relationship diagram between active learning and model updating in the present invention;

[0040] Figure 6 It is a structural block diagram of the system of the present invention. DETAILED DESCRIPTION

[0041] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present disclosure. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of well-known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present disclosure.

[0042] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "comprise," "include," etc. used herein indicate the presence of the features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0043] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0044] like Figure 1 As shown, a method for recognizing track images in plateau tunnel construction includes the following steps:

[0045] Ultrasonic imaging sensors, short-wave infrared sensors, and hyperspectral imaging equipment are installed at the end of the train transporting materials. During the train operation, ultrasonic depth data, optical image data, and spectral feature data in the tunnel are collected, and data synchronization and pre-processing are performed to generate synchronized cleaning data.

[0046] Preferably, in the data synchronization and preprocessing steps, the ultrasonic depth data, optical image data and spectral feature data are time aligned and noise removed by a time synchronization module.

[0047] In train tunnel construction scenarios at altitudes above 5,000 meters, the tunnels are characterized by high temperatures and humidity, and pre-construction sensors and lighting facilities are inadequate. Under these conditions, trains transporting supplies require a precise forward obstacle recognition system to ensure safety while operating on the tracks. To achieve this, the present invention proposes installing ultrasonic imaging sensors, short-wave infrared sensors, and hyperspectral imaging equipment at the ends of the trains to collect multi-source environmental data within the tunnel, ensuring that the system can accurately identify the tracks and surrounding obstacles under complex conditions.

[0048] Ultrasonic imaging sensors, shortwave infrared sensors, and hyperspectral imaging equipment are deployed at the train's ends. Each sensor has a specific function: ultrasonic imaging sensors collect geometric depth data of the track, providing precise structural information; shortwave infrared sensors acquire optical image data, capturing morphological details of the track and surrounding environment; and hyperspectral imaging equipment collects spectral signature data to identify the optical properties of different materials and the reflective characteristics of surrounding substances. This combination enables real-time data collection while the train is in motion, providing rich multimodal data for subsequent processing.

[0049] In extreme environments of high temperature and humidity, the temporal and spatial consistency of data is particularly important. To this end, the data synchronization and preprocessing steps utilize a time synchronization module to time-align and remove noise from data from different sensors. The time synchronization module timestamps each data source and uses a timing matching algorithm to ensure data consistency within the same time frame. The core of this step is the application of timing matching formulas, such as:

[0050] T aligned (i) = T sensor (i)-Δt

[0051] Among them, T aligned (i) indicates the timestamp after alignment; T sensor (i) is the original timestamp of sensor data acquisition; Δt is the time difference between sensors.

[0052] The noise removal process uses a median filter to pre-process the ultrasonic depth data to reduce the interference of high-frequency noise; for optical image data, Gaussian filtering is applied to smooth the image to improve the overall image quality and reduce noise; hyperspectral data is spectrally smoothed using a denoising algorithm to improve its accuracy in material identification.

[0053] The high temperature and humidity conditions in plateau tunnels can affect sensor stability and data accuracy. Therefore, the time synchronization module and data processing system of the present invention incorporate an environmental compensation mechanism. During the noise removal process, the preprocessing algorithm includes dynamic parameter adjustments to adapt to real-time environmental changes. For example, when the temperature changes, the algorithm parameter k(T) automatically adjusts according to the ambient temperature T:

[0054]

[0055] Where k0 is the initial parameter; β is the temperature coefficient; T ref This adjustment ensures the stability of data preprocessing under different environmental conditions.

[0056] Through the aforementioned data acquisition and preprocessing steps, this method can generate synchronized clean data under the extreme conditions of plateau tunnels, ensuring the temporal and spatial consistency and quality of the input data, providing a reliable foundation for subsequent multimodal data fusion and feature extraction. This method exhibits strong adaptability and data stability in high-temperature and high-humidity environments, enabling the system to accurately identify tracks and nearby obstacles in complex tunnel environments, ensuring safe train operation.

[0057] The synchronized cleaned data is input into the multimodal data fusion module, and geometric feature data and spectral enhancement feature data are extracted through multi-scale convolution and hyperspectral fusion, and feature fusion is performed to generate fused feature data;

[0058] Preferably, Figure 2 As shown, the multimodal data fusion module adopts a multi-scale convolutional network to extract geometric feature data layer by layer through convolution kernels with different receptive fields, and embeds spectral enhancement feature data in each convolution layer to generate fused feature data.

[0059] In high-altitude tunnel construction environments at altitudes exceeding 5,000 meters, trains face the challenges of high temperature and humidity. This places extremely high demands on the forward obstacle recognition systems of material trains, especially given the lack of comprehensive sensor and lighting infrastructure during early construction. To overcome these challenges, this paper proposes a multimodal data fusion module. This module extracts geometric and spectral enhancement feature data from the tunnel environment through multi-scale convolution and hyperspectral fusion, and then fuses these features to generate highly accurate fused feature data, enabling precise identification and timely response to obstacles on and around the track.

[0060] The synchronized cleaned data generated through data synchronization and preprocessing is input into the multimodal data fusion module. This module uses a multi-scale convolutional network to extract geometric feature data layer by layer, ensuring that features at different scales can be captured and analyzed. The multi-scale convolutional network uses convolution kernels with different receptive fields (such as k1×k1 and k2×k2) to capture geometric features from details to the global level. The formula is as follows:

[0061]

[0062] in, is the geometric feature output of the lth layer; is the convolution kernel weight of the lth layer; X (l-1) Input data for the l-1 layer; b (l) is the bias term; * indicates the convolution operation.

[0063] In each convolutional layer, the hyperspectral data is embedded into the convolutional network to form spectrally enhanced feature data. This spectral enhancement is achieved by pixel-by-pixel embedding, ensuring that the surface material properties of each geometric feature are taken into account. The formula for this step is as follows:

[0064]

[0065] in, represents the enhanced spectral feature, S is the spectral feature of the corresponding position, Represents an embed operation.

[0066] The multimodal data fusion module uses layer-by-layer feature extraction from a multi-scale convolutional network, combined with spectral enhancement features, to enable the system to effectively handle complex optical variations such as reflection and scattering in the high-temperature and high-humidity tunnel environment. This improves the robustness of feature extraction and the ability to discern detail. The ultimate result of feature fusion is fused feature data that integrates geometric and spectral information, exhibiting good spatiotemporal consistency and environmental adaptability.

[0067] This approach ensures that trains can accurately identify obstacles on and around the track during operation. For example, if there are water stains or muddy areas caused by humidity in a tunnel, the system can quickly identify the presence and material properties of these obstacles through a combined analysis of geometric and spectral enhancement features, generating reliable obstacle identification data for real-time decision-making.

[0068] In an experiment simulating a high-altitude tunnel environment, the multimodal data fusion module demonstrated high-precision feature extraction under varying lighting and humidity conditions. Multi-scale convolution captured both detailed and overall structural information, while spectral enhancement helped identify the material of slippery surfaces and debris, enabling the system to make accurate path selection and obstacle avoidance decisions.

[0069] The present invention demonstrates extremely strong robustness in high temperature and high humidity environments, ensuring the stable operation and safety of the train, and meeting the high requirements for accurate obstacle identification and timely response under plateau tunnel construction conditions.

[0070] like Figure 3 As shown, the fused feature data is input into the adaptive modeling module, the fused feature data is optimized using the spectral-geometric feature collaborative optimization algorithm, and the model parameters of the adaptive modeling module are updated through adaptive training to generate optimized feature data;

[0071] In order to achieve efficient detection of tracks and surrounding obstacles, the present invention proposes an adaptive modeling module, which receives fused feature data and optimizes it through a spectral-geometric feature collaborative optimization algorithm. At the same time, an adaptive training method is used to update the model parameters to generate optimized feature data, thereby improving the system's responsiveness in extreme environments.

[0072] The fused feature data generated by the multimodal data fusion module is input into the adaptive modeling module. The adaptive modeling module processes the fused feature data using a spectral-geometric feature collaborative optimization algorithm, focusing on leveraging the correlation between spectral and geometric features to enhance the expressive power of features. Spectral features provide information on material properties and optical characteristics, while geometric features provide information on spatial structure and depth. This collaborative optimization process is performed using the following formula:

[0073] W opt (x,y)=Wgeo (x,y)+α·W spec (x,y)

[0074] Among them, W opt (x,y) represents the optimized weight combination at the pixel position (x,y); W geo (x,y) represents the geometric feature weight; W spec (x, y) represents the spectral feature weight; α is the dynamic adjustment coefficient; it is used to adaptively adjust the impact of spectral features on the optimization process according to environmental changes.

[0075] To overcome the impact of high temperature and humidity on data stability, the adaptive modeling module embeds an environmental compensation factor into the collaborative optimization algorithm. This factor is adjusted using real-time environmental data obtained by temperature and humidity sensors, ensuring that the algorithm maintains optimized stability even under extreme conditions. The dynamic adjustment coefficient α involved in the formula is calculated using the following relationship:

[0076]

[0077] Among them, α0 is the basic adjustment coefficient; β is the environmental adjustment factor; H is the current humidity; H ref is the reference humidity.

[0078] After completing collaborative optimization, the fused feature data is further used for adaptive training. The adaptive modeling module uses an online learning mechanism to automatically adjust the model weights and learning rate based on real-time feedback data to improve the model's stability in changing environments. The online learning mechanism uses a meta-learning algorithm to update parameters, and its formula is:

[0079]

[0080] Among them, θ new is the updated model parameter; θ old is the current model parameter; η is the learning rate; is the gradient loss function based on the feature data X.

[0081] Through spectral-geometric feature collaborative optimization and adaptive training, the adaptive modeling module can adjust the model in real time to adapt to dynamic changes in high-temperature and high-humidity environments. This design ensures that the system can continuously output high-precision optimized feature data under the complex optical and material conditions in the tunnel, improving the robustness and stability of obstacle recognition. For example, when the spectral reflectance characteristics change due to increased humidity in the tunnel, the adaptive adjustment coefficient α can dynamically adjust the weight of the spectral features, allowing the model to maintain high recognition accuracy when dealing with slippery tracks or obstacles of different materials.

[0082] In actual experiments, the module demonstrated outstanding performance in a simulated high-altitude tunnel environment. Through online adaptive training and collaborative optimization, the system achieved stable recognition of tracks and obstacles, effectively reducing false alarms and missed alarms, and ensuring safe train operation and real-time responsiveness. This sophisticated optimization strategy is particularly important during the construction phase, meeting the stringent requirements for real-time image recognition and high adaptability.

[0083] Preferably, the spectral-geometric feature collaborative optimization algorithm optimizes the geometric feature data and the spectral feature data using the following calculation formula:

[0084] W opt =W geo +α·W spec

[0085] Among them, the optimized weight combination is W opt , used for subsequent feature analysis and decision making; the geometric feature weight is W geo , mainly describes the spatial layout of track structure and obstacles; the spectral feature weight is W spec , reflecting the optical and material properties of the object surface; the weight adjustment coefficient is α, which controls the influence of spectral features in the final feature combination.

[0086] In plateau tunnel construction scenarios at altitudes above 5,000 meters, trains transporting supplies operate in complex environments, facing the challenges of high temperature and high humidity, as well as the imperfections of early sensors and lighting facilities. These conditions place extremely high demands on the obstacle recognition system. The system must not only overcome the adverse effects of the environment, but also accurately identify obstacles on and around the track during real-time operation. To achieve this goal, the spectral-geometric feature collaborative optimization algorithm used in this invention optimizes the data by fusing geometric and spectral features to improve the model's adaptability and recognition accuracy in complex environments.

[0087] The core of this optimization algorithm lies in the effective combination of geometric and spectral feature data, leveraging their combined strengths to enhance the recognition system's performance. In practical applications, geometric feature data provides information on the shape and spatial location of tracks and obstacles, while spectral feature data provides information on material properties and surface optical characteristics. Through a collaborative optimization algorithm, these two features complement each other in high-temperature and high-humidity environments, forming a more reliable feature combination.

[0088] To improve the algorithm's adaptability in high-temperature and high-humidity environments, the weight adjustment coefficient α can be dynamically adjusted based on environmental data. For example, when humidity and temperature affect the reliability of spectral data, the system will appropriately reduce the value of α to balance the contributions of geometric and spectral features. This adjustment can be achieved by obtaining real-time humidity HHH and temperature TTT through environmental parameter sensors:

[0089]

[0090] Among them, α0 is the basic adjustment coefficient; γ is the adjustment factor used to reflect the influence of humidity on the spectral feature weight; H ref is the reference humidity value, used to normalize the humidity effect.

[0091] This optimization strategy ensures that the system can dynamically adjust the weights of feature combinations when faced with environmental changes, thereby improving the model's recognition accuracy and stability of tracks and obstacles.

[0092] By applying this collaborative optimization algorithm, the system can accurately distinguish between tracks and obstacles in complex optical environments. For example, the complex reflective properties of slippery track surfaces can easily lead to misidentification using traditional methods. However, this algorithm effectively addresses these optical variations by enhancing the ability of spectral features to identify materials. In tests simulating a high-altitude tunnel environment, when humidity reached high levels, the system automatically adjusted the α value, ensuring that the recognition model maintained a high detection accuracy, thus achieving precise distinction between obstacles and tracks.

[0093] This optimization process generates highly robust and adaptable optimized feature data, which can be used for subsequent path planning and real-time decision-making, enabling safe and efficient train operation under complex construction conditions. This invention demonstrates superior performance in high-temperature and high-humidity environments, meeting the high-precision requirements of image recognition systems under extreme construction conditions.

[0094] Preferably, during the adaptive training process, the adaptive modeling module introduces a meta-learning mechanism so that the adaptive modeling module automatically adjusts the learning rate and weight parameters according to feedback data.

[0095] In high-altitude tunnel construction environments at altitudes exceeding 5,000 meters, trains transporting supplies must operate under extreme conditions of high temperature and humidity, facing the challenges of a complex environment and inadequate sensors and lighting. This places extremely high demands on the train's obstacle recognition system, requiring it to adapt to environmental changes in real time while maintaining high-precision recognition. This paper proposes an adaptive modeling module and introduces a meta-learning mechanism during the adaptive training process, enabling the system to automatically adjust the learning rate and weight parameters based on real-time feedback data to cope with dynamic environmental changes.

[0096] The core of the adaptive modeling module is the dynamic adjustment of model parameters through a meta-learning mechanism. Meta-learning is a technique that "learns how to learn," enabling the model to quickly adapt to new situations based on constantly changing input data. In high-temperature and high-humidity environments such as those found in plateau tunnels, external conditions can affect the stability and quality of sensor data. By introducing a meta-learning mechanism into the adaptive modeling module, the system can analyze feedback data in real time and automatically adjust the model's learning rate and weight parameters based on environmental changes to maintain robustness and efficiency.

[0097] In adaptive training, the model uses the gradient update method to optimize parameters and combines the meta-learning mechanism to achieve dynamic adjustment of the learning rate. The formula is as follows:

[0098]

[0099] Among them, θ new is the updated model parameter; θ old is the current model parameter; η adaptive It is an adaptive learning rate, which is dynamically adjusted according to real-time feedback data; is the gradient loss function based on the input data X.

[0100] Adaptive learning rate η adaptive The adjustment is based on performance indicators in the feedback data, such as the error change rate. The adjustment formula is:

[0101]

[0102] Among them, η0 is the basic learning rate; λ is the adjustment factor used to control the variation of the learning rate; It represents the gradient of the loss function with respect to temperature T, reflecting the impact of environmental changes on model error.

[0103] High temperatures and high humidity in plateau tunnels can cause image blur and abnormal light reflections, which can reduce the system's recognition performance. The adaptive modeling module, through the introduction of a meta-learning mechanism, can rapidly respond to these dynamic environmental changes and adjust the learning rate and model weights in real time. When identifying tracks and surrounding obstacles, the model no longer relies on fixed parameters but instead automatically updates according to environmental conditions to maintain recognition accuracy and stability.

[0104] For example, when the ambient humidity increases and the spectral characteristic data fluctuates, the system adjusts the learning rate η by real-time analysis of the feedback data. adaptive , thereby reducing the training error and optimizing the weight parameter θ. This adjustment enables the system to quickly adapt to data differences caused by humidity changes, maintaining high accuracy and reliability of track identification.

[0105] In experiments, this adaptive modeling module was applied to an image recognition system simulating a high-altitude tunnel environment. Results showed that, after dynamically adjusting the learning rate, the model's recognition accuracy improved by over 15% compared to traditional fixed-parameter models. Furthermore, the model's false positive and false negative rates significantly decreased when subjected to sudden increases in humidity or temperature fluctuations. This performance improvement ensures safe train operation in complex environments and meets the stringent requirements for efficient obstacle recognition during the construction phase.

[0106] like Figure 4 As shown, the optimized feature data is input into the real-time prediction and multi-objective decision-making system, the latent space distribution is generated by the variational autoencoder, and the path simulation and path optimization are performed in combination with the reinforcement learning module to generate path prediction data and path optimization data;

[0107] In the construction environment of plateau tunnels at an altitude of more than 5,000 meters, trains transporting materials face extreme conditions of high temperature and high humidity during operation. This environment requires the track image recognition system to have extremely high accuracy and adaptability. Due to the incompleteness of sensors and lighting facilities in the early stages of tunnel construction, the recognition system needs to overcome the impact of the environment on data quality and system stability in order to identify the track and surrounding obstacles in real time and respond quickly. The present invention proposes inputting optimized feature data into a real-time prediction and multi-objective decision-making system. The system generates a latent space distribution through a variational autoencoder, combines a reinforcement learning module to perform path simulation and path optimization, and ultimately generates path prediction data and path optimization data, thereby ensuring the safety and efficiency of train operation.

[0108] After being processed by the adaptive modeling module, the optimized feature data is input into the real-time prediction and multi-objective decision-making system. The system first uses a variational autoencoder (VAE) to generate a latent space distribution that captures the underlying structure and characteristics of the track and its surrounding environment. The VAE generates a probability distribution of the latent variable Z through two neural networks—an encoder and a decoder. Specifically, the encoder network converts the input feature data X into the latent distribution parameters μ(X) and σ(X):

[0109] Z~N(μ(X),σ(X))

[0110] Where Z represents the latent space variable; μ(X) is the mean of the input feature data; σ(X) is the standard deviation of the input feature data.

[0111] The latent space distribution captures global information about track characteristics and obstacles, allowing the system to infer even with incomplete environmental data. This is particularly important in tunnel environments with high temperature and humidity, where environmental variations can cause partial data distortion or incompleteness. The generated latent space distribution is fed into the reinforcement learning module, which receives this latent space information and performs path simulation and optimization.

[0112] The reinforcement learning module consists of a policy network and a value network. The policy network generates path simulation plans, while the value network evaluates the safety and resource efficiency of each path. The reinforcement learning algorithm uses a reward-based approach to select the optimal path after simulating multiple paths. The path optimization formula is as follows:

[0113]

[0114] Where Q(s,a) represents the value of taking action a in the current state s; R(s,a) is the immediate reward of the current state-action pair; γ is a discount factor used to balance the weights of current and future rewards; s′ and a′ represent the next state and action, respectively.

[0115] During the construction phase, complex tunnel conditions necessitated rapid decision-making in the event of unexpected or unforeseen obstacles. Using a latent space distribution generated by a variational autoencoder, the system was able to predict track characteristics even with incomplete data. Combined with a reinforcement learning module for path simulation and real-time optimization, the system generated timely path prediction and optimization data in the event of sudden track changes, obstacles, or slippery surfaces.

[0116] For example, in high humidity environments, where the track surface is reflective or slippery due to water stains, the variational autoencoder captures the underlying distribution of spectral and geometric features. The reinforcement learning module simulates the feasibility of different routes and selects a path based on safety and stability. The resulting output of path prediction and optimization data can be used to guide train adjustments in real time, avoiding potential safety risks.

[0117] In experiments simulating a high-altitude tunnel environment, this real-time prediction and multi-objective decision-making system demonstrated high adaptability. Under extremely high temperature and humidity conditions, the system maintained a high success rate in route optimization, reducing the risk of misjudgments and route deviations caused by slippery tracks. Experimental results showed that the system's response time in dynamic route planning was reduced by over 20% compared to traditional models, while also improving overall recognition and emergency response accuracy. This makes the invention particularly suitable for high-altitude tunnel construction train operation scenarios, which require efficient image recognition and real-time decision-making.

[0118] Preferably, the real-time prediction and multi-objective decision-making system generates a latent space distribution through a variational autoencoder, and the latent space distribution is described by the following formula:

[0119] p(Z|X)=N(μ(X),σ(X))

[0120] Among them, the latent space variable is Z, which represents the representation of the optimized feature data in the latent space; the optimized feature data is X; the mean of the feature data is μ(X), which reflects the central tendency of the feature data; the standard deviation of the feature data is σ(X), which represents the diffusion of the data distribution.

[0121] In plateau tunnel construction scenarios at altitudes above 5,000 meters, trains transporting materials face the complexity and data collection difficulties brought by high temperature and high humidity environments. This places extremely high demands on the track image recognition system, especially under the conditions of imperfect sensors and lighting facilities in the early stages of construction. The system needs to overcome the influence of harsh environments, accurately identify tracks and surrounding obstacles, and perform image processing and path decision-making in real time. In the present invention, the real-time prediction and multi-objective decision-making system generates a potential space distribution through a variational autoencoder (VAE), providing a solid data foundation for path optimization.

[0122] The variational autoencoder (VAE) is a probabilistic generative model that compresses input data into a latent space through an encoder-decoder structure and then reconstructs the input data from this space. Its core function is to generate a latent space distribution p(Z|X), which represents the mapping of input data X into the latent space. For optimized feature data X collected in a high-altitude tunnel environment, the VAE leverages its statistical properties to generate a latent space distribution, enabling the system to perform path prediction and decision-making even when data is incomplete or the environment is complex and changing.

[0123] This latent space distribution can capture the intrinsic connections and changes in track and obstacle characteristics in high temperature and high humidity environments, ensuring that the model can effectively perform feature reconstruction and path decision-making when the environment is disturbed and sensor data is unstable.

[0124] After inputting the optimized feature data into the real-time prediction and multi-objective decision-making system, the latent space distribution generated by the VAE provides multiple possible predictions for route optimization. The reinforcement learning module uses this latent space distribution to simulate multiple paths and select the optimal path for safe and efficient train operation. The latent space variable Z enables the system to perform efficient route simulation and risk assessment even when the environment changes or data is missing.

[0125] For example, when increased humidity in a tunnel causes increased noise in the optical data, the latent space distribution p(Z|X) can still capture the overall characteristics of the track and obstacles through its mean μ(X) and standard deviation σ(X). This allows the system to generate more stable path prediction and optimization data in real-time response, reducing the risk of misjudgment or path deviation.

[0126] In a simulated high-altitude tunnel environment, the system fed optimized feature data into a variational autoencoder, and the resulting latent space distribution was used by a reinforcement learning module to simulate the path. Results demonstrated that the system maintained efficient recognition and path decision-making under conditions of high humidity and drastic temperature fluctuations, reducing false alarm rates by 15% and shortening overall path planning time by approximately 20%. This approach ensures the system can rapidly adjust and output reliable path optimization data in harsh environments, enhancing the safety and adaptability of train operations.

[0127] The present invention uses the latent space distribution generated by the variational autoencoder, combined with real-time path simulation and optimization technology, to achieve accurate identification and real-time response to tracks and obstacles, meeting the complex needs of the plateau tunnel construction stage.

[0128] Preferably, the reinforcement learning module includes a policy network and a value network, the policy network performs path simulation based on the generated latent space distribution, and the value network is used to evaluate the safety and efficiency of the path simulation to generate path prediction data and path optimization data.

[0129] In high-altitude tunnel construction environments above 5,000 meters, train operation faces extreme conditions, such as high temperature and humidity, and a pre-existing environment with inadequate sensors and lighting. This places extremely high demands on track image recognition systems. The system must overcome these limitations in complex environments, accurately identifying tracks and surrounding obstacles in real time and responding efficiently. This is achieved through a reinforcement learning module comprised of a policy network and a value network. These modules simulate and optimize paths based on latent space distributions, generating path prediction and optimization data.

[0130] The core of the reinforcement learning module lies in the collaborative work of the policy network and the value network to ensure the system's efficiency and safety in real-time track path planning. A latent space distribution, generated by a variational autoencoder, provides a high-dimensional representation of track characteristics. The policy network uses this distribution as input to simulate paths, while the value network evaluates the safety and efficiency of paths to optimize decision-making.

[0131] The policy network is a deep neural network-based structure that receives a latent space variable Z as input and outputs a series of path selection actions. The network is optimized using a policy gradient algorithm to maximize the expected cumulative reward. The calculation formula for path simulation is:

[0132]

[0133] Among them, π(a|Z) represents the probability of taking action a given the latent space variable Z; Q(Z,a) represents the value of taking action a in path selection; a′ represents all possible actions.

[0134] The policy network performs path simulation based on this formula and selects the optimal path sequence to improve the system's response speed to track changes and obstacles.

[0135] The value network is used to evaluate the safety and efficiency of each path simulation. It evaluates the path's multi-objective indicators and outputs the path's cumulative value. The value function is defined as follows:

[0136]

[0137] Among them, V(Z) represents the cumulative value of the path under the latent space variable Z; R t is the immediate reward obtained at time step t, taking into account the safety and efficiency of the path; γ is the discount factor used to balance the importance of future rewards and current rewards.

[0138] The value network incorporates environmental characteristics such as track slipperiness, obstacle distances, and real-time optical data into its assessment. By incorporating environmental feedback, the value network accurately assesses the safety and expected outcomes of a path.

[0139] By combining the policy network and the value network, the reinforcement learning module can simulate multiple possible paths in the complex environment of plateau tunnels and optimize the selection based on real-time data. For example, in high-temperature and high-humidity environments, when the tunnel is slippery or poorly lit, resulting in increased optical data noise, the policy network can quickly generate multiple path simulations based on inputs distributed in the latent space, while the value network selects the optimal path through real-time evaluation, ensuring safe and efficient train operation even in deteriorating conditions.

[0140] In a practical implementation, the reinforcement learning module was tested in a simulated high-altitude tunnel environment. The test demonstrated that, through the synergistic effect of the policy network and the value network, the system was able to quickly respond to complex environmental changes and identify and avoid obstacles in route selection. The route prediction and optimization data generated by the system demonstrated a 30% reduction in response time, while track recognition stability and route planning accuracy increased by over 20%. This performance significantly enhances the system's adaptability to extreme environments and train safety.

[0141] Through this reinforcement learning module design, the system not only achieves efficient path simulation and optimization, but also ensures the stability of train operation in the special environment of plateau tunnels, meeting the needs of accurate identification and response under complex construction conditions.

[0142] The active learning module collects execution feedback of path optimization data and performs active learning and model updating on the adaptive modeling module.

[0143] Preferably, Figure 5As shown, when collecting execution feedback of path optimization data, the active learning module generates labeled feedback data through an automatic labeling algorithm and performs multi-level data cleaning to remove outliers and redundant information.

[0144] In high-altitude tunnel construction environments at altitudes exceeding 5,000 meters, trains face significant obstacles in identifying obstacles due to challenges such as high temperature and humidity, as well as inadequate sensors and lighting during initial construction. This places extremely high demands on the real-time and accuracy of the track image recognition system. To improve the system's responsiveness and adaptability, the present invention introduces an active learning module to collect feedback from execution of path optimization data and to update and optimize the adaptive modeling module. This design enables the system to continuously learn in a dynamic environment and enhance its recognition and decision-making capabilities.

[0145] The core of the active learning module lies in its continuous collection and processing of path optimization data, which it uses to actively learn and update the adaptive modeling module. Active learning is a machine learning strategy that selectively labels and learns from new data samples, enabling more efficient model updates and optimization. Specifically, within the complex operating environment of a high-altitude tunnel, this module monitors the path optimization data output by the system, collects feedback from its execution, and adjusts the model based on this feedback.

[0146] The active learning module extracts execution feedback data from the output of the real-time path simulation and optimization process. This data includes the actual path the train took on the track, identified obstacles, and the performance of the path decision. The collected execution feedback data is processed by an automatic labeling algorithm to generate labeled feedback data to assist in model training. The labeling algorithm automatically labels the data for subsequent supervised learning of the model. The labeling process uses the following algorithms:

[0147] y i =f(x i )+∈ i

[0148] Among them, y i is the annotation feedback data; x i is the input path data; ∈ i is the noise term, reflecting the potential disturbance in the environment.

[0149] To ensure the reliability of model training data, the active learning module performs multi-stage data cleaning to remove outliers and redundant information. This multi-stage data cleaning process includes the following steps: primary data denoising, outlier detection, and redundancy removal. Primary denoising uses a median filter to process feedback data to eliminate high-frequency noise. Outlier detection uses statistical methods, where data with a standard deviation exceeding a certain threshold is considered an anomaly. Redundancy removal is achieved by checking the similarity of time series data to reduce redundant model training.

[0150] After collecting and cleaning the execution feedback data, the active learning module inputs it into the adaptive modeling module for model updating. Model updating uses a meta-learning mechanism to adjust model parameters such as learning rate and weights by analyzing the feedback data. The calculation expression for adaptive model updating is as follows:

[0151]

[0152] Among them, θ new represents the updated model parameters; θ old is the current model parameter; η update The learning rate adjusted from the active learning module; is the gradient loss function with respect to the feedback data y.

[0153] This active learning mechanism enables the system to continuously perform adaptive optimization and learning in complex high-temperature and high-humidity environments, improving track and obstacle recognition accuracy. For example, when recognition accuracy decreases in a tunnel due to humidity or temperature fluctuations, the active learning module can quickly collect new execution feedback and generate feedback data for learning through an automatic labeling algorithm. After analyzing and cleaning this data, the system updates model parameters, allowing the adaptive modeling module to quickly adjust to the new environmental conditions.

[0154] In experiments, the active learning module was applied to a simulated plateau tunnel environment. Results showed that, based on multiple rounds of feedback, the system improved recognition accuracy by 10% and reduced path optimization time by 15%. This enhanced model learning and updating capability enables the system to maintain real-time response and efficient recognition in harsh environments, meeting the stringent requirements for safe and efficient train operation.

[0155] like Figure 6 As shown, a system for implementing the plateau construction tunnel track image recognition method, the system includes:

[0156] Ultrasonic imaging sensors, shortwave infrared sensors, and hyperspectral imaging equipment installed at the end of the train transporting supplies collect ultrasonic depth data, optical image data, and spectral signature data from within the tunnel while the train is in motion. The core of the system is a variety of sensors installed at the end of the train transporting supplies: ultrasonic imaging sensors, shortwave infrared sensors, and hyperspectral imaging equipment. These sensors collect ultrasonic depth data, optical image data, and spectral signature data from within the tunnel while the train is in motion. This combination of sensors enables the acquisition of diverse information about the tunnel environment from different physical levels, such as geometry and object material characteristics, providing rich input for subsequent data fusion and feature extraction.

[0157] The time synchronization module is used to time-align and pre-process the collected ultrasonic depth data, optical image data, and spectral signature data to generate synchronized cleaned data. To ensure temporal and spatial consistency of data collected by different sensors, the time synchronization module performs time alignment and pre-processing on the collected ultrasonic depth data, optical image data, and spectral signature data to generate synchronized cleaned data. This step provides preliminary data pre-processing through timestamp calibration and noise removal, thereby improving the accuracy and stability of data fusion in subsequent steps.

[0158] The multimodal data fusion module, which includes multiscale convolution and hyperspectral fusion units, is used to extract geometric feature data and spectral enhancement feature data from the synchronized cleaned data and perform feature fusion to generate fused feature data. The multimodal data fusion module integrates multiscale convolution and hyperspectral fusion units, which can extract geometric feature data and spectral enhancement feature data from the synchronized cleaned data. The multiscale convolution network uses convolution kernels with different receptive fields to extract geometric feature data layer by layer to capture the morphological features and texture details inside the tunnel. The spectral enhancement unit further enriches the feature space with spectral data, allowing the system to maintain robust recognition in the face of optical interference in high temperature and high humidity environments. The fused feature data is the final output feature set, which contains the combined information of various sensor data.

[0159] The adaptive modeling module is configured to receive fused feature data and optimize it using a spectral-geometric co-optimization algorithm. Adaptive training updates the model parameters of the adaptive modeling module to generate optimized feature data. The adaptive modeling module receives fused feature data and processes it using a spectral-geometric co-optimization algorithm. This algorithm optimizes the overall data by balancing the weights of geometric and spectral features, ensuring model stability and recognition accuracy in complex environments. Furthermore, the module incorporates a built-in adaptive training mechanism that dynamically updates model parameters through real-time feedback, enabling the model to maintain its self-adjustment capabilities in a changing environment and generate optimized feature data.

[0160] The real-time prediction and multi-objective decision-making system, including a variational autoencoder and a reinforcement learning module, is used to generate a latent space distribution based on the optimized feature data, perform path simulation and path optimization to generate path prediction data and path optimization data. The optimized feature data is input into the real-time prediction and multi-objective decision-making system, which includes a variational autoencoder (VAE) and a reinforcement learning module. The VAE is used to map the optimized feature data to the latent space, thereby modeling and predicting the distribution of complex track environments. Combined with the reinforcement learning module, the system can perform path simulation and path optimization, outputting path prediction data and path optimization data. This combination enables the system to make effective decisions and path adjustments even in incomplete or noisy environments.

[0161] The active learning module collects feedback from route optimization data and actively learns and updates the adaptive modeling module to improve the system's recognition accuracy and adaptability in high-temperature and high-humidity environments. The active learning module collects feedback from route optimization data and further updates the model parameters of the adaptive modeling module. Using an automatic labeling algorithm, the module annotates the feedback data and performs multi-level data cleaning to remove noise and redundant data. This ensures the model maintains efficient learning capabilities in harsh environments such as high temperature, high humidity, and sensor instability, improving the overall system's recognition accuracy and environmental adaptability.

[0162] By leveraging the multi-module collaboration of the aforementioned systems, this invention enables trains to efficiently identify track and surrounding obstacles, respond promptly, and optimize their routes even in complex tunnel environments at altitudes exceeding 5,000 meters. Compared to existing technologies, this invention not only overcomes the effects of high temperature and humidity on sensors and data processing, but also achieves continuous performance improvements through adaptive modeling and active learning, ensuring the safety and stability of train operations.

[0163] A storage medium stores a computer program, which, when executed by a processor, implements the steps of the plateau construction tunnel track image recognition method.

[0164] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware.

[0165] The above are merely embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should be included within the scope of the claims of the present application.

Claims

1. A method for recognizing track images in plateau tunnel construction, characterized in that: The following steps are involved: Ultrasonic imaging sensors, short-wave infrared sensors, and hyperspectral imaging equipment are installed at the end of the train transporting materials. During the train operation, ultrasonic depth data, optical image data, and spectral feature data in the tunnel are collected, and data synchronization and pre-processing are performed to generate synchronized cleaning data. The synchronized cleaned data is input into the multimodal data fusion module. The geometric feature data and spectral enhancement feature data are extracted through multi-scale convolution and hyperspectral fusion. Feature fusion is performed to generate fused feature data. The geometric feature data is extracted layer by layer using a multi-scale convolutional network. In each convolutional layer, the hyperspectral data is embedded in the convolutional network to form spectral enhancement feature data. The fused feature data is input into the adaptive modeling module, and the fused feature data is optimized using the spectral-geometric feature collaborative optimization algorithm. The model parameters of the adaptive modeling module are updated through adaptive training to generate optimized feature data. The spectral-geometric feature collaborative optimization algorithm optimizes the geometric feature data and the spectral feature data using the following calculation formula: W opt =W geo +α·W spec Among them, the optimized weight combination is W opt ; The geometric feature weight is W geo ; The spectral feature weight is W spec ; α is the dynamic adjustment coefficient; it is used to adaptively adjust the influence of spectral characteristics on the optimization process according to environmental changes; The optimized feature data is input into the real-time prediction and multi-objective decision-making system, and the latent space distribution is generated through the variational autoencoder. The reinforcement learning module is combined to perform path simulation and path optimization to generate path prediction data and path optimization data. The active learning module collects execution feedback of path optimization data and performs active learning and model updating on the adaptive modeling module.

2. The plateau construction tunnel track image recognition method according to claim 1 is characterized in that: In the data synchronization and preprocessing steps, the ultrasonic depth data, optical image data, and spectral feature data are time-aligned and noise-removed through the time synchronization module.

3. The plateau construction tunnel track image recognition method according to claim 1, characterized in that: The multimodal data fusion module adopts a multi-scale convolutional network to extract geometric feature data layer by layer through convolution kernels with different receptive fields, and embeds spectral enhancement feature data in each convolution layer to generate fused feature data.

4. The plateau construction tunnel track image recognition method according to claim 1, characterized in that: During the adaptive training process, the adaptive modeling module introduces a meta-learning mechanism to enable the adaptive modeling module to automatically adjust the learning rate and weight parameters according to feedback data.

5. The plateau construction tunnel track image recognition method according to claim 1, characterized in that: The real-time prediction and multi-objective decision-making system generates a latent space distribution through a variational autoencoder, and the latent space distribution is described by the following formula: p(Z|X)=N(μ(X),σ(X)) Among them, the latent space variable is Z; the optimized feature data is X; the mean of the feature data is μ(X); and the standard deviation of the feature data is σ(X).

6. The plateau construction tunnel track image recognition method according to claim 5, characterized in that: The reinforcement learning module includes a policy network and a value network. The policy network performs path simulation based on the generated latent space distribution, and the value network is used to evaluate the safety and efficiency of the path simulation to generate path prediction data and path optimization data.

7. The plateau construction tunnel track image recognition method according to claim 1, characterized in that: When collecting execution feedback of path optimization data, the active learning module generates labeled feedback data through an automatic labeling algorithm and performs multi-level data cleaning to remove outliers and redundant information.

8. A system for implementing the plateau construction tunnel track image recognition method according to any one of claims 1 to 7, characterized in that: The system includes: Ultrasonic imaging sensors, short-wave infrared sensors, and hyperspectral imaging equipment installed at the end of the train transporting materials are used to collect ultrasonic depth data, optical image data, and spectral feature data in the tunnel while the train is running; A time synchronization module is used to time align and preprocess the collected ultrasonic depth data, optical image data and spectral feature data to generate synchronized cleaning data; The multimodal data fusion module includes a multi-scale convolution and hyperspectral fusion unit, which is used to extract geometric feature data and spectral enhancement feature data from the synchronized cleaned data and perform feature fusion to generate fused feature data. The geometric feature data is extracted layer by layer using a multi-scale convolutional network. In each convolution layer, the hyperspectral data is embedded in the convolutional network to form spectral enhancement feature data. The adaptive modeling module is configured to receive the fused feature data and optimize it using a spectral-geometric feature collaborative optimization algorithm, and to update the model parameters of the adaptive modeling module through adaptive training to generate optimized feature data. The spectral-geometric feature collaborative optimization algorithm optimizes the geometric feature data and the spectral feature data using the following calculation formula: W opt =W geo +α·W spec Among them, the optimized weight combination is W opt ; The geometric feature weight is W geo ; The spectral feature weight is W spec ; α is the dynamic adjustment coefficient; it is used to adaptively adjust the influence of spectral characteristics on the optimization process according to environmental changes; A real-time prediction and multi-objective decision-making system, including a variational autoencoder and a reinforcement learning module, is used to generate a latent space distribution based on optimized feature data, perform path simulation and path optimization to generate path prediction data and path optimization data; The active learning module is used to collect execution feedback of path optimization data and perform active learning and model updates on the adaptive modeling module to improve the recognition accuracy and adaptability of the system in high temperature and high humidity environments.

9. A storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the plateau construction tunnel track image recognition method described in any one of claims 1 to 7 are implemented.

Images