Steam leakage data analysis method and system based on collaborative manifold learning

Abstract

The embodiment of the invention provides a steam leakage data analysis method and system based on collaborative manifold learning, and is applied to the technical field of industrial monitoring systems. According to the method, after a to-be-recognized image is acquired, a candidate area can be recognized from the to-be-recognized image by using a target detection model. And based on the candidate region information, dynamic evolution characteristics of the candidate region in the continuous to-be-identified image are extracted. And then embedding the dynamic evolution characteristics into a low-dimensional space through collaborative manifold learning to obtain low-dimensional embedded characteristics, thereby generating analysis result information according to the low-dimensional embedded characteristics and preset steam leakage physical behavior characteristics. According to the method, the data dimension can be compressed on the basis of keeping the physical semantic consistency, the characteristics related to the steam leakage evolution process are effectively extracted, and the accuracy of the steam leakage data analysis result is improved.

Description

Technical Field

[0001] The present application relates to the technical field of industrial monitoring systems, and in particular to a steam leakage data analysis method and system based on collaborative manifold learning. Background Art

[0002] Steam leakage occurs during industrial production due to damage or seal failure of pipes, valves, equipment, and other components, resulting in the escape of steam from pipelines. Because steam has a high temperature and may be accompanied by the release of toxic substances, steam leaks pose risks such as personal injury, equipment corrosion, production interruptions, and environmental pollution. Therefore, to mitigate the hazards of steam leaks, real-time monitoring of steam leaks is necessary to promptly identify leak points and implement appropriate countermeasures.

[0003] In industrial scenarios, real-time monitoring of steam leaks can be achieved through image acquisition. This involves capturing images of the pipeline system using infrared imaging or video capture equipment, generating infrared images or video data. This data is then analyzed to determine whether a steam leak is present. Because the infrared images and video data generated during a steam leak contain complex spatial structures, temporal dynamics, and temperature gradient information, they exhibit high dimensionality, strong spatiotemporal coupling, nonlinearity, and multi-scale characteristics. Therefore, effectively extracting key features to support subsequent analysis, such as principal component analysis, is fundamental to achieving intelligent steam leak perception and behavioral modeling.

[0004] However, data analysis methods such as principal component analysis (PCA) suffer from poor adaptability when processing high-dimensional, nonlinear data. This makes it difficult to accurately express the evolutionary patterns and physical characteristics of steam leaks, resulting in insufficient feature expression capabilities. Furthermore, the lack of physical consistency modeling capabilities for industrial leak data prevents the full integration of important semantic features such as temperature field diffusion and dynamic trajectories. This reduces the accuracy of steam leak data analysis results and limits the practical application of steam leak analysis in industrial scenarios. Summary of the Invention

[0005] In view of this, an embodiment of the present application provides a steam leakage data analysis method and system based on collaborative manifold learning to solve the problem of low accuracy of steam leakage data analysis results.

[0006] According to a first aspect of the present application, a steam leakage data analysis method based on collaborative manifold learning is provided, the method comprising: Acquire an image to be identified, where the image to be identified is an infrared image acquired by an infrared video surveillance device performing image acquisition on a live operation scene; Identifying candidate regions from the image to be identified using a target detection model to obtain candidate region information, wherein the target detection model is a deep learning model trained using a training dataset; the candidate region information includes a region prediction box, a region category, and a confidence score; Extracting dynamic evolution features of the candidate regions in the continuous images to be identified based on the candidate region information, wherein the dynamic evolution features include temperature distribution features, spatial diffusion morphology features, and time series change features; Embedding the dynamic evolution feature into a low-dimensional space through collaborative manifold learning to obtain a low-dimensional embedding feature, wherein the low-dimensional embedding feature is generated by extracting a nonlinear structural relationship within the data from the dynamic evolution feature; Analysis result information is generated according to the low-dimensional embedding features and preset steam leakage physical behavior features.

[0007] According to a second aspect of the present application, a steam leakage data analysis system based on collaborative manifold learning is provided, the system comprising: an infrared video surveillance device, a server, a communication device, and a terminal device; wherein the infrared video surveillance device and the terminal device establish a communication connection with the server via the communication device; the infrared video surveillance device is configured to perform image acquisition of an on-site operation scene; the server comprises: An image acquisition module is used to acquire an image to be identified, wherein the image to be identified is an infrared image obtained by performing image acquisition on a live operation scene by an infrared video surveillance device; A candidate region recognition module is configured to use a target detection model to identify candidate regions from the image to be recognized, to obtain candidate region information, wherein the target detection model is a deep learning model trained using a training dataset; the candidate region information includes a region prediction box, a region category, and a confidence score; A feature extraction module is used to extract the dynamic evolution features of the candidate region in the continuous image to be identified based on the candidate region information, wherein the dynamic evolution features include temperature distribution features, spatial diffusion morphology features and time series change features; A dimensionality reduction module, configured to embed the dynamic evolution features into a low-dimensional space through collaborative manifold learning to obtain low-dimensional embedded features, wherein the low-dimensional embedded features are generated by extracting nonlinear structural relationships within the data from the dynamic evolution features; An analysis module is used to generate analysis result information based on the low-dimensional embedding features and preset steam leakage physical behavior features.

[0008] According to a third aspect of the present application, a computer device is provided, comprising a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, wherein when the processor executes the program, the steam leakage data analysis method based on collaborative manifold learning is implemented.

[0009] According to a fourth aspect of the present application, a storage medium is provided, on which a computer program is stored. When the program is executed by a processor, the program implements the above-mentioned steam leakage data analysis method based on collaborative manifold learning.

[0010] By means of the above technical solution, the embodiment of the present application provides a steam leakage data analysis method and system based on collaborative manifold learning. After acquiring the image to be identified, the method can use the target detection model to identify the candidate area from the image to be identified. Then, based on the candidate area information, the dynamic evolution characteristics of the candidate area in the continuous image to be identified are extracted. Then, the dynamic evolution characteristics are embedded into the low-dimensional space through collaborative manifold learning to obtain low-dimensional embedding features, so as to generate analysis result information based on the low-dimensional embedding features and the preset steam leakage physical behavior characteristics. The method can integrate physical features such as spatial structure, temperature information and time dynamics to construct a high-dimensional feature expression with physical consistency, and realize nonlinear dimensionality reduction through manifold embedding. The method can compress the data dimension on the basis of maintaining physical semantic consistency, effectively extract features related to the steam leakage evolution process, and improve the accuracy of the steam leakage data analysis results.

[0011] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings: Figure 1 A schematic flow chart of a steam leakage data analysis method based on collaborative manifold learning provided in an embodiment of the present application; Figure 2 A schematic diagram of the system deployment structure provided in an embodiment of the present application; Figure 3 A schematic diagram of the candidate region extraction process provided in an embodiment of the present application; Figure 4 A schematic diagram of the steam leak detection effect provided in an embodiment of the present application; Figure 5 A schematic diagram of the collaborative manifold learning dimensionality reduction process provided in an embodiment of the present application; Figure 6 A schematic diagram of the collaborative manifold learning physical feature dimensionality reduction process provided in an embodiment of the present application; Figure 7 A schematic diagram of the process of generating analysis result information provided in an embodiment of the present application; Figure 8 Schematic diagram of the structure of the steam leakage data analysis system based on collaborative manifold learning provided in an embodiment of the present application. DETAILED DESCRIPTION

[0013] The present application will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that, unless there is a conflict, the embodiments and features in the embodiments of the present application can be combined with each other.

[0014] In the embodiments of the present application, steam leakage is a production abnormality in an industrial production process. Steam leakage is caused by damage or seal failure of pipes, valves, equipment and other components, resulting in steam escaping from the pipes.

[0015] Since steam has a high temperature and may be accompanied by the release of toxic substances, steam leakage may cause personal injury, equipment corrosion, production interruption, and environmental pollution. Therefore, in order to reduce the harm of steam leakage, it is necessary to monitor steam leakage in real time, detect the steam leakage point in time, and take appropriate countermeasures.

[0016] To achieve real-time monitoring of steam leaks, in some embodiments, images of steam-prone areas at the production site, including pipes, valves, and equipment, can be captured to obtain infrared images or video data. The infrared images or video data can then be analyzed to determine whether a steam leak exists.

[0017] During the analysis of infrared images or video data, the large amount of infrared images and video data generated by steam leaks contains complex spatial structures, temporal dynamics, and temperature gradient information, and is characterized by high dimensionality, strong spatiotemporal coupling, nonlinearity, and multi-scale characteristics. Therefore, effectively extracting key features to support subsequent analysis is the foundation for achieving intelligent steam leak perception and behavioral modeling.

[0018] As industrial process automation continues to improve, the demand for steam leak monitoring is growing. The associated image and sensor data exhibit significant characteristics such as high dimensionality, dynamics, and nonlinearity. In particular, when using infrared imaging equipment to capture steam leaks, not only is there significant temperature field diffusion behavior, but it is also accompanied by dynamic temporal and spatial variations and complex environmental interference. Therefore, the processing and analysis of this data, particularly feature extraction and dimensionality reduction modeling, has become a fundamental step in achieving intelligent identification and early warning.

[0019] In some embodiments, a linear method such as principal component analysis (PCA) can be used to convert a set of potentially correlated variables into a set of linearly uncorrelated variables through orthogonal transformation, thereby reducing the dimensionality of the data while preserving the variability of the original data as much as possible. However, the infrared images and video data generated by the steam leakage process have typical spatiotemporal coupling characteristics, containing multidimensional information such as temperature diffusion, dynamic changes, and local disturbances. In actual steam leakage images, the temperature diffusion path, local abnormal hot spots, dynamic morphological changes, etc. often do not conform to the linear assumption. As a result, the above-mentioned linear dimensionality reduction method is difficult to retain key information and cannot effectively capture the nonlinear structure in the data. The dimensionality reduction results lack representativeness, which affects subsequent identification and analysis.

[0020] In some embodiments, nonlinear manifold algorithms such as t-distributed Stochastic Neighbor Embedding (t-SNE), isometric mapping (Isomap), and locally linear embedding (LLE) can be used to represent high-dimensional data in a low-dimensional space while preserving the data's intrinsic structure as much as possible. While manifold learning algorithms can be used for nonlinear data dimensionality reduction, in practice they fail to consider the physical relationships of steam leak data, such as temperature evolution, spatial structure, and time series. This results in a lack of physical consistency in the dimensionality reduction results, making them difficult to use for downstream leak behavior analysis tasks. These nonlinear manifold algorithms also lack the ability to model physical semantics and cannot fully incorporate important semantic features such as temperature field diffusion and dynamic trajectories, making them difficult to use for industrial leak data compression. This limits their practicality in industrial scenarios. For example, while dimensionality reduction algorithms can reveal the manifold structure of data, they fail to consider the unique temporal, spatial diffusion, and thermophysical characteristics of steam leaks, resulting in a lack of engineering interpretability in the dimensionality reduction results.

[0021] Furthermore, steam leak data typically consists of a multi-frame image sequence, and the steam distribution pattern and temperature field within each frame vary significantly. However, the dimensionality reduction method described in the above embodiment only processes single-frame images or static sampling. Performing dimensionality reduction on each frame individually tends to overlook the temporal continuity and diffusion evolution trends between frames, causing the embedded features after dimensionality reduction to lose their temporal structure and making it difficult to characterize complete behavioral patterns. This makes it difficult to fuse multi-source heterogeneous data, and the data dimensionality reduction method described in the above embodiment lacks synergy and the ability to synergistically compress spatiotemporal dynamic features. This makes it impossible to model the continuous evolution trajectory and behavioral patterns during the leak process, limiting its applicability in subsequent diagnostic tasks.

[0022] In order to solve the problem of low accuracy of steam leakage data analysis results, some embodiments of the present application provide a steam leakage data analysis method based on collaborative manifold learning, which can perform dimensionality reduction processing on steam leakage data based on collaborative manifold learning. Combining the physical features and behavioral dynamics of multi-channel images, high-dimensional feature compression and semantic preservation are achieved through nonlinear embedding, providing an interpretable low-dimensional representation for subsequent intelligent recognition and behavioral modeling. By fusing three types of physical features: spatial structure, temperature information, and time dynamics, a physically consistent high-dimensional feature expression space is constructed, and nonlinear dimensionality reduction is achieved through manifold embedding. While maintaining physical semantic consistency, the data dimension is compressed, and potential features closely related to the steam leakage evolution process are effectively extracted, providing low-dimensional expression support for subsequent recognition, judgment, and modeling.

[0023] The steam leakage data analysis method based on collaborative manifold learning can be applied to electronic devices with data processing and communication capabilities. The electronic devices include but are not limited to computers, servers, mobile terminals, smart wearable devices, industrial control machines, etc. For ease of description, in the embodiment of the present application, the server is used as the execution subject of the method. It should be understood that the method can also be applied to other types of execution subjects, which will not be shown one by one in the embodiment of the present application. Figure 1 As shown, the method includes: S101: Obtain an image to be recognized.

[0024] To monitor the on-site operation scene, the server can first acquire an image to be identified. This image is an infrared image captured by infrared video surveillance equipment. The infrared video surveillance equipment can establish a communication connection with the server and, in conjunction with the user's terminal device, establish an intelligent steam leak monitoring system for real-time monitoring of the on-site operation scene.

[0025] In some embodiments, as Figure 2 As shown in the figure, the intelligent steam leak monitoring system adopts a layered architecture design, including a front-end perception layer, a data transmission layer, and a back-end processing layer. The front-end perception layer is used to collect monitoring-related data in the field operation scene. Therefore, the front-end perception layer can include infrared video surveillance equipment deployed in the field operation scene.

[0026] For example, the front-end sensing layer deploys a high-precision infrared thermal imaging camera within a safe 2-3 meter radius around the steam pipe using a universally adjustable bracket. Its spherical node supports full ±45° adjustment, ensuring optimal coverage of the critical pipeline area with a horizontal downward viewing angle of 15-20°. The bracket is constructed of 304 stainless steel and securely connects to building facades, such as walls, columns, and ceilings, using M12 corrosion-resistant expansion bolts, ensuring both stability and adaptability to thermal expansion.

[0027] The data transmission layer facilitates data transmission between the front-end perception layer, back-end processing layer, and client layer. It can include network cables, switches, and other communication equipment. For example, an infrared thermal imaging camera is directly connected to an industrial-grade PoE switch via a Category 6 double-shielded Ethernet cable. After connecting to the core switch via a redundant ring network, the thermal image data is uploaded to the back-end processing layer in real time via a fiber optic link.

[0028] The backend processing layer processes monitoring process data and can be built by installing applications related to the steam leak intelligent monitoring system on a server. For example, the backend processing layer includes a monitoring server that is connected to the enterprise's distributed control system (DCS) for data integration and processing.

[0029] The back-end processing layer can also include user-used terminal devices to display identification and processing results. For example, the terminal device is a mobile terminal such as a mobile phone or tablet computer. The monitoring server can run an AI algorithm. When the AI ​​algorithm identifies the characteristics of a steam leak, it automatically triggers a three-level alarm mechanism, such as an immediate warning from an on-site sound and light alarm, a pop-up 3D positioning alarm window on the management platform, and a disposal work order pushed to the operation and maintenance personnel's mobile terminal via an SMS gateway. At the same time, an electronic plan is generated that includes the leak rate, impact range, and disposal recommendations, providing multi-dimensional data support for emergency decision-making.

[0030] In some embodiments, to obtain images to be identified, the server can receive data analysis instructions input by the user during real-time monitoring and, in response to the data analysis instructions, send a data acquisition request to the infrared video surveillance device. Upon receiving the data acquisition request, the infrared video surveillance device can activate its image acquisition function, capturing infrared images or recording infrared video of the on-site operating scene to generate an image data stream. This image data stream is then sent to the server, allowing the server to retrieve the images to be identified from the image data stream.

[0031] To meet the requirements of subsequent steam leak data analysis, in some embodiments, the server can perform frame segmentation and normalization during the acquisition of the image to be identified, obtaining an image that meets the input requirements. Therefore, the server can first receive the video stream data transmitted by the infrared video surveillance device, then perform frame segmentation on the video stream data to obtain a continuous image sequence comprising multiple consecutive image frames. The image frames in the continuous image sequence are then normalized based on a preset normalization factor to obtain the image to be identified.

[0032] Among them, the preset normalization items include size normalization, format normalization, and pixel normalization for adapting to the input requirements of the target detection model. For example, an infrared camera has high-resolution thermal imaging capabilities and can record the thermal distribution information of the equipment surface and the surrounding environment in real time. Infrared video monitoring equipment deployed around industrial equipment collects video stream data during on-site operation, and performs frame processing to convert the video stream data into a continuous infrared image sequence. The framed images are then preprocessed through size standardization, format conversion, and pixel normalization, so that the preprocessed images can adapt to the input requirements of the deep learning target detection model, providing a high-quality data foundation for subsequent candidate region extraction and physical feature modeling.

[0033] S102: Using the target detection model to identify candidate regions from the image to be identified, to obtain candidate region information.

[0034] After acquiring the image to be recognized, the server may invoke an object detection model. The object detection model is a deep learning model trained using a training dataset. In some embodiments, the server may pre-train the model to obtain the object detection model. During model training, the server may first acquire the training dataset.

[0035] The training data set may include a certain amount of sample images, and each sample image may be provided with a region label. That is, the training data set includes sample images and region labels, and the region labels are used to mark the steam leakage region in the sample images. For example, Figure 3 As shown, the server can use image annotation tools to annotate steam leak areas in infrared images to construct a standardized training dataset. This annotation tool, such as labelimg, is used to annotate steam leak areas in infrared image sequences, generate annotation files, and construct a standardized training dataset. Furthermore, during annotation, bounding boxes must be drawn for each steam leak target in each image, and each bounding box must be assigned a category.

[0036] For the acquired training dataset, the sample images in the training dataset can be sequentially input into the trained deep learning model to obtain the training results output by the deep learning model. The training results are then compared with the region labels to calculate the training loss. The deep learning model is then iteratively trained based on the training loss to obtain the object detection model.

[0037] After obtaining a standardized training dataset through annotation, the server can input the standardized training dataset into the deep learning model for training. The deep learning model can be configured to extract spatial and semantic features from the sample images and perform classification calculations based on the extracted features to obtain training results. This includes determining whether the sample image contains a steam leak target, as well as the predicted region bounding box, region category, and confidence score for the steam leak target in the sample image. The training loss is then calculated by comparing the training results with the region labels. The training loss is used to optimize the deep learning model's ability to detect steam leak targets. Therefore, the training loss is the sum of the class classification loss, bounding box regression loss, and distributional regression loss.

[0038] For example, the deep learning model being trained can be a model such as YOLO v8. YOLO v8 uses an anchor-free design, allowing the detection head to directly regress the object bounding box and category, simplifying the training process. When the server inputs the standardized training dataset into the YOLO v8 deep learning model for training, the deep learning model can extract the spatial and semantic features of the infrared image to optimize the model's ability to detect steam leaks.

[0039] To improve the model's generalization capabilities, we employed YOLO v8's built-in Automatic Mixed Precision (AMP), optical flow transformation, and adaptive sample selection to enrich the variability and diversity of training samples. Data augmentation effectively alleviates the detection challenges associated with the complexity of steam leak scenarios. During training, model parameters were optimized by minimizing a comprehensive loss function. The overall loss function can be expressed as: L total = L cls + L box + L dfl ; in, L cls is the category classification loss; L box It is the bounding box regression loss, which calculates the overlap, center distance and aspect ratio difference between the predicted box and the real box; L dfl It is a distributed regression loss used to improve the accuracy of bounding box regression.

[0040] After calculating the training loss, the server can compare the training loss based on a pre-set loss threshold. When the training loss is greater than the loss threshold, it indicates that the current deep learning model has not yet been trained to convergence. Therefore, the model parameters of the deep learning model can be adjusted based on the training loss. Iterative training continues based on the deep learning model with adjusted parameters until the training loss is less than or equal to the loss threshold. The model parameters corresponding to the deep learning model at this time are output to obtain the target detection model. It can be seen that through continuous iterative training, the deep learning model's ability to detect steam leak targets can be gradually optimized, allowing the target detection model to identify candidate areas from the image to be identified and accurately obtain information about the candidate areas.

[0041] After obtaining the target detection model through iterative training, the server can input the trained target detection model into the real-time image to be identified for target detection. The target detection model then outputs candidate region information. This candidate region information includes the predicted region bounding box, region category, and confidence score. In other words, the target detection model can output the predicted region bounding box, category, and confidence score for the candidate steam leak region.

[0042] To obtain candidate region information, in some embodiments, when the server uses an object detection model to identify candidate regions from the image to be identified, the server may first input the image to be identified into the object detection model, extract the spatial and semantic features of the image to be identified, and then calculate the recognition probability of the steam leak region in the image to be identified based on the spatial and semantic features. The server then determines the region category based on the region label with the highest recognition probability and calculates a confidence score based on the recognition probability, thereby delineating a region prediction box in the image to be identified based on the recognition probability.

[0043] For example, after inputting real-time infrared images into a standardized and trained YOLO v8 object detection model, the object detection model can output the predicted boxes, categories, and confidence scores for candidate steam leak regions. The predicted boxes are then filtered using the non-maximum suppression (NMS) algorithm, removing boxes with high overlap and retaining only the optimal candidate regions. The final output is the coordinates, category label, and corresponding confidence score p for each candidate region. The candidate region information can be expressed as: ; in, Indicates candidate region information; ( x , y , w , h ) are the coordinates and dimensions of the bounding box, c For categories, pis the confidence score, and I is the input image. Through the above target detection process, the candidate region image blocks for physical feature modeling can be obtained, and the detection effect of the candidate region is shown as follows: Figure 4 shown.

[0044] It should be noted that the use of an object detection model to identify candidate regions from the image to be identified and obtain candidate region information, as described in the above embodiments, can serve as the first stage of a steam leak data analysis method based on collaborative manifold learning. In this first stage, the candidate region information obtained by the object detection model provides high-confidence candidate region locations, providing a spatially defined range for physical feature modeling. This provides candidate region coordinates and imagery for feature extraction and dimensionality reduction modeling in the subsequent (second) stage.

[0045] S103 : Based on the candidate region information, extract dynamic evolution features of the candidate region in the continuous to-be-identified images.

[0046] After using the object detection model to identify candidate regions from the image to be identified and obtaining candidate region information, the server can execute the second phase of the collaborative manifold learning-based steam leak data analysis. In this second phase, the server can perform feature extraction and dimensionality reduction modeling based on the candidate region information. Specifically, based on the candidate region information, the server extracts the dynamic evolution characteristics of the candidate region in the continuous image to be identified. These dynamic evolution characteristics include temperature distribution characteristics, spatial diffusion morphology characteristics, and time series variation characteristics.

[0047] like Figure 5 As shown, in order to extract dynamic evolution features, in some embodiments, when extracting the dynamic evolution features of the candidate region in the continuous images to be identified based on the candidate region information, the server may first obtain associated video data based on the image to be identified. The associated video data includes multiple consecutive image frames, and the multiple consecutive image frames include the image to be identified.

[0048] To obtain associated video data, the server can extract a certain number or duration of image frames from a continuous image sequence based on the position of the current image to be identified in the video data stream. The extracted image frames can then form a video of a certain length together with the current image to be identified, i.e., the associated video data. For example, for a frame of image to be identified in a video data stream, the server needs to obtain associated video data of at least 10 seconds in length. Therefore, the server needs to extract multiple consecutive image frames within 10 seconds after the current image to be identified from the video data stream, starting from the current image to be identified. In other words, if the infrared video surveillance device is a high-speed thermal imaging infrared camera with a frame rate of 60, the server needs to extract 600 consecutive infrared image frames from the video data stream to form the associated video data.

[0049] After acquiring the associated video data, the server then labels the steam leak region in multiple consecutive image frames based on the candidate regions. It then extracts physical features from the associated video data based on the steam leak region, including temperature distribution data, spatial diffusion patterns, and time series variation characteristics. Dynamic evolution features are then constructed by statistically analyzing the physical features across multiple consecutive image frames.

[0050] Specifically, based on the candidate steam leak areas detected in the first phase, the server extracts temperature distribution data, spatial diffusion patterns, and time series variation characteristics from continuous infrared video. These characteristics reflect the dynamic evolution of steam under actual operating conditions. The extracted physical information is then integrated into a high-dimensional physical feature set, taking into account its spatial structure, thermal distribution patterns, and temporal evolution trends. This construction of a complete, multi-dimensional candidate region feature space provides the foundation for subsequent dimensionality reduction.

[0051] For example, based on the candidate steam leak regions identified in the first stage of target detection, the server can perform feature extraction on the candidate region data to extract its physical characteristics within the continuous infrared video. These physical characteristics include, but are not limited to, temperature distribution data T, spatial diffusion pattern S, and time series variation characteristics f. By performing statistical analysis on the continuous frame data of the candidate region, a multidimensional physical feature set F = {T, S, f} is constructed for the candidate region. This is then input into the collaborative manifold learning model for dimensionality reduction modeling.

[0052] S104. Embed the dynamic evolution features into a low-dimensional space through collaborative manifold learning to obtain low-dimensional embedded features.

[0053] After extracting dynamic evolution features from multiple consecutive frames associated with the image to be identified through feature extraction, the server can then perform the second phase of the method, dimensionality reduction modeling, which embeds the dynamic evolution features into a low-dimensional space through collaborative manifold learning to obtain low-dimensional embedded features. The low-dimensional embedded features are generated by extracting nonlinear structural relationships within the data from the dynamic evolution features.

[0054] In some embodiments, when embedding dynamically evolving features into a low-dimensional space through collaborative manifold learning to obtain low-dimensional embedded features, the server may first call a collaborative manifold learning module. The collaborative manifold learning module is a learning module constructed by combining manifold learning and a collaborative learning algorithm framework. The collaborative manifold learning module can be configured to mine the inherent manifold structure of the dynamically evolving features and the collaborative relationship between different features.

[0055] Based on the called collaborative manifold learning module, the server can input the dynamic evolution features into the collaborative manifold learning module to obtain a fused feature representation. Based on the fused feature representation, the neighborhood relationship between different modalities is constructed, and the features to be embedded are obtained by adaptive neighbor selection according to the neighborhood relationship. The nonlinear manifold dimensionality reduction algorithm is then used to calculate the low-dimensional embedding representation of the features to be embedded to obtain low-dimensional embedding features. The fused feature representation is a splicing result obtained by performing multi-modal feature weighted splicing on the dynamic evolution features. The nonlinear manifold dimensionality reduction algorithm includes at least one of local linear embedding, t-distributed random neighbor embedding, and isometric mapping embedding.

[0056] The server can input high-dimensional physical features into the collaborative manifold learning module. The collaborative manifold learning module can combine spatial, temporal and temperature information to construct a high-dimensional physical feature space and adopt nonlinear manifold dimensionality reduction algorithms such as local linear embedding (LLE), t-distributed stochastic neighbor embedding (t-SNE) or isomap to obtain a low-dimensional embedding representation to express the potential physical pattern of the candidate target.

[0057] For example, to more intuitively demonstrate the collaborative manifold learning method, when modeling the physical features of the candidate region, the server can use Figure 6 The core processing shown performs dimensionality reduction modeling. This process can include key steps such as high-dimensional feature construction, nonlinear dimensionality reduction, and low-dimensional embedding generation. When performing dimensionality reduction modeling, the server can first input multimodal features. Multimodal feature input can include three different types of feature inputs: spatial features, which represent the spatial characteristics of the candidate region in the image to be identified; temporal features, which represent the characteristics of the candidate region data in the image to be identified over time; and temperature image features, which represent the temperature data of the candidate region in the image to be identified, namely, the temperature distribution in the thermal image.

[0058] Based on the multimodal feature input, the server can perform multimodal weighted concatenation at the feature fusion layer. This involves setting multimodal weights and performing a weighted concatenation of the three features, combining them to form a comprehensive feature representation. During this weighted concatenation and fusion process, features from different modalities can be weighted to highlight the importance of certain features.

[0059] After forming a fused feature representation through weighted concatenation, the server can perform adaptive neighbor selection for cross-modal neighborhood construction. To find similar or related data points across modalities for better understanding and analysis, the server can construct neighborhood relationships between modalities and perform adaptive neighbor selection to obtain features to be embedded.

[0060] After obtaining the features to be embedded, the server can also use a manifold optimization algorithm to maintain the local structure of the data. Manifold learning is a nonlinear dimensionality reduction technique that can preserve the data's inherent geometric structure in a low-dimensional space while minimizing the dimensionality of the data.

[0061] Ultimately, the server can express the physical pattern in a low-dimensional embedding space by mapping the processed data into a low-dimensional embedding space. In this low-dimensional space, the data representation can reflect its pattern or structure in the physical world, thereby extracting and expressing the physical pattern in the data.

[0062] In some embodiments, when the server embeds the dynamic evolution features into a low-dimensional space through collaborative manifold learning to obtain low-dimensional embedded features, it can first extract the characteristic trajectories of the candidate regions in consecutive image frames based on the low-dimensional embedded features. It then applies a neighborhood consistency analysis algorithm to calculate the consistency metric parameters of the candidate regions in the characteristic trajectories. Furthermore, it uses a trajectory clustering algorithm to identify abnormal trajectory behavior based on the spatiotemporal distribution and change trends of the trajectory points in the characteristic trajectories to obtain an anomaly identification result. Based on the consistency metric parameters and the anomaly identification results, the diffusion pattern and thermal dynamic characteristics of the characteristic trajectories are evaluated.

[0063] For example, the server can use collaborative manifold learning technology to process the above high-dimensional physical features, extract the nonlinear structural relationships within the data, and embed them into a low-dimensional space. This embedding result significantly reduces the data dimension while preserving the original physical behavior characteristics as much as possible, thereby improving the efficiency and accuracy of subsequent analysis. The dimensionality reduction goal can be formally expressed as:

[0064] in, Z Represents the feature set to be reduced; N Indicates the total number of data points; F i and F j Respectively represent i and j The characteristics of each data point; N ( i ) represents the sample i Neighborhood, ω ij is the reconstruction weight.

[0065] The server can then model and analyze the feature trajectories of the candidate regions under continuous image frames in the embedded low-dimensional manifold space. In the obtained low-dimensional manifold space Z, the neighborhood consistency analysis and dynamic trajectory clustering algorithm are applied to model and verify the physical behavior evolution trajectory of the candidate regions. By calculating the trajectory consistency measure of the candidate target in the embedded space D ( z i , z j ), to assess whether it conforms to the diffusion pattern and thermal dynamic characteristics of a typical steam leak. The neighborhood consistency metric can be formalized as:

[0066] in, D ( z i , z j ) represents the trajectory consistency measure; z i and z j Represent the i-th and j-th state vectors respectively; ∆ t Indicates a time interval.

[0067] At the same time, trajectory clustering algorithms such as DBSCAN and OPTICS are introduced to identify abnormal trajectory behaviors through the spatiotemporal distribution and change trends of trajectory points, and to determine whether the candidate targets meet the diffusion patterns and thermal dynamic characteristics of typical steam leaks.

[0068] S105. Generate analysis result information based on the low-dimensional embedded features and the preset steam leakage physical behavior features.

[0069] After obtaining the low-dimensional embedded features, the server can determine whether the image to be identified meets the preset steam leak physical behavior features based on the low-dimensional embedded features and the preset steam leak physical behavior features, and generate analysis result information based on the determination result. Specifically, if the low-dimensional embedded features corresponding to the image to be identified meet the preset steam leak physical behavior features, an analysis result can be obtained that indicates that the image to be identified contains a steam leak target, thereby indicating the presence of a steam leak in the current on-site operation scene. If the low-dimensional embedded features corresponding to the image to be identified do not meet the preset steam leak physical behavior features, an analysis result can be obtained that indicates that the image to be identified does not contain a steam leak target. In this case, the server can continue to monitor the on-site operation scene according to the analysis method provided in the above embodiment.

[0070] For example, in the reduced low-dimensional space, the server can further analyze the evolution trajectory and diffusion pattern of candidate regions to determine whether they conform to the physical behavior characteristics of a typical steam leak. For regions that meet these characteristics, their embedded representations are output as valid low-dimensional feature results. Abnormal samples are removed or marked, providing reliable input for subsequent identification, early warning, and other applications.

[0071] To this end, in some embodiments, when generating analysis result information based on the low-dimensional embedded features and the preset physical behavior characteristics of the steam leak, the server may first extract behavioral features from the low-dimensional embedded features and obtain a typical steam leak diffusion pattern. The behavioral features include spatial structural features and temporal evolution trajectories. The typical steam leak diffusion pattern includes the preset physical behavior characteristics of the steam leak.

[0072] By comparing the behavior features with the preset physical behavior features of steam leaks, if the behavior features are consistent with the preset physical behavior features of steam leaks, the low-dimensional embedded features corresponding to the behavior features are retained. If the behavior features are inconsistent with the preset physical behavior features of steam leaks, the low-dimensional embedded features corresponding to the behavior features are removed or marked.

[0073] For example, the server can construct behavioral features by combining the spatial structure and temporal evolution of low-dimensional embedded features. If the behavioral features are highly consistent with the typical diffusion pattern of a steam leak, the low-dimensional embedded features are retained and a low-dimensional feature set is output. This output low-dimensional feature set serves as the input data for subsequent recognition, clustering, or prediction models, and is used for subsequent analysis and modeling. If the behavioral features do not meet the consistency criteria for the typical diffusion pattern of a steam leak, they are discarded to improve the effectiveness and expressiveness of the final dimensionality reduction results.

[0074] As can be seen, by combining multidimensional physical feature modeling with manifold embedding methods, the server can effectively reduce data dimensionality, retain key information, and provide strong feature support for subsequent steam leak identification, trend analysis, and early warning. Furthermore, to adapt to the computing resource and real-time requirements of different application scenarios, the dimensionality reduction algorithm and candidate region extraction method can be flexibly adjusted according to actual application needs.

[0075] By applying the technical solutions of the above-mentioned embodiments, the steam leak data analysis method based on collaborative manifold learning provided in the above-mentioned embodiments can integrate temperature, spatial, and temporal features to construct a high-dimensional physical feature set and perform dimensionality reduction processing. A nonlinear embedding algorithm is also used to achieve manifold dimensionality reduction of candidate region features, improving the physical consistency and interpretability of the dimensionality reduction results. Trajectory clustering and neighborhood consistency analysis are also used to screen low-dimensional features that conform to the evolutionary laws of steam leaks, thereby improving the effectiveness and expressiveness of the dimensionality reduction data. Therefore, the described method not only effectively compresses data dimensions and preserves key evolutionary patterns during steam leaks, but also exhibits good physical consistency and semantic interpretability, providing high-quality low-dimensional embedded features for subsequent anomaly detection, behavior modeling, and intelligent analysis.

[0076] In some embodiments, as a refinement and extension of the specific implementation of the above embodiment, in order to fully illustrate the specific implementation process of this embodiment, some embodiments of this application also provide a steam leakage data analysis method based on collaborative manifold learning. On the basis of the above embodiment, the method can also generate analysis result information based on the retained low-dimensional embedding features, such as Figure 7 As shown, the method includes: S201, generating an effective low-dimensional feature set based on the retained low-dimensional embedded features; S202, calculating the classification probability of the image to be identified for the steam leak classification label based on the effective low-dimensional feature set; S203: If the classification probability is greater than or equal to the probability threshold, generate first analysis result information; S204: If the classification probability is less than the probability threshold, generate second analysis result information.

[0077] The server can filter the low-dimensional embedded features based on the comparison results of the behavioral features with the typical diffusion pattern of steam leaks, and then collect and organize the retained low-dimensional embedded features to generate an effective low-dimensional feature set. The effective low-dimensional feature set is then input into the analysis model, so that the analysis model can calculate the classification probability of the image to be identified for the steam leak classification label based on the effective low-dimensional feature set. The analysis model can be a neural network model trained using sample images containing low-dimensional embedded features. The training process of the analysis model can be similar to the training method of the target recognition model described in the above embodiment, with corresponding differences in the classification labels and loss function, which will not be further described here.

[0078] After inputting the effective low-dimensional feature set into the trained analysis model, the analysis model can output the classification probability of the effective low-dimensional feature set for the analysis result label. The server can then compare the classification probability with a preset probability threshold to determine different analysis results. Specifically, if the classification probability is greater than or equal to the probability threshold, a first analysis result is generated. This first analysis result indicates that the image to be identified contains a steam leak target. If the classification probability is less than the probability threshold, a second analysis result is generated, indicating that the image to be identified does not contain a steam leak target.

[0079] By applying the technical solutions of the aforementioned embodiments, the steam leak data analysis method based on collaborative manifold learning provided in these embodiments can model the data based on the physical properties of steam leak image data by integrating multidimensional information, thereby improving the problems of insufficient feature modeling and weak dimensionality reduction expression. Furthermore, the use of collaborative manifold learning for nonlinear dimensionality reduction can better preserve the spatial structure and dynamic characteristics of the leak region, improving the structural preservation and semantic expression of the data dimensionality reduction results. Furthermore, by introducing trajectory consistency analysis and clustering mechanisms, low-dimensional features are effectively screened, enhancing the stability and applicability of the dimensionality reduction results, and providing high-quality feature input for subsequent recognition, clustering, and prediction tasks.

[0080] As can be seen, the proposed method is able to focus on steam leak infrared image data, which is characterized by high dimensionality, nonlinearity, strong spatiotemporal coupling, and significant dynamic features. A data dimensionality reduction technique based on collaborative manifold learning is proposed. Through candidate region extraction, multidimensional physical feature construction, and nonlinear manifold embedding, this method achieves a structured representation of steam leak data in a low-dimensional space. Unlike shallow processing methods based solely on edge, brightness, or temperature thresholds, the proposed method integrates spatial diffusion morphology, temperature evolution trends, and temporal dynamic information. While maintaining the consistency of the original physical features, it significantly reduces feature dimensionality, facilitating subsequent tasks such as clustering, recognition, and trend analysis.

[0081] Therefore, the method can be applied to the efficient compression and feature extraction of infrared image data under complex working conditions, and can be widely used in steam system anomaly analysis, industrial intelligent monitoring, behavior modeling and data-driven fault warning, etc., with good engineering practicality and industrial promotion prospects.

[0082] In some embodiments, as a specific implementation of the steam leakage data analysis method based on collaborative manifold learning described in the above embodiments, some embodiments of the present application also provide a steam leakage data analysis system based on collaborative manifold learning, such as Figure 8 As shown, the system includes: infrared video monitoring equipment, servers, communication equipment and terminal equipment.

[0083] Wherein, the infrared video monitoring device and the terminal device establish a communication connection with the server through the communication device; the infrared video monitoring device is configured to perform image acquisition on the on-site operation scene.

[0084] For example, to support the implementation of the collaborative manifold learning-based steam leak data analysis method described in the above embodiment, the system may include an infrared camera monitoring device, a server, and a monitored object. The infrared camera monitoring device, serving as the image acquisition end, can be installed around key pipelines or equipment at industrial sites to continuously capture infrared image sequences and obtain thermal distribution information on the equipment surface and surrounding areas.

[0085] The server, deployed with a deep learning object detection model and collaborative manifold learning algorithm, receives infrared image data and performs candidate region detection, physical feature extraction, and dimensionality reduction. The monitored targets are critical thermal equipment or pipelines within industrial sites that pose a potential risk of steam leaks.

[0086] Through the linkage between the image acquisition terminal and the back-end server, standardized acquisition and dimensionality reduction analysis processing of high-dimensional image data can be achieved, providing high-quality, structurally consistent low-dimensional feature representation for subsequent steam leakage behavior modeling and intelligent analysis. The server includes: An image acquisition module is used to acquire an image to be identified, wherein the image to be identified is an infrared image obtained by performing image acquisition on a live operation scene by an infrared video surveillance device; A candidate region recognition module is configured to use a target detection model to identify candidate regions from the image to be recognized, to obtain candidate region information, wherein the target detection model is a deep learning model trained using a training dataset; the candidate region information includes a region prediction box, a region category, and a confidence score; A feature extraction module is used to extract the dynamic evolution features of the candidate region in the continuous image to be identified based on the candidate region information, wherein the dynamic evolution features include temperature distribution features, spatial diffusion morphology features and time series change features; A dimensionality reduction module, configured to embed the dynamic evolution features into a low-dimensional space through collaborative manifold learning to obtain low-dimensional embedded features, wherein the low-dimensional embedded features are generated by extracting nonlinear structural relationships within the data from the dynamic evolution features; An analysis module is used to generate analysis result information based on the low-dimensional embedding features and preset steam leakage physical behavior features.

[0087] By applying the technical solutions of the above-described embodiments, the steam leak data analysis system based on collaborative manifold learning provided in the above-described embodiments can, after the image acquisition module acquires the image to be identified, use the target detection model to identify candidate regions from the image to be identified by the candidate region identification module. The feature extraction module then extracts the dynamic evolution features of the candidate regions in the continuous image to be identified based on the candidate region information. The dimensionality reduction module then embeds the dynamic evolution features into a low-dimensional space through collaborative manifold learning to obtain low-dimensional embedded features. Finally, the analysis module generates analysis results based on the low-dimensional embedded features and preset steam leak physical behavior characteristics. The system can integrate physical features such as spatial structure, temperature information, and temporal dynamics to construct a physically consistent high-dimensional feature representation, and achieve nonlinear dimensionality reduction through manifold embedding. The system can compress data dimensions while maintaining physical semantic consistency, effectively extract features related to the steam leak evolution process, and improve the accuracy of steam leak data analysis results.

[0088] It should be noted that for other corresponding descriptions of the various functional units involved in the steam leakage data analysis system based on collaborative manifold learning provided in an embodiment of the present application, reference can be made to the corresponding descriptions in the steam leakage data analysis method based on collaborative manifold learning provided in the above embodiment, and will not be repeated here.

[0089] The embodiment of the present application also provides a computer device, which can be specifically a personal computer, a server, a network device, etc. The computer device includes a bus, a processor, a memory and a communication interface, and may also include an input and output interface and a display device. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store location information. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, the steps in each method embodiment are implemented.

[0090] Those skilled in the art will understand that the structure of the above-mentioned computer device is only a partial structure related to the solution of the present application and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components, or combine certain components, or have a different component arrangement.

[0091] In one embodiment, a computer-readable storage medium is further provided. The computer-readable storage medium may be non-volatile or volatile, and stores a computer program thereon. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0092] In one embodiment, a computer program product is further provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0093] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0094] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing related hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods.

[0095] Any reference to a memory, database, or other medium used in the embodiments provided herein may include at least one of a non-volatile memory and a volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc.

[0096] Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0097] The database involved in each embodiment provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchain. The processor involved in each embodiment provided herein may be, but is not limited to, a general-purpose processor, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, and the like.

[0098] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0099] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A steam leakage data analysis method based on collaborative manifold learning, characterized in that: The method comprises: Acquire an image to be identified, where the image to be identified is an infrared image acquired by an infrared video surveillance device performing image acquisition on a live operation scene; Identifying candidate regions from the image to be identified using a target detection model to obtain candidate region information, wherein the target detection model is a deep learning model trained using a training dataset; the candidate region information includes a region prediction box, a region category, and a confidence score; Extracting dynamic evolution features of the candidate regions in the continuous images to be identified based on the candidate region information, wherein the dynamic evolution features include temperature distribution features, spatial diffusion morphology features, and time series change features; Embedding the dynamic evolution feature into a low-dimensional space through collaborative manifold learning to obtain a low-dimensional embedding feature, wherein the low-dimensional embedding feature is generated by extracting a nonlinear structural relationship within the data from the dynamic evolution feature; Analysis result information is generated according to the low-dimensional embedding features and preset steam leakage physical behavior features.

2. The method according to claim 1, characterized in that Obtain the image to be recognized, including: Receiving video stream data sent by the infrared video surveillance device; Performing frame processing on the video stream data to obtain a continuous image sequence, wherein the continuous image sequence includes a plurality of continuous image frames; Based on preset normalization items, the image frames in the continuous image sequence are normalized to obtain the image to be identified. The preset normalization items include size normalization, format normalization and pixel normalization for adapting to the input requirements of the target detection model.

3. The method according to claim 1, characterized in that The method further comprises: Acquire a training data set, the training data set including sample images and region labels, the region labels being used to mark steam leakage regions in the sample images; Inputting the sample image into a trained deep learning model to obtain a training result output by the deep learning model; Calculating a training loss by comparing the training result and the region label, where the training loss is the sum of a category classification loss, a bounding box regression loss, and a distributional regression loss; Iterative training is performed on the deep learning model based on the training loss to obtain the object detection model.

4. The method according to claim 1, wherein Identifying candidate regions from the image to be identified using an object detection model to obtain candidate region information, including: Inputting the image to be identified into the object detection model to extract spatial features and semantic features of the image to be identified through the object detection model; Calculating a recognition probability of a steam leakage area in the image to be recognized based on the spatial features and the semantic features; Determining the region category based on the region label with the highest recognition probability and calculating the confidence score according to the recognition probability; According to the recognition probability, a region prediction box is defined in the image to be recognized.

5. The method according to claim 1, wherein Extracting dynamic evolution features of the candidate region in the continuous images to be identified based on the candidate region information includes: Acquire associated video data based on the image to be identified, wherein the associated video data includes a plurality of continuous image frames, and the plurality of continuous image frames include the image to be identified; marking steam leakage regions in the plurality of consecutive image frames in the associated video data according to the candidate regions; Extracting physical features of the associated video data according to the steam leakage area, wherein the physical features include temperature distribution data, spatial diffusion morphology, and time series change characteristics; The dynamic evolution feature is constructed by counting the physical features in the multiple consecutive image frames.

6. The method according to claim 1, characterized in that The dynamic evolution features are embedded into a low-dimensional space through collaborative manifold learning to obtain low-dimensional embedding features, including: Invoking a collaborative manifold learning module, wherein the collaborative manifold learning module is a learning module constructed by combining manifold learning and collaborative learning algorithm frameworks; Inputting the dynamic evolution feature into the collaborative manifold learning module to obtain a fused feature representation, wherein the fused feature representation is a splicing result obtained by performing multimodal feature weighted splicing on the dynamic evolution feature; Constructing a neighborhood relationship between different modalities based on the fused feature representation, and obtaining features to be embedded by adaptive neighbor selection according to the neighborhood relationship; A nonlinear manifold dimensionality reduction algorithm is used to calculate a low-dimensional embedding representation of the feature to be embedded to obtain the low-dimensional embedding feature; the nonlinear manifold dimensionality reduction algorithm includes at least one of local linear embedding, t-distributed random neighbor embedding, and isometric mapping embedding.

7. The method according to claim 1, characterized in that The dynamic evolution features are embedded into a low-dimensional space through collaborative manifold learning to obtain low-dimensional embedding features, including: Extracting feature trajectories of the candidate regions in consecutive image frames based on the low-dimensional embedded features; Applying a neighborhood consistency analysis algorithm to calculate consistency measurement parameters of the candidate region on the feature trajectory; According to the trajectory clustering algorithm, the abnormal trajectory behavior is identified through the temporal and spatial distribution and change trend of the trajectory points in the characteristic trajectory to obtain an abnormality identification result; Based on the consistency metric parameter and the anomaly identification result, a diffusion pattern and thermal dynamic characteristics of the characteristic trajectory are evaluated.

8. The method according to claim 1, characterized in that Generate analysis result information based on the low-dimensional embedding features and the preset steam leakage physical behavior features, including: Extracting behavioral features from the low-dimensional embedded features, the behavioral features including spatial structural features and time evolution trajectories; Acquire a typical diffusion pattern of steam leakage, wherein the typical diffusion pattern of steam leakage includes the preset physical behavior characteristics of steam leakage; If the behavior feature is consistent with the preset steam leakage physical behavior feature, retaining the low-dimensional embedding feature corresponding to the behavior feature; If the behavior feature is inconsistent with the preset steam leakage physical behavior feature, the low-dimensional embedding feature corresponding to the behavior feature is removed or marked.

9. The method according to claim 8, characterized in that Generate analysis result information based on the low-dimensional embedding features and the preset steam leakage physical behavior features, including: Generating an effective low-dimensional feature set based on the retained low-dimensional embedding features; Calculating the classification probability of the image to be identified for the steam leak classification label based on the effective low-dimensional feature set; If the classification probability is greater than or equal to the probability threshold, generating first analysis result information, wherein the first analysis result information is used to indicate that the image to be identified contains a steam leak target; If the classification probability is less than the probability threshold, second analysis result information is generated, where the second analysis result information is used to indicate that the image to be identified does not contain a steam leakage target.

10. A steam leakage data analysis system based on collaborative manifold learning, characterized in that: The system includes: an infrared video surveillance device, a server, a communication device, and a terminal device; wherein the infrared video surveillance device and the terminal device establish a communication connection with the server via the communication device; the infrared video surveillance device is configured to perform image acquisition of an on-site operation scene; the server includes: An image acquisition module is used to acquire an image to be identified, wherein the image to be identified is an infrared image obtained by performing image acquisition on a live operation scene by an infrared video surveillance device; A candidate region recognition module is configured to use a target detection model to identify candidate regions from the image to be recognized, to obtain candidate region information, wherein the target detection model is a deep learning model trained using a training dataset; the candidate region information includes a region prediction box, a region category, and a confidence score; A feature extraction module is used to extract the dynamic evolution features of the candidate region in the continuous image to be identified based on the candidate region information, wherein the dynamic evolution features include temperature distribution features, spatial diffusion morphology features and time series change features; A dimensionality reduction module, configured to embed the dynamic evolution features into a low-dimensional space through collaborative manifold learning to obtain low-dimensional embedded features, wherein the low-dimensional embedded features are generated by extracting nonlinear structural relationships within the data from the dynamic evolution features; An analysis module is used to generate analysis result information based on the low-dimensional embedding features and preset steam leakage physical behavior features.

Images