Fire site and fire point tracing method and system, and storage medium

Abstract

The invention relates to a source tracing method and system for a fire site and a fire point, and a storage medium, and belongs to the technical field of information source tracing based on space correlation and fire scene trace image intelligent analysis. Comprising the following steps: data acquisition and database construction, and standardized processing and tagged management are carried out; extracting fire scene trace features in a three-dimensional space, wherein the fire scene trace features comprise spreading direction trace feature extraction, fire site local feature extraction and fire point local trace feature extraction; constructing a mapping model from a three-dimensional space to a two-dimensional image, and establishing a cross-modal feature correlation model; carrying out machine learning analysis, training a large amount of collected fire scene data and laboratory simulation data by using a deep learning algorithm, comparing a scene image with a database image, calculating the similarity and judging the position of a fire point; and model optimization and dynamic learning: through a dynamic incremental learning mechanism, continuously introducing new data, optimizing model performance, and improving accuracy and generalization ability. According to the invention, the scientificity and accuracy of fire investigation are greatly improved.

Description

Technical Field

[0001] The present invention relates to a method and system for tracing the origin of a fire location and a storage medium, which are based on spatial correlation and intelligent analysis of fire scene trace images and belong to the field of information tracing technology. Background Art

[0002] Fire is a sudden and destructive disaster with 800,000 to 900,000 fire-related incidents every year, posing a serious threat to public safety. At the same time, fire scenes are complex and changeable, and investigation is difficult. The project focuses on key technologies for identifying the starting points of fire-related incidents: first, we collect fire scene data and build a database platform based on the construction of a "data acquisition" standard system, and then focus on the development of efficient multi-source data fusion algorithms to accurately extract key information. Based on big data and deep learning, we analyze the fire range, spread direction, fire location, and fire point trace characteristics to provide support for the identification of unknown fire points, establish "data analysis" and "data application" standards and specifications, and help accurately identify the starting points of fire-related incidents. This will safeguard the professional law enforcement of public security, fire, emergency and housing and construction departments, ensure the scientific and accurate handling of fire-related incidents, and maintain judicial fairness.

[0003] In traditional fire investigations, the determination of the fire point often relies on the experience of investigators, who make inferences based on observing two-dimensional features such as smoke marks and combustion residues. This is highly subjective and inefficient. Existing technologies, such as the fire detection method disclosed in CN114998822A, can identify fire conditions but lack quantitative analysis of three-dimensional spatial features. The fire source location technology in CN119445802A relies on sensor data and is difficult to apply to complex scenes. Currently, three-dimensional scanning technology can accurately restore the spatial structure of a fire scene, but combining three-dimensional data with two-dimensional trace feature images to achieve intelligent tracing of the fire location and point of origin remains a technical challenge. Summary of the Invention

[0004] In response to the shortcomings of the existing technology, the present invention proposes a fire location and ignition point tracing method, system and storage medium based on spatial correlation and intelligent analysis of fire scene trace images. By systematically collecting three-dimensional scanning on-site data and laboratory simulation data of fire-related incidents, a fire point database with multiple dimensions such as fire location, fire cause, fire location, ignition point, time and geographic information is constructed. Deep learning technology is used to develop a large model for fire location and ignition point analysis, breaking through the limitations of traditional methods and achieving efficient and accurate identification of fire locations and even specific ignition points.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0006] A method for tracing the origin of a fire location is provided, which is based on spatial correlation and intelligent analysis of fire scene trace images and includes the following steps:

[0007] S 1. Data collection and database construction: Collect 3D scanning data and image data from the fire scene, combine it with laboratory simulated fires to generate a standard data set, establish a fire point database containing multi-dimensional and multi-modal information such as fire location, fire cause, fire location, fire point, and time, and perform standardized processing and labeling management;

[0008] S2. Extract fire trace features in three-dimensional space, including trace features in the direction of spread, local features of the fire location, and local trace features at the fire point;

[0009] S3. Construct a 3D-to-2D image mapping model, projecting 3D data through multi-view rendering to generate a 2D image set, and use convolutional neural networks to extract texture features, color changes, and edge contour features. Feature fusion and association modeling: deeply fuse 3D spatial features with 2D image features to establish a cross-modal feature association model.

[0010] S4. Machine learning analysis: Using deep learning algorithms, a large amount of collected fire scene data and laboratory simulation data is trained to build a large model for analyzing fire locations and ignition points. By inputting a large amount of labeled three-dimensional spatial feature data and two-dimensional image data, the model learns the characteristic patterns and laws of fire locations and ignition points in different fire scenarios; based on the twin neural network architecture combined with the attention mechanism, intelligent image comparison is performed to compare the scene image with the database image, calculate the similarity and determine the location of the fire point; S5. Model optimization and dynamic learning: Through the dynamic incremental learning mechanism, new data is continuously introduced to optimize model performance and improve accuracy and generalization ability.

[0011] Furthermore, in S1, the data sources include:

[0012] 3D scanning data of fire-related incidents: Use 3D laser scanning equipment to obtain spatial data of the real fire scene, and fully record the smoke, combustion, and deformation information at the scene;

[0013] Laboratory simulation data: Build experimental models of different fire causes and situations in a controlled environment, and use 3D scanning technology to record the fire process and characteristic information at each stage;

[0014] Data preprocessing: MeshLab and CloudCompare 3D processing software were used to remove noise, reconstruct discrete point clouds, and standardize the collected data. At the same time, multi-level and multi-dimensional labels were added to the fire points and important feature locations.

[0015] Furthermore, in S2, the three-dimensional fire scene trace feature extraction specifically includes:

[0016] Extraction of trace features of the spread direction: Identify the carbonization, sooting, and melting trace features on the surfaces of combustible and non-combustible materials to infer the fire spread path;

[0017] Extraction of fire trace features: Analyze the material properties and local deformation damage signs of the fire site to establish a preliminary judgment of the fire location;

[0018] Fire point trace feature extraction and comprehensive determination: Refine the fire point trace features and determine the fire point location based on comprehensive fire scene features.

[0019] Furthermore, in said S2, the fire point trace features are refined to include ash accumulation and electrical fault traces.

[0020] Furthermore, in S3, the mapping modeling of the two-dimensional image adopts a method combining orthogonal projection and perspective projection to convert the three-dimensional spatial features into a multi-view two-dimensional image, retaining the spatial topological relationship, and performing feature extraction and feature enhancement on the mapped two-dimensional image through the convolutional neural network CNN to ensure the integrity and robustness of the two-dimensional image in spatial representation capability.

[0021] Furthermore, in S4, the intelligent image comparison adopts a twin network combined with an attention mechanism to perform high-dimensional feature matching and fine-grained similarity judgment on the target image and the database image; the similarity judgment adopts a comprehensive judgment method of weighted fusion of cosine similarity and Euclidean distance, thereby improving the accuracy and stability of the fire point location matching.

[0022] Furthermore, in S5, the model optimization specifically includes:

[0023] Incremental learning: Introducing newly collected fire case data, the deep neural network is continuously updated with small steps to enhance the model's adaptability to new types of fire scenarios;

[0024] Knowledge distillation: By migrating knowledge from large and complex models to lightweight sub-models, we improve model deployment efficiency while maintaining judgment accuracy and adapt to the operational needs of terminal devices.

[0025] Automatic hyperparameter optimization: Use the Bayesian optimization algorithm to automatically tune key hyperparameters such as learning rate, regularization coefficient, and loss function weight to improve the overall training efficiency and prediction performance of the model.

[0026] A fire location and point tracing system based on spatial correlation and intelligent analysis of fire scene trace images, which adopts the above-mentioned fire location and point tracing method, includes:

[0027] Data acquisition device: used to collect 3D scanning data, image data and laboratory simulation data of the fire scene;

[0028] Data storage unit: used to build a fire point database containing multi-dimensional data such as fire location, fire cause, fire location, fire point, time information and geographical location;

[0029] Trace feature extraction module: used to extract fire-related spatial features from 3D reconstruction data, including trace features of the spread direction of combustible and non-combustible materials, trace features of the fire location, and trace features of the fire starting point;

[0030] 3D to 2D mapping modeling module: used to generate a set of 2D images through multi-view rendering projection of 3D data, and use convolutional neural networks to extract features such as texture features, color changes, and edge contours;

[0031] Feature fusion and association modeling module: used to deeply fuse three-dimensional spatial features with two-dimensional image features and establish a cross-modal feature association model;

[0032] Machine learning analysis module: used to train the collected data using deep learning algorithms and build a large model for analyzing fire locations and starting points;

[0033] Image intelligent comparison module: Based on the twin neural network architecture and the attention mechanism, it is used to compare the scene image with the database image, calculate the similarity and determine the location of the fire point;

[0034] Model dynamic optimization module: Introduces incremental learning mechanism, knowledge distillation technology and Bayesian optimization algorithm to dynamically adjust and optimize the model.

[0035] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the above-mentioned method for tracing the source of a fire location and a fire point.

[0036] After adopting the above technical solution, the present invention has at least one of the following beneficial effects compared with the prior art:

[0037] This invention uses 3D scanning to collect fire scene data and laboratory simulation data to construct a multidimensional and multimodal fire scene database, covering elements such as the location of the fire, the cause of the fire, the direction of spread, the location of the fire, the point of fire, the time of fire, and geographic information. The data is categorized and managed using a hierarchical labeling system and technical nomenclature to ensure the systematicity, standardization, and traceability of the information. On this basis, a large model system based on deep learning has been developed, breaking through the limitations of traditional reliance on human experience analysis and judgment, and promoting the intelligent and scientific nature of fire incident investigations.

[0038] The technical process of the system of the present invention includes data collection and database construction, fire scene trace feature extraction based on three-dimensional spatial association, two-dimensional mapping modeling, image intelligent comparison, model optimization and actual combat testing. The three-dimensional spatial trace feature extraction focuses on analyzing the fire spread direction, fire location and fire point characteristic traces. The specific method includes: using three-dimensional reconstruction and image processing technology to extract smoke trace characteristics, analyzing the characteristics and distribution patterns of heat loss traces of combustible and non-combustible materials, establishing the association between traces and heating conditions and heating processes, and inferring the fire spread path and direction; on this basis, combined with trace comparison and correlation analysis, a scientific basis is provided for tracing the fire location and fire point.

[0039] To accurately determine the origin of a fire, this method, after determining the direction of fire spread and the location of the fire, meticulously extracts the local features of typical heat damage traces of combustible and non-combustible materials at the fire point. Through 3D-to-2D mapping modeling, the 3D features are projected onto a 2D image. Combining a deep convolutional neural network and a twin network architecture, intelligent image comparison and feature similarity calculation are performed to accurately trace the origin of the fire. Furthermore, the system incorporates dynamic incremental learning and hyperparameter optimization mechanisms to continuously improve the model's accuracy and generalization capabilities, ensuring the system's reliability in various fire scenarios.

[0040] This invention establishes a scientific, efficient, and scalable fire origin tracing technology system through systematic data construction, refined feature extraction, deep learning model analysis, and dynamic optimization and updating. This system not only significantly improves the scientific nature and accuracy of fire investigations, but also provides strong technical support for public security, fire protection, emergency management, and housing and construction departments, making a positive contribution to public safety and judicial justice. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 This is a technical flow chart of the present invention, which shows the entire process of data collection, feature extraction, mapping modeling, comparison judgment and optimization. DETAILED DESCRIPTION

[0042] The following is combined with Figure 1 The present invention will be further described in detail with specific implementations to facilitate a clear understanding of the present invention, but they do not constitute a limitation to the present invention.

[0043] In the description of the present invention, it should be noted that the terms "upper", "lower", "front", "back", "left", "right", "vertical", "inside", "outside", etc., indicating directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operated in a specific direction. Therefore, they cannot be understood as limiting the present invention.

[0044] In the description of the present invention, it should be noted that, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood broadly. For example, they may refer to fixed, detachable, or integral connections; mechanical or electrical connections; and direct or indirect connections through an intermediary. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on the specific circumstances.

[0045] Example 1:

[0046] This embodiment uses a residential fire scene as an example to illustrate a fire location and point tracing method based on spatial correlation and intelligent analysis of fire scene trace images, including the following steps:

[0047] S 1. Data collection and database construction: Collect 3D scanning data and image data from the fire scene, combine it with laboratory simulated fires to generate a standard data set, establish a fire point database containing multi-dimensional and multi-modal information such as fire location, fire cause, fire location, fire point, and time, and perform standardized processing and labeling management;

[0048] Specifically, the data sources include:

[0049] 3D scanning data of fire-related incidents: Use 3D laser scanning equipment to obtain spatial data of real fire scenes, and fully record the smoke, combustion, and deformation information at the scene. In this embodiment, it is necessary to collect 10,000 fire scene data, covering various scenes such as residences, warehouses, and forests, and generate information including smoke marks, carbonized wood areas, metal discoloration areas, building material structural characteristics, combustion residue characteristics, thermal radiation and thermal convection characteristics, physical and chemical change characteristics, etc.

[0050] Laboratory simulation data: Experimental models of different fire causes and situations are constructed in a controlled environment, and 3D scanning technology is used to record the fire process and characteristic information of each stage. In this embodiment, scenarios such as electrical short circuits and ignition of flammable liquids are simulated in a closed combustion chamber to generate annotated 3D data sets. The annotations include the location of the fire point (accurate to the centimeter level), the smoke diffusion path (with the diffusion direction and speed marked), the carbonization depth of wood (with the carbonization depth marked at different temperatures), and the metal discoloration area (with the discoloration range under high temperature).

[0051] Data preprocessing: MeshLab and CloudCompare 3D processing software are used to remove noise, reconstruct discrete point clouds, and standardize the collected data. Feature labels such as the coordinates of the fire point, the location of the fire, and the cause of the fire are accurately labeled. The fire point and important feature locations are also labeled with multi-level, multi-dimensional labels. The database includes the fire location (such as residence, entertainment venue, storage warehouse, factory, etc.), the cause of the fire (electrical short circuit, open flame ignition, arson, equipment failure, etc.), the fire location (living room, bedroom, warehouse corner, etc.), the location of the fire point (socket, wiring, etc.), time information (daytime / nighttime), and geographic location (province, city, township, etc.). A technical nomenclature is used to classify the case data. Through this hierarchical labeling system, the system can quickly locate and analyze the fire characteristics in specific scenarios, facilitating subsequent rapid retrieval and model training. In this embodiment, the first level is the fire location (such as residence, entertainment venue, factory, etc.), the fire cause (such as electrical short circuit, arson, etc.), the fire location (such as living room, kitchen, etc.), and the fire point (such as socket, wiring, etc.). The second level is time (year, month, day, specific to the hour) and geographic information (province, city, town, specific to the street).

[0052] S2. Extract fire trace features in three-dimensional space, including trace features in the direction of spread, local features of the fire location, and local trace features at the fire point; extract fire-related spatial features from the three-dimensional reconstruction data using a deep learning model. The three-dimensional fire trace feature extraction specifically includes:

[0053] Extraction of trace features of the spread direction: Identify the carbonization, sooting, and melting trace features on the surfaces of combustible and non-combustible materials to infer the fire spread path; specifically, comprehensively survey the combustibles at the fire scene, clarify their types, distribution locations, and forms, observe the burning traces on the surface of combustibles, and consider the influence of factors such as heat conduction, convection, and radiation; analyze the thermal discoloration phenomenon on the surface of non-combustibles, the distribution and flow direction of melting traces, and the traces left on the non-combustibles by the burning of surrounding combustibles to infer the direction of fire spread.

[0054] Extracting trace features at the fire site: Analyze the material properties and local signs of deformation and damage at the fire site to establish a preliminary assessment of the fire's location. In this embodiment, based on determining the direction of fire spread, local features of the fire site are extracted, including its shape, size, and structural characteristics, such as whether there are cracks in the wall, whether doors and windows are intact, and whether electrical wiring is exposed. These features may be closely related to the cause of the fire. Simultaneously, the characteristics of the building materials and decorative materials at the fire site are analyzed. The combustion properties and thermal conductivity of different materials can affect the spread of the fire and the trace features at the fire site. Finally, the extracted trace features of the fire site are compared and correlated with known trace features in the database to identify similar feature patterns, providing a reference for determining the cause of the fire.

[0055] Extraction and comprehensive determination of fire point trace features: Refine the fire point trace features such as ash accumulation and electrical fault traces, and determine the fire point location based on the fire scene features. In this embodiment, trace features of the fire point are further extracted within the fire site, such as the ash accumulation form, the distribution center of residual liquid traces, and the fault point of the electrical equipment. These trace features are usually highly directional and can directly indicate the location of the fire point. Combining the trace features of the spread direction, the trace features of the fire site, and the trace features of the fire point, a comprehensive analysis and determination are performed to determine the final location of the fire point. The cause of the fire is inferred based on the trace features of the fire point, providing a scientific basis for the investigation and handling of fire accidents.

[0056] S3. Construct a 3D-to-2D image mapping model, projecting 3D data through multi-view rendering to generate a 2D image set, and use convolutional neural networks to extract texture features, color changes, and edge contour features. Feature fusion and association modeling: deeply fuse 3D spatial features with 2D image features to establish a cross-modal feature association model.

[0057] Among them, the mapping modeling of two-dimensional images adopts a method that combines orthogonal projection and perspective projection to convert three-dimensional spatial features into multi-perspective two-dimensional images, retaining spatial topological relationships, and uses convolutional neural networks (CNN) to extract and enhance features of the mapped two-dimensional images to ensure the integrity and robustness of the two-dimensional images in spatial representation capabilities.

[0058] Specifically, this embodiment includes the following methods:

[0059] Multi-view rendering and projection: Advanced multi-view rendering technology is used to project 3D data from multiple angles, generating a rich set of 2D images that display detailed information about the fire scene from different perspectives.

[0060] Maintain data integrity: During the entire projection process, ensure that the topological relationship and texture information of the original 3D data are not lost, and completely retain the key features of the original data, providing an accurate data basis for subsequent analysis.

[0061] Convolutional neural network feature extraction: Use convolutional neural network (CNN) to extract important features such as texture features, color changes, edge contours, etc. from the generated two-dimensional image, give full play to the advantages of CNN in image feature extraction, and obtain discriminative image features.

[0062] Feature fusion and association modeling: Deeply fuse the extracted 3D spatial features with 2D image features at the feature level, establish a cross-modal feature association model, achieve effective association between 3D spatial features and 2D image features, break the barriers between data dimensions, and provide a unified and comprehensive feature description for subsequent image comparison and analysis.

[0063] S4. Machine learning analysis: Using deep learning algorithms, we train a large amount of collected fire scene data and laboratory simulation data to build a large-scale model for analyzing fire locations and starting points. By inputting a large amount of labeled three-dimensional spatial feature data and two-dimensional image data, the model learns the characteristic patterns and regularities of fire locations and starting points in different fire scenarios, enabling accurate analysis and judgment of new fire scene data. Based on the twin neural network architecture and the attention mechanism, we perform intelligent image comparison, comparing scene images with database images, calculating similarity, and determining the location of the fire point.

[0064] Among them, the intelligent image comparison uses a twin network combined with an attention mechanism to perform high-dimensional feature matching and fine-grained similarity judgment between the target image and the database image; the similarity judgment adopts a comprehensive judgment method that combines cosine similarity and Euclidean distance weighted fusion, thereby improving the accuracy and stability of the fire point location matching. The details are as follows:

[0065] Network Architecture Design: Based on the powerful Siamese Network architecture and incorporating an attention mechanism, an image comparison network is constructed. The target scene image and sample images from the database are simultaneously fed into the network. The attention mechanism automatically focuses on key areas in the image, such as trace features near the fire point, highlighting key information and improving comparison accuracy.

[0066] Similarity calculation: By calculating the cosine similarity and Euclidean distance between images and using a weighted fusion method, the degree of matching between images is comprehensively obtained, and the similarity between the target scene image and the database image is quantified, providing a reliable numerical basis for judgment.

[0067] Determination and Output: A preset similarity threshold is set. When the calculated similarity exceeds this threshold, the fire point location is determined to match the database. The 3D coordinates of the fire point and the matching confidence level are output, providing fire investigators with intuitive and accurate fire point location information, assisting in the investigation and handling of fire accidents. In this embodiment, when the similarity exceeds a threshold (e.g., 0.85), the fire point is determined to match, and the 3D coordinates of the fire point and the confidence level (e.g., 92%) are output.

[0068] S5. Model optimization and dynamic learning: Through a dynamic incremental learning mechanism, new data is continuously introduced to optimize model performance and improve accuracy and generalization capabilities.

[0069] The model optimization specifically includes:

[0070] Incremental learning: Incorporating newly collected fire case data, the deep neural network is continuously updated in small steps to enhance the model's adaptability to new types of fire scenarios. Specifically, this involves introducing an incremental learning mechanism monthly to continuously incorporate new fire scenario data and dynamically adjust and update the model. By dynamically adjusting the deep network parameters, the model can continuously adapt to emerging fire scenarios and feature changes, improving its generalization capabilities in complex and changing environments and ensuring its long-term effectiveness and accuracy in practical applications.

[0071] Knowledge distillation: By migrating knowledge from large, complex models to lightweight sub-models, we improve model deployment efficiency while maintaining judgment accuracy and adapting to the operational needs of terminal devices. Specifically, knowledge distillation technology is used to transfer the rich knowledge and experience learned from large models to lightweight models, achieving both lightweight and efficient models. This makes lightweight models easier to deploy and apply on field devices while retaining key knowledge, expanding the system's practical application scenarios and scope.

[0072] Automatic Hyperparameter Optimization: Utilizes the Bayesian optimization algorithm to automatically tune key hyperparameters such as the learning rate, regularization coefficient, and loss function weight to improve the model's overall training efficiency and predictive performance. Specifically, the Bayesian optimization algorithm automatically adjusts hyperparameters such as the learning rate and loss function weight. Based on the model's performance during training and validation, the hyperparameter values ​​are dynamically optimized, continuously improving the model's overall performance and generalization capabilities. This ensures the model maintains optimal performance in various fire scenarios, providing more reliable fire source tracing results for fire investigations.

[0073] In actual fire investigation applications, the system operates through the following process:

[0074] On-site data collection: In actual fire investigations, high-quality 3D scanning data and image data are first collected at the fire scene to obtain comprehensive information about the fire scene and provide a data basis for subsequent analysis.

[0075] Automatic feature extraction: The system automatically processes the collected data, extracts the direction of fire spread and preliminary features of the fire location, quickly narrows the scope of the investigation, and determines the key analysis area.

[0076] Machine Learning Analysis: Utilizing deep learning algorithms, the system is trained on collected fire scene data and laboratory simulation data. By inputting a large amount of labeled 3D spatial feature data and 2D image data, a large-scale model for analyzing fire locations and starting points is constructed. The model learns the characteristic patterns and patterns of fire locations and starting points in different fire scenarios. This enables the system to accurately analyze and judge new fire scene data, providing strong support for subsequent image comparison and fire starting point determination.

[0077] Intelligent image matching: Through intelligent image matching technology, images collected on-site are quickly and accurately matched with images in the database. Utilizing pre-trained models and algorithms, the fire point can be accurately identified in a large amount of sample data, thereby improving the efficiency and accuracy of fire investigations.

[0078] Result output and application: The final output is the spatial coordinates of the fire location and the inference results of the fire cause, providing a scientific and accurate basis for fire responsibility determination and related law enforcement work, and assisting relevant departments in handling fire accidents and determining responsibility in a fair and reasonable manner.

[0079] Example 2

[0080] A fire location and point tracing system based on spatial correlation and intelligent analysis of fire scene trace images, which adopts the fire location and point tracing method of embodiment 1 above, includes:

[0081] Data acquisition device: used to collect 3D scanning data, image data and laboratory simulation data of the fire scene;

[0082] Data storage unit: used to build a fire point database containing multi-dimensional data such as fire location, fire cause, fire location, fire point, time information and geographical location;

[0083] Trace feature extraction module: used to extract fire-related spatial features from 3D reconstruction data, including trace features of the spread direction of combustible and non-combustible materials, trace features of the fire location, and trace features of the fire starting point;

[0084] 3D to 2D mapping modeling module: used to generate a set of 2D images through multi-view rendering projection of 3D data, and use convolutional neural networks to extract features such as texture features, color changes, and edge contours;

[0085] Feature fusion and association modeling module: used to deeply fuse three-dimensional spatial features with two-dimensional image features and establish a cross-modal feature association model;

[0086] Machine learning analysis module: used to train the collected data using deep learning algorithms and build a large model for analyzing fire locations and starting points;

[0087] Image intelligent comparison module: Based on the twin neural network architecture and the attention mechanism, it is used to compare the scene image with the database image, calculate the similarity and determine the location of the fire point;

[0088] Model dynamic optimization module: Introduces incremental learning mechanism, knowledge distillation technology and Bayesian optimization algorithm to dynamically adjust and optimize the model.

[0089] Example 3

[0090] This embodiment verifies the accuracy and efficiency of the system through comparative experiments and compares its performance with traditional methods.

[0091] 1. Experimental Design

[0092] Dataset preparation:

[0093] A scanner was used to collect 3D data from 100 fire scenes. 20 different fire scenarios were simulated in the laboratory to generate annotated 3D datasets.

[0094] Experimental objective: To verify the accuracy and efficiency of the system in determining the fire point in a complex fire scene.

[0095] 2. Experimental Procedure

[0096] Data preprocessing: noise removal and feature annotation (technical nomenclature) of the collected 3D data.

[0097] Feature extraction: The PointNet++ algorithm is used to extract key spatial features from preprocessed 3D data. This algorithm can effectively capture local and global features in 3D data, providing accurate feature descriptions for subsequent image comparison.

[0098] Image Comparison: Using a twin network combined with an attention mechanism, the target scene image is matched with the database image. Inputs are the target scene image (2D) and the database image (2D), and the output is the similarity score and the location of the fire.

[0099] Result evaluation: Compare the accuracy (the proportion of correctly identified fire points), recall rate (the proportion of actual fire points identified) and processing time (the time required for a single processing) of the traditional method (such as experience judgment) and the method of the present invention.

[0100] 3. Experimental Results

[0101] Experimental results show that the method of the present invention has significant improvements in accuracy and recall rate compared with traditional methods, indicating that the system can more accurately determine the fire point; at the same time, the single processing time of the method of the present invention is significantly shortened, reflecting its advantage in processing efficiency.

[0102] 4. Results Analysis

[0103] Advantages: The system of the present invention fully leverages the advantages of deep learning technology by cleverly combining three-dimensional spatial features with two-dimensional image features. It can more accurately capture the key features of a fire scene, thereby significantly improving the accuracy and efficiency of determining the fire location and starting point. It effectively overcomes the limitations of traditional methods in terms of subjectivity and efficiency, and is particularly suitable for tracing the source of fire points at complex fire scenes.

[0104] Limitations: In extremely complex scenarios (e.g., multiple fire points), the system may need to further optimize model parameters to improve performance.

[0105] 5. Comparative Verification

[0106] Traditional methods: rely on the experience of investigators, are highly subjective, and are inefficient;

[0107] The method of the present invention is based on deep learning and three-dimensional spatial analysis, has strong objectivity and high efficiency, and is suitable for complex fire scenes.

[0108] Example 4

[0109] A computer-readable storage medium of this embodiment stores a computer program, which, when executed by a processor, implements the method for tracing the source of the fire location and point of fire of the above-mentioned embodiment 1.

[0110] The above is merely a preferred embodiment of the present invention and does not constitute any formal limitation on the structure of the present invention. The layout and number of the present invention are not limited to this example and can be optimized according to actual engineering practices. Any modifications, equivalent changes, and decorations to the above embodiment based on the technical principles of the present invention that do not depart from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A method for tracing the origin of a fire location, based on spatial correlation and intelligent analysis of fire scene trace images, is characterized by: The steps include: S1. Data Collection and Database Construction: Collect 3D scanning data and image data from the fire scene, combine it with laboratory simulated fires to generate a standard data set, and establish a fire point database containing multi-dimensional and multimodal information such as fire location, fire cause, fire location, fire point, and time. This database is then standardized and managed using labels. S2. Extract fire trace features in three-dimensional space, including trace features in the direction of spread, local features of the fire location, and local trace features at the fire point; S3. Build a 3D-to-2D image mapping model, project the 3D data through multi-view rendering to generate a 2D image set, and use a convolutional neural network to extract texture features, color changes, and edge contour features. feature Fusion and association modeling: deeply integrate 3D spatial features with 2D image features to establish a cross-modal feature association model; S4. Machine learning analysis: Using deep learning algorithms, we train a large amount of collected fire scene data and laboratory simulation data to build a large-scale model for analyzing fire locations and starting points. By inputting a large amount of labeled three-dimensional spatial feature data and two-dimensional image data, the model learns the characteristic patterns and regularities of fire locations and starting points in different fire scenarios. Based on the twin neural network architecture and the attention mechanism, we perform intelligent image comparison, comparing the scene images with the database images, calculating the similarity and determining the location of the fire point. S5. Model optimization and dynamic learning: Through a dynamic incremental learning mechanism, new data is continuously introduced to optimize model performance and improve accuracy and generalization capabilities.

2. The method for tracing the source of a fire location according to claim 1, characterized in that: In S1, the data sources include: 3D scanning data of fire-related incidents: Use 3D laser scanning equipment to obtain spatial data of the real fire scene, and fully record the smoke, combustion, and deformation information at the scene; Laboratory simulation data: Build experimental models of different fire causes and situations in a controlled environment, and use 3D scanning technology to record the fire process and characteristic information at each stage; Data preprocessing: MeshLab and CloudCompare 3D processing software were used to remove noise, reconstruct discrete point clouds, and standardize the collected data. At the same time, multi-level and multi-dimensional labels were added to the fire points and important feature locations.

3. The method for tracing the fire location and starting point according to claim 2, characterized in that: In S2, the three-dimensional fire scene trace feature extraction specifically includes: Extraction of trace features of the spread direction: Identify the carbonization, sooting, and melting trace features on the surfaces of combustible and non-combustible materials to infer the fire spread path; Extraction of fire trace features: Analyze the material properties and local deformation damage signs of the fire site to establish a preliminary judgment of the fire location; Fire point trace feature extraction and comprehensive determination: Refine the fire point trace features and determine the fire point location based on comprehensive fire scene features.

4. The method for tracing the source of a fire location according to claim 3, characterized in that: In said S2, the fire point trace features are refined to include ash accumulation and electrical fault traces.

5. The method for tracing the fire location and starting point according to claim 4, characterized in that: In S3, the mapping modeling of the two-dimensional image adopts a method combining orthogonal projection and perspective projection to convert the three-dimensional spatial features into multi-view two-dimensional images, retaining the spatial topological relationship, and performing feature extraction and feature enhancement on the mapped two-dimensional image through the convolutional neural network CNN to ensure the integrity and robustness of the two-dimensional image in spatial representation capability.

6. The method for tracing the source of a fire location according to claim 5, characterized in that: In S4, the intelligent image comparison uses a twin network combined with an attention mechanism to perform high-dimensional feature matching and fine-grained similarity judgment on the target image and the database image; the similarity judgment adopts a comprehensive judgment method of weighted fusion of cosine similarity and Euclidean distance, thereby improving the accuracy and stability of the fire point location matching.

7. The method for tracing the fire location and starting point according to claim 6, characterized in that: In S5, the model optimization specifically includes: Incremental learning: Introducing newly collected fire case data, the deep neural network is continuously updated with small steps to enhance the model's adaptability to new types of fire scenarios; Knowledge distillation: By migrating knowledge from large and complex models to lightweight sub-models, we improve model deployment efficiency while maintaining judgment accuracy and adapt to the operational needs of terminal devices. Automatic hyperparameter optimization: Use the Bayesian optimization algorithm to automatically tune key hyperparameters such as learning rate, regularization coefficient, and loss function weight to improve the overall training efficiency and prediction performance of the model.

8. A fire location and origin tracing system based on spatial correlation and intelligent analysis of fire scene trace images, which adopts the fire location and origin tracing method of claim 7, characterized in that: include: Data acquisition device: used to collect 3D scanning data, image data and laboratory simulation data of the fire scene; Data storage unit: used to build a fire point database containing multi-dimensional data such as fire location, fire cause, fire location, fire point, time information and geographical location; Trace feature extraction module: used to extract fire-related spatial features from 3D reconstruction data, including trace features of the spread direction of combustible and non-combustible materials, trace features of the fire location, and trace features of the fire starting point; 3D to 2D mapping modeling module: used to generate a set of 2D images through multi-view rendering projection of 3D data, and use convolutional neural networks to extract features such as texture features, color changes, and edge contours; Feature fusion and association modeling module: used to deeply fuse three-dimensional spatial features with two-dimensional image features and establish a cross-modal feature association model; Machine learning analysis module: used to train the collected data using deep learning algorithms and build a large model for analyzing fire locations and starting points; Image intelligent comparison module: Based on the twin neural network architecture and the attention mechanism, it is used to compare the scene image with the database image, calculate the similarity and determine the location of the fire point; Model dynamic optimization module: Introduces incremental learning mechanism, knowledge distillation technology and Bayesian optimization algorithm to dynamically adjust and optimize the model.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for tracing the fire location and point of origin as described in any one of claims 1 to 7 is implemented.

Images