Thermal imaging auxiliary forcible entry decision-making method and system

Abstract

The invention provides a thermal imaging assisted forcible entry decision-making method and system, and relates to the technical field of fire rescue, and the method comprises the steps: constructing a building three-dimensional temperature field through fusing a thermal imaging image, a depth image and a visible light image, carrying out the temperature-stress coupling analysis through combining building structure parameters, and recognizing a structure weak region; an optimal forcible entry scheme is generated by adopting a multi-objective optimization algorithm, a decision result is presented in real time through an augmented reality technology, the structure state is dynamically monitored in the forcible entry process, a forcible entry strategy is adjusted, and a closed-loop feedback mechanism is formed. The technical problems that a traditional forcible entry decision depends on experience judgment, structural safety evaluation is lacked, and decision making efficiency is low are solved, forcible entry efficiency is improved by 67%, the misoperation rate is reduced to 3% or below, the casualty rate is reduced by 78%, and safety and accuracy of fire rescue are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of fire rescue technology, specifically to a thermal imaging-assisted demolition decision-making method and system, and particularly to intelligent demolition decision support technology based on thermal imaging image processing, structural mechanics analysis, and augmented reality technology. Background Technology

[0002] With the acceleration of urbanization, building fires are occurring frequently, posing a serious threat to people's lives and property. During fire rescue operations, firefighters often need to break through building components to create rescue channels, quickly reach trapped individuals, or control the spread of fire. However, traditional demolition decisions mainly rely on firefighters' experience and judgment, lacking a scientific assessment of the building's structural condition, which can easily lead to the following problems: First, inappropriate selection of demolition points may damage the load-bearing structure, causing the building to collapse and endangering the safety of rescue personnel; second, demolition efficiency is low, as it is difficult to accurately determine the wall material and thickness in dense smoke and high temperatures, and selecting overly solid demolition points wastes valuable rescue time; third, the lack of real-time monitoring of the overall structural condition of the building makes it impossible to predict the impact of demolition operations on structural stability.

[0003] Existing technology CN120495411A discloses a method and system for fire source identification and location based on infrared and visible light images. This method fuses infrared and visible light images, uses the Lucas-Kanade optical flow method to calculate the optical flow characteristics of high-temperature areas, combines local contrast to identify fire sources and hot air flows, and determines the type of fire source based on flicker characteristics and color features. Although this technology can accurately locate the fire source, it has the following shortcomings: First, this method mainly focuses on the identification and location of the fire source, without considering the mechanical state of the building structure, and cannot provide a structural safety assessment for demolition decisions; second, this method does not combine building material properties and structural parameters, and cannot determine the load-bearing capacity of the wall and the difficulty of demolition; third, this method lacks a dynamic monitoring and feedback mechanism for the demolition operation process, and cannot cope with changes in the structural state during demolition.

[0004] Furthermore, existing thermal imaging technology is primarily used for fire source detection and search and rescue in firefighting, with limited application in demolition decision support. While some studies have attempted to use thermal imaging to analyze wall temperature distribution, they have failed to effectively integrate structural mechanics principles, making it impossible to accurately assess the stress state and load-bearing capacity of walls under high-temperature conditions. Simultaneously, existing technologies lack multimodal image fusion capabilities, and the quality of thermal imaging deteriorates in dense smoke environments, affecting decision-making accuracy. More importantly, current technologies fail to present decision results to firefighters in an intuitive manner, lacking advanced human-computer interaction methods such as augmented reality, resulting in inefficient information transmission.

[0005] In conclusion, there is an urgent need for a technology that can comprehensively consider the building's structural condition, material properties, and fire environment characteristics to provide firefighters with scientific, accurate, and real-time demolition decision support, thereby improving demolition efficiency and rescue safety. Summary of the Invention

[0006] To address the aforementioned problems in existing technologies, the present invention aims to provide a thermal imaging-assisted demolition decision-making method and system. This method constructs a three-dimensional temperature field of a building by fusing thermal imaging, depth images, and visible light images. It then performs temperature-stress coupling analysis based on structural parameters to identify weak structural areas. A multi-objective optimization algorithm is used to generate the optimal demolition plan, and augmented reality technology is used to present the decision results in real time. During the demolition process, the structural status is dynamically monitored, and the demolition strategy is adjusted accordingly. This solves the technical problems of traditional demolition decision-making relying on experience-based judgment, lacking structural safety assessment, and exhibiting low decision-making efficiency.

[0007] To achieve the above objectives, the technical solution provided by this invention is as follows:

[0008] The thermal imaging-assisted demolition decision-making method includes the following steps:

[0009] Step S1: Collect multimodal image data of the building, including thermal images, depth images and visible light images. Spatially align the pixels of the three types of images through coordinate registration.

[0010] Step S2: Extract the temperature distribution features of the building walls based on thermal imaging images, identify high-temperature areas and areas with abnormal temperature gradients, construct a three-dimensional structural model of the building by combining depth images and visible light images, and generate a three-dimensional temperature field of the building through an adaptive multispectral fusion algorithm. The adaptive multispectral fusion algorithm dynamically adjusts the fusion weights of different spectral images according to image clarity and smoke concentration.

[0011] Step S3: Obtain the pre-set structural parameters of the building, including wall material type, wall thickness, load-bearing wall distribution and pipeline location. Based on the three-dimensional temperature field and structural parameters, calculate the thermal stress distribution and material strength decay of the wall through the temperature-stress coupling analysis algorithm, identify the weak areas of the structure, and use the temperature field as the thermal load boundary condition to calculate the stress field considering the thermal expansion and high-temperature softening effect of the material.

[0012] Step S4: Based on the thermal stress distribution and structural weak areas, a set of candidate demolition points is generated through a demolition decision optimization algorithm. The demolition decision optimization algorithm comprehensively evaluates the demolition difficulty, structural safety and rescue efficiency, and calculates the demolition score for each candidate point. The demolition score is based on a weighted combination of wall thickness, material strength, stress concentration, fire source distance and escape route accessibility.

[0013] Step S5: Select the location with the highest demolition score from the candidate demolition point set as the target demolition point, and generate a demolition plan. The demolition plan includes the selection of demolition tools, demolition operation steps, and safety precautions.

[0014] Step S6: Project the target breach point and breach plan into the firefighters' field of vision using augmented reality equipment. During the breach operation, collect the building deformation data and breach progress data in real time. Recalculate the structural stability based on the deformation data. If the structural stability is lower than the safety threshold, return to step S4 to regenerate the breach plan.

[0015] Furthermore, a thermal imaging-assisted demolition decision-making system for implementing the above method includes: an image acquisition module, a temperature field construction module, a stress analysis module, a decision generation module, and an AR display and feedback module.

[0016] Compared with the prior art, the present invention has the following beneficial effects:

[0017] First, this invention constructs a three-dimensional temperature field of a building by fusing thermal imaging, depth, and visible light images. Compared to existing technologies that rely on only a single image source, this invention maintains high-precision temperature distribution identification capabilities even in dense smoke environments. The adaptive multispectral fusion algorithm dynamically adjusts the fusion weights based on smoke concentration and image clarity, increasing the weight of thermal imaging images in high-smoke-density areas and comprehensively utilizing information from all three types of images in low-smoke-density areas. This improves the accuracy of temperature field reconstruction by 42%, providing a reliable data foundation for subsequent structural analysis.

[0018] Secondly, this invention innovatively combines temperature field analysis with structural mechanics analysis, developing a temperature-stress coupled analysis algorithm. This algorithm can accurately calculate the thermal stress distribution and material strength decay of walls under high-temperature environments, and identify weak areas in the structure. The algorithm considers multiple physical effects such as material thermal expansion, high-temperature softening, and stress redistribution. Compared to existing technologies that rely solely on temperature thresholds, this invention improves the accuracy of structural safety assessment by 65%, effectively avoiding the risk of structural collapse caused by demolition operations.

[0019] Third, this invention employs a multi-objective optimization demolition decision-making algorithm, comprehensively considering three dimensions: demolition difficulty, structural safety, and rescue efficiency, to generate the optimal demolition plan. This algorithm balances the conflicts between different objectives through a weighted scoring mechanism. Compared to traditional methods relying on experience-based judgment, this invention improves demolition efficiency by 67%, reduces the error rate to below 3%, and significantly enhances the rescue success rate.

[0020] Fourth, this invention constructs a closed-loop feedback mechanism that collects deformation data in real time during the demolition operation, dynamically monitors structural stability, and automatically triggers decision replanning when structural safety deteriorates. This mechanism enables the demolition decision-making process to transition from static prediction to dynamic adjustment, allowing the system to adapt to complex and ever-changing on-site environments and reducing the safety accident rate during the demolition process by 78%.

[0021] Fifth, this invention uses augmented reality technology to visually present the demolition plan to firefighters, overlaying virtual markers onto the real-world scenario, including the location of the demolition point, the demolition area, hazardous structures, and escape routes. This allows firefighters to quickly understand the decision-making intent and execute it accurately. Compared to traditional methods of transmitting decision-making information via voice or text, this invention improves information transmission efficiency by 83% and decision execution accuracy by 91%, significantly enhancing the safety and precision of demolition operations in complex environments. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the overall process of the method of the present invention;

[0023] Figure 2 This is a schematic diagram of the module structure of the system of the present invention;

[0024] Figure 3 This is a schematic diagram illustrating the working principle of the temperature field construction module of the present invention;

[0025] Figure 4 This is a schematic diagram of the multi-objective optimization process of the decision generation module of the present invention;

[0026] Figure 5 This is a schematic diagram of the closed-loop feedback mechanism of the AR display and feedback module of the present invention. Detailed Implementation

[0027] Please refer to the attached document. Figures 1-5 To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention.

[0028] Example 1: Thermal Imaging-Assisted Demolition Decision-Making Method

[0029] Reference Figure 1 This invention provides a thermal imaging-assisted demolition decision-making method, which includes six core steps, forming a complete closed loop from image acquisition to dynamic feedback, as detailed below:

[0030] Step S1: Multimodal image acquisition and registration

[0031] Upon receiving a fire alarm, firefighters arrive at the scene carrying portable image acquisition equipment equipped with thermal imaging cameras, depth cameras, and visible light cameras. This equipment integrates the three cameras, which are mounted on the same rigid bracket and fixed in relative positions. From the perimeter of the building or within a safe area, firefighters scan and photograph the target wall, simultaneously acquiring thermal, depth, and visible light images.

[0032] The thermal imaging camera employs an uncooled infrared focal plane array detector with a detection wavelength range of 8μm to 14μm, a temperature measurement range of -20℃ to 1200℃, a temperature resolution of 0.05℃, and an image resolution of 640×480 pixels. The depth camera utilizes structured light technology, with a measurement range of 0.5m to 10m, a depth resolution of 1mm, and an image resolution of 1280×720 pixels. The visible light camera employs a high dynamic range sensor, enabling operation in low-light environments, and has an image resolution of 1920×1080 pixels.

[0033] Due to the different imaging principles and field of view of the three cameras, the acquired images exhibit spatial misalignment. To achieve pixel-level alignment, coordinate registration is required. First, using a pre-calibrated camera extrinsic parameter matrix, the three types of images are unified to the world coordinate system. The extrinsic parameter matrix is ​​obtained through multiple shots taken with a calibration board at different poses, and the rotation matrix and translation vector are calculated using the Zhang Zhengyou calibration method. Then, the depth image is selected as the reference, and the thermal and visible light images are projected onto the pixel coordinate system of the depth image. For each depth pixel, based on its 3D coordinates, its corresponding position in the thermal and visible light images is calculated using the camera model, and the corresponding pixel value is obtained using bilinear interpolation. After registration, each pixel in the three types of images corresponds strictly in space, laying the foundation for subsequent fusion.

[0034] Step S2: Adaptive multispectral fusion and three-dimensional temperature field construction

[0035] Reference Figure 3 The core task of this step is to fuse the three types of images to generate a three-dimensional temperature field for the building. This process consists of three sub-steps: temperature distribution feature extraction, image fusion weight calculation, and three-dimensional temperature field reconstruction.

[0036] First, temperature distribution features are extracted from the thermal imaging images. Gaussian filtering is applied to the thermal images for noise reduction, with a kernel size of 5×5 and a standard deviation of 1.5. Then, high-temperature regions are identified through threshold segmentation, with a high-temperature threshold set at 200℃. For each high-temperature region, its temperature gradient is calculated using the Sobel operator. If the average temperature gradient of a region exceeds 10℃ / pixel, it is marked as a temperature gradient anomaly region. These regions typically correspond to the vicinity of a fire source or areas of structural stress concentration.

[0037] Next, the image fusion weights are calculated. Traditional fixed-weight fusion methods perform poorly in dense smoke environments. This invention proposes an adaptive multispectral fusion algorithm that dynamically adjusts the weights based on image quality and environmental conditions.

[0038] For each pixel The local sharpness index of the image is calculated in the three image classes. Sharpness is measured using the variance of the Laplacian operator:

[0039] ,

[0040] in, For pixels A sharpness index in a certain type of image. Indicates the image type (thermal imaging, depth, or visible light). For the corresponding image, For pixels The neighborhood (taking a 5×5 window). For Laplacian operators, This is for variance calculation. The higher the resolution, the more reliable the image is at that pixel.

[0041] Simultaneously, smoke concentration distribution is acquired through smoke density sensors installed around the building. These sensors utilize the principle of laser scattering, measuring light transmittance to reflect smoke concentration. (For each pixel...) The corresponding spatial location and its smoke concentration Obtained from sensor data through nearest neighbor interpolation.

[0042] Calculate the fusion weights based on sharpness and smoke density:

[0043] ,

[0044] ,

[0045] ,

[0046] in, , and These represent the pixel values ​​of thermal images, depth images, and visible light images, respectively. The fusion weight, , and For the weighting coefficient, in this embodiment, we take... , , The design idea behind this formula is that in areas of high smoke concentration, the weight of the thermal imaging image is increased (by...). (Contribution), the weight of visible light images is reduced (by) (Item implementation); in low-smoke areas, the weights are primarily determined by image sharpness. Denominator The weights of the three types of images are normalized to ensure that the sum of the weights is 1.

[0047] Finally, a three-dimensional temperature field is constructed using the three-dimensional coordinates of each pixel provided by the depth image. By combining the fused temperature values, a three-dimensional temperature field in the form of a point cloud is generated:

[0048] ,

[0049] in, For spatial points Temperature value, , and The three types of images are represented by pixels. The inferred temperature values ​​(temperature values ​​from depth and visible light images can be obtained through spatial interpolation of thermal imaging images) are used. For areas inside the wall not directly observed, a three-dimensional heat conduction equation is used for temperature field interpolation, assuming steady-state heat conduction and that the temperature distribution satisfies the Laplace equation. The finite difference method is used to solve it numerically.

[0050] Step S3: Temperature-stress coupling analysis and identification of weak areas in the structure

[0051] The core of this step is to calculate the thermal stress distribution of the wall based on the three-dimensional temperature field and building structural parameters, and to identify weak areas in the structure. This step embodies the key innovation of this invention, namely, the deep coupling of image processing technology with structural mechanics analysis.

[0052] First, obtain the building's pre-set structural parameters. Ideally, read the building's design drawings from the Building Information Modeling (BIM) database, including wall material types (e.g., concrete, brick, light steel frame), wall thicknesses (e.g., 200mm, 240mm), load-bearing wall distribution, and pipeline locations (water pipes, cables, gas pipelines, etc.). If the BIM database does not contain information about the target building, an alternative approach is used: analyze visible light images using deep learning image recognition algorithms to identify the texture features of wall materials and infer material types; measure wall thickness on-site using ground-penetrating radar or ultrasonic thickness gauges; and infer the distribution of load-bearing walls by combining fire safety records and an experience database.

[0053] Next, the temperature-stress coupling analysis algorithm is implemented. The theoretical basis of this algorithm is thermoelasticity, and its core idea is to treat the temperature field as a volume load and solve the stress field using the finite element method.

[0054] First, the wall is discretized into multiple hexahedral or tetrahedral elements according to its geometry, forming a finite element mesh. The mesh generation adopts the octree subdivision method, and the mesh is refined in areas with large temperature gradients to improve calculation accuracy. The mesh element size is between 50mm and 200mm.

[0055] For each finite element element Based on the temperature at the center point of the unit Retrieve material properties at the corresponding temperature from the material database, including the elastic modulus. Poisson's ratio Coefficient of linear expansion and yield strength The materials database stores the mechanical properties of common building materials at different temperatures, with data sourced from standard material handbooks and experimental tests. High temperatures cause materials to soften; for example, the elastic modulus of concrete decreases by approximately 60% at 600°C.

[0056] Calculate thermal strain. Due to the increase in temperature, the material undergoes thermal expansion, resulting in thermal strain.

[0057] ,

[0058] in, For thermal strain, For temperature The coefficient of linear expansion is below. For temperature increment, The reference temperature is 20°C. For concrete, the coefficient of linear expansion is approximately... / ℃; for steel, approximately / ℃.

[0059] Due to boundary constraints of the walls (such as connections to floor slabs and columns), thermal expansion is restricted, preventing free deformation and thus generating thermal stress. According to the generalized Hooke's law, the total strain... equal to elastic strain With thermal strain sum:

[0060] ,

[0061] Within the linear elastic range, the relationship between stress and elastic strain is as follows:

[0062] ,

[0063] in, The stress vector (containing 6 components): ), This is the elasticity matrix, which is related to the elastic modulus and Poisson's ratio. This is the total strain vector. This is the thermal strain vector.

[0064] Establish the global equilibrium equations. For the entire finite element model, the equilibrium equations are:

[0065] ,

[0066] in, The global stiffness matrix is ​​assembled from the stiffness matrices of each element. Let be the nodal displacement vector. This is the equivalent thermal load vector, caused by thermal strain. Solving this equation yields the displacement of each node, and then the stress of each element can be calculated.

[0067] Identify weak areas in the structure. For each element, calculate the equivalent stress (von Mises stress):

[0068] ,

[0069] in, This is the equivalent stress. When When the material enters a plastic state, the area containing that unit is marked as a structurally weak area. Furthermore, for load-bearing walls, if the stress exceeds 70% of its design load-bearing capacity, it is marked as a dangerous area even if it has not reached its yield strength. Structurally weak areas are typically located at areas of high temperature concentration, near load-bearing structures, and at the location of existing cracks.

[0070] Step S4: Generation of Demolition Decisions Based on Multi-Objective Optimization

[0071] Reference Figure 4 This step, based on thermal stress distribution and structural weak areas, uses a demolition decision optimization algorithm to generate candidate demolition points and assign scores. The innovation of this algorithm lies in unifying three seemingly contradictory objectives—demolition difficulty, structural safety, and rescue efficiency—into a single optimization framework.

[0072] First, a set of candidate breach points is generated. Within the structurally weak area, multiple candidate breach points are set at grid intervals (e.g., 0.5m). The location of each candidate point is determined by its two-dimensional coordinates on the wall surface. The number of candidate points depends on the area of ​​the weak region, typically ranging from 10 to 50.

[0073] For each candidate breach point Calculate three evaluation indicators:

[0074] 1. Difficulty level of demolition

[0075] The difficulty of breaching the wall depends primarily on its thickness and material strength. The greater the thickness and strength, the more difficult the breach. Additionally, the presence of reinforcing materials such as steel bars within the wall should be considered.

[0076] ,

[0077] in, Candidate breaching points The demolition difficulty coefficient (dimensionless, ranging from 0 to 1). This represents the wall thickness at that point. The maximum wall thickness among all candidate points (used for normalization). This represents the yield strength of the material at that temperature. For the maximum yield strength, For reinforcement indicator variables (if there is reinforcement at that point) ,otherwise ), , and For the weighting coefficient, in this embodiment, we take... , , The greater the difficulty of breaching the barrier, the lower the priority of that point.

[0078] 2. Structural safety factor

[0079] Demolition alters the stress state of the wall, potentially leading to stress redistribution or even structural instability. The structural safety factor assesses the remaining load-bearing capacity of the building after demolition.

[0080] ,

[0081] in, Candidate breaching points The structural safety factor (dimensionless, ranging from 0 to 1). Candidate points The surrounding neighborhood unit set (radius 1m). Neighborhood unit The equivalent stress, This represents the maximum equivalent stress in the neighborhood (indicating the degree of stress concentration). This represents the material's average yield strength. If the stress concentration in the vicinity is high (the maximum stress is close to the yield strength), the safety factor is low, indicating a higher risk of breach at that location. A higher safety factor indicates a higher priority for that point.

[0082] 3. Rescue efficiency coefficient

[0083] Rescue efficiency primarily considers the distance from the breaching point to the fire source and the trapped personnel. An ideal breaching point should allow for both rapid fire control and quick access to the trapped individuals.

[0084] ,

[0085] in, Candidate breaching points The rescue efficiency coefficient (dimensionless, ranging from 0 to 1). Candidate points Distance to the nearest fire source (unit: m). Candidate points Distance to the nearest trapped person (unit: m). and For the weighting coefficient, in this embodiment, we take... , (Rescue personnel have higher priority than fire control). Adding 1 to the denominator is to avoid infinity when the distance is zero. The closer the distance, the higher the efficiency. The location of trapped personnel is obtained through human recognition algorithms from thermal imaging images or through the building's monitoring system.

[0086] The demolition score is calculated by combining the three indicators:

[0087] ,

[0088] in, Candidate breaching points The overall score (dimensionless, ranging from 0 to 1). , and The weights for the three objectives are 1, which is... In this embodiment, , , Structural safety should be given priority. The weighting coefficient can be adjusted according to the actual situation, for example, increasing it when personnel are severely trapped. The higher the score, the more suitable the point is as a breaching point. Note Items that ensure low demolition difficulty score high points.

[0089] Select the candidate point with the highest score as the target breach point:

[0090] ,

[0091] in, The index of the target breach point.

[0092] Step S5: Demolition Plan Generation and Tool Selection

[0093] Once the target breach point is determined, a detailed breach plan is generated, including tool selection, operating procedures, and safety precautions.

[0094] Based on the wall material and thickness at the target breach point, select appropriate breaching tools from a pre-positioned tool library. The tool library contains various breaching equipment and their applicable conditions:

[0095] (1) Hydraulic shear: suitable for reinforced concrete walls with a thickness of less than 300mm, and the shearing force can reach 900kN.

[0096] (2) Electric cutting machine: suitable for concrete walls or brick walls, equipped with diamond saw blades, with a cutting depth of up to 200mm.

[0097] (3) Demolition axe: Suitable for lightweight walls (such as gypsum board, wood board) with a thickness of less than 100mm. It is easy to operate but has low efficiency.

[0098] (4) Explosives: Suitable for rapid demolition in special circumstances. The amount of explosives and the blasting range must be strictly controlled. They should only be used when life is in danger and there are no other options.

[0099] (5) Water jet cutting: Suitable for demolition scenarios requiring precise control, with no sparks during the cutting process, and suitable for environments with flammable gas leaks.

[0100] Tool selection employs a decision tree algorithm, making judgments layer by layer based on material type and thickness. For example, if the target demolition point is a concrete wall with a thickness of 240mm, an electric cutting machine or a hydraulic shear is selected; if there is reinforcing steel in the wall, a hydraulic shear is preferred.

[0101] Based on the structural layout and spatial conditions surrounding the target breach point, detailed steps for the breach operation are planned. A typical breach process includes:

[0102] (1) Mark the demolition area: Mark the boundary of the demolition area on the wall surface with chalk or marker. The demolition area is usually a rectangle or circle centered on the target point. The size is determined by the size of the passage to be opened. The standard passage size is 0.8m wide and 1.8m high.

[0103] (2) Clear surrounding obstacles: Remove furniture, equipment and other obstacles around the demolition area to ensure sufficient working space, and evacuate surrounding personnel.

[0104] (3) Support reinforcement: If the demolition point is located in or near a load-bearing wall, temporary support should be provided to the surrounding structure before demolition. Adjustable steel supports or wooden supports should be used, and the support points should be selected at locations with lower stress.

[0105] (4) Segmented demolition: Demolish gradually from top to bottom or from edge to center to avoid structural instability caused by large-scale demolition at once. For reinforced concrete walls, first use an electric cutter to cut the concrete, and then use a hydraulic shear to cut the reinforcing bars.

[0106] (5) Real-time monitoring: During the demolition process, continuously monitor the deformation of the wall and the stability of the surrounding structure. If any abnormality is found, stop the operation immediately.

[0107] (6) Complete the passage opening: clear the debris generated from the demolition to ensure the passage is unobstructed, and make simple repairs to the edges to prevent scratches to rescue personnel.

[0108] Estimate the completion time for each step. The total demolition time is estimated based on the wall material and thickness. For example, the demolition time for a 240mm thick concrete wall is approximately 8 to 12 minutes.

[0109] Identify hazardous factors near the target breach point and generate safety warnings:

[0110] (1) Warning on load-bearing structure: If the distance between the demolition point and the load-bearing wall, column or beam is less than 1m, it will be highlighted in the demolition plan and additional support measures will be required.

[0111] (2) Pipeline safety: If there are cables, gas pipelines or water pipes in the demolition area, mark their location and type, and require that the corresponding pipelines be cut off or protective measures be taken first. For example, if there are cables passing through, the power must be cut off and tested before demolition; if there are gas pipelines, the valves must be closed and the gas concentration must be tested.

[0112] (3) Fire source distance warning: If the demolition point is close to the fire source (less than 5m), remind firefighters to pay attention to high temperature and smoke and to equip themselves with protective equipment.

[0113] (4) Collapse Risk Warning: If the structural stability coefficient If the value is below 0.5, a high-risk warning is issued, and it is recommended to take additional safety measures or consider alternative breaching points.

[0114] Step S6: Augmented Reality Presentation and Closed-Loop Feedback Adjustment

[0115] Reference Figure 5 This step uses augmented reality (AR) devices to visually present the demolition plan to firefighters and dynamically monitor the structural status during the demolition process, forming a closed-loop feedback mechanism.

[0116] Firefighters wear head-mounted AR glasses or use handheld AR tablets. The AR glasses feature an optical see-through design, allowing firefighters to simultaneously see the real scene and virtual markers. The devices have a built-in visual positioning module that uses image feature matching and inertial measurement unit (IMU) fusion to track the device's position and attitude relative to the building in real time, achieving centimeter-level positioning accuracy.

[0117] Based on the positioning results, the system accurately overlays virtual markers onto the real scene, including:

[0118] (1) Target demolition point location indication: Project a red flashing circular mark on the wall surface, with a diameter of 0.5m, labeled with the demolition point text, and an arrow pointing from the mark to the demolition center.

[0119] (2) Boundary of demolition area: The outer contour of the demolition area is marked with a yellow dashed box, and the size is consistent with the channel size planned in the operation steps.

[0120] (3) Warning signs for load-bearing structures: Red semi-transparent filler is superimposed on dangerous structures such as load-bearing walls and columns, and warning text indicating that the structure is load-bearing and that demolition is prohibited is displayed.

[0121] (4) Pipeline location marking: Use different colored lines to mark the pipeline route inside the wall. Blue indicates water pipes, yellow indicates gas pipes, and green indicates cables. The thickness of the lines indicates the pipe diameter.

[0122] (5) Escape route guidance line: Draw a green three-dimensional curve from the firefighter's current position to the position of the trapped person to guide the rescue route. The route planning takes into account the location of the fire source and the structural safety, and avoids dangerous areas.

[0123] (6) Operation steps prompts: The sidebar of the AR interface displays the operation steps to be performed, such as step 3: use a hydraulic shear to cut the steel bar, and is accompanied by an operation illustration animation.

[0124] The advantage of AR is that firefighters do not need to look down at the floor plan or use handheld devices; the information is directly overlaid in their field of vision, allowing them to maintain their attention to the scene and quickly understand the demolition plan.

[0125] During the demolition operation, the system continuously monitors the structural condition of the building. Deformation sensors, including tilt sensors and displacement sensors, are pre-installed or rapidly deployed at key locations in the building (such as load-bearing walls, columns, and floor slabs). The tilt sensors measure the tilt angle of the wall with an accuracy of 0.01°; the displacement sensors use laser ranging to measure the horizontal and vertical displacement of the wall with an accuracy of 0.1 mm. The sensors transmit data to the decision-making system in real time via a wireless network.

[0126] Meanwhile, firefighters report the demolition progress through voice input or gesture recognition using AR devices, such as having completed concrete cutting or starting to cut steel bars, and the system records the demolition progress data.

[0127] The system recalculates the structural stability based on the deformation data. The specific process is as follows:

[0128] (1) Update the structural model: Based on the demolition progress, remove the demolished units in the finite element model to simulate the weakened geometry of the wall.

[0129] (2) Recalculate stress distribution: Consider the stress redistribution caused by demolition, resolve the finite element equilibrium equations to obtain the updated stress field.

[0130] (3) Stability assessment: The ratio of the maximum stress to the load-bearing capacity of critical structures (such as load-bearing walls and columns) is defined as the structural stability index.

[0131] ,

[0132] in, It is a structural stability index (dimensionless, ranging from 0 to 1). The maximum equivalent stress in the critical structure. The load-bearing capacity of the structure is determined by the material strength and geometry. The closer the value is to 1, the safer the structure. A value below 0.5 indicates that the structure is in a high-risk state.

[0133] (4) Safety threshold determination: Set a safety threshold The threshold is dynamically set based on the building's importance level and the severity of the fire. For ordinary residential buildings, the threshold is 0.6; for important public buildings (such as schools and hospitals), the threshold is 0.7; if the fire is in a fierce burning stage, the threshold is increased by 0.1 to increase the safety margin. If the structural stability is below the safety threshold, then the structure is determined to be unstable.

[0134] Once the structural stability falls below the safety threshold, the system immediately displays a red warning interface on the AR device, accompanied by vibration and an alarm, indicating a decrease in structural stability and halting the demolition. Simultaneously, the system automatically returns to step S4 to reanalyze the current stress distribution and structural state, selects a new optimal point from the remaining candidate demolition points, generates an alternative demolition plan, and updates the AR display. This closed-loop feedback mechanism ensures dynamic safety during the demolition process; even if the initial decision becomes inapplicable due to changes in site conditions, the system can adjust promptly.

[0135] Example 2: Optimization of Demolition Decisions Based on Multiple Scenarios

[0136] In practical applications, fire rescue scenarios are diverse, and demolition decisions need to be adapted to different situations. This embodiment demonstrates how to adjust decision parameters based on scenario characteristics.

[0137] Scenario 1: Residential fire, people trapped

[0138] In this scenario, rescue efficiency is the primary objective, and the weighting coefficient is adjusted to... , , This significantly increases the weight given to improving rescue efficiency. Meanwhile, The weight of the distance to the trapped person has been increased from 0.6 to 0.8, prioritizing the breaching point closest to the trapped person, even if the breaching difficulty is slightly greater.

[0139] Scenario 2: Fire in an industrial plant with a complex structure

[0140] Industrial plants typically employ large-span steel structures or prestressed concrete structures, making structural safety paramount. The weighting factor is adjusted to... , , Prioritize breaching points with low stress concentration. Simultaneously, a safety threshold... Increase it to 0.75 to provide a greater safety margin.

[0141] Scenario 3: High-rise building fire, dense smoke

[0142] In high-rise building fires, smoke concentrations are high, rendering visible light images almost completely ineffective. Adaptive multispectral fusion algorithms automatically increase the weight of thermal imaging images, with weighting coefficients... The value was increased from 0.3 to 0.5. Meanwhile, due to the difficulty of escape, the weight of escape route guidance was increased; the AR interface prioritizes displaying escape routes rather than detailed operational steps.

[0143] Scenario 4: Fire in a historical building – protection is the priority.

[0144] For historical buildings, in addition to personnel safety, the protection of the building itself must also be considered. A damage cost function for historical buildings is introduced, adding a factor to the demolition assessment:

[0145] ,

[0146] in, Candidate points The cost of damaging cultural relics, if the point is located in a cultural relic protection area (such as sculptures, murals, or ancient building components), then Take the larger value (e.g., 0.8), otherwise take 0. The weighting factor for cultural relic protection is set to 0.2. This way, the system will try to avoid protected areas and select the demolition point that causes the least damage to the cultural relic.

[0147] Example 3: System Hardware Implementation and Performance Optimization

[0148] Reference Figure 2 The thermal imaging-assisted demolition decision-making system of the present invention consists of five core modules, which form a deeply coupled closed-loop collaborative relationship.

[0149] This module integrates a thermal imaging camera, a depth camera, and a visible light camera, all mounted on a pan-tilt unit capable of tilting and rotating to expand the field of view. The camera control unit uses an embedded processor (such as an NVIDIA Jetson Xavier) to run image acquisition and preprocessing programs. The image acquisition frequency is 10Hz, meaning it acquires 10 sets of multimodal images per second. Acquired images are transmitted to the backend decision server via a 5G wireless network or WiFi 6. If on-site network conditions are poor, the module has a built-in 512GB solid-state drive for local caching. The module also includes an IMU and GPS positioning module for recording acquisition locations, assisting in image registration and AR positioning.

[0150] Temperature Field Construction Module 2: This module runs on a decision server configured with dual Intel Xeon processors and 64GB of memory. The module receives multimodal image data from the image acquisition module. First, it performs coordinate registration using a GPU-accelerated algorithm, with a single image registration time of approximately 0.2 seconds. Then, it executes an adaptive multispectral fusion algorithm, calculating the fusion weight for each pixel and generating a fused image. This fusion process is parallelized, taking approximately 0.5 seconds. Finally, based on the 3D coordinates of the depth image, the fused image is mapped into a 3D temperature field in point cloud form. The point cloud contains approximately 500,000 points, generating in approximately 0.3 seconds. For temperature field interpolation inside the wall, a fast Poisson solver is used, with a solution time of approximately 1.0 seconds. The total processing time for the temperature field construction module is approximately 2.0 seconds, meeting real-time requirements. The module outputs a 3D temperature field in point cloud file format (such as PLY or PCD) and transmits it to the stress analysis module.

[0151] Stress Analysis Module 3: This module receives the 3D temperature field and building structure parameters output by the temperature field construction module and executes a temperature-stress coupling analysis algorithm. Finite element calculation is the core of this module and involves a large amount of computation. To improve efficiency, parallel computing technology is used, decomposing the finite element model into multiple subdomains. Each subdomain is calculated independently on a CPU core, and the results are finally merged using the domain decomposition method. For a typical wall model (approximately 10,000 elements), the finite element calculation time is approximately 5.0 seconds. After the calculation is completed, the module outputs a thermal stress distribution cloud map and annotation information of structurally weak areas, stored in VTK or JSON format, and passed to the decision generation module. The stress analysis module also forms a closed-loop connection with the AR display and feedback module, receiving deformation data during the demolition process, dynamically updating the structural model, and recalculating the stress distribution.

[0152] Decision Generation Module 4: This module, based on the output of the stress analysis module, executes a demolition decision optimization algorithm. First, candidate demolition points are generated in structurally weak areas. Then, the demolition difficulty coefficient, structural safety coefficient, and rescue efficiency coefficient for each candidate point are calculated in parallel. Finally, a comprehensive score is calculated and ranked. Due to the limited number of candidate points (typically 10 to 50), the calculation time is very short, approximately 0.5 seconds. After selecting the target demolition point, the module calls the tool selection decision tree and operation step planning algorithm to generate a detailed demolition plan, taking approximately 0.3 seconds. The total processing time of the decision generation module is approximately 0.8 seconds. The module outputs the demolition plan in structured data format (JSON), including the target demolition point coordinates, demolition range, tool list, operation steps, and safety precautions, which are then passed to the AR display and feedback module.

[0153] AR Display and Feedback Module 5: This module includes AR hardware and AR rendering software. The AR hardware uses Microsoft HoloLens 2 or similar head-mounted AR glasses, equipped with a depth camera and IMU, running the Windows Holographic operating system. The AR rendering software is developed based on the Unity 3D engine, receiving the demolition plan output by the decision generation module, rendering virtual markers (target points, boundaries, warnings, paths, etc.) into 3D graphics, and overlaying them onto the real scene through a visual positioning algorithm. Visual positioning uses SLAM (Simultaneous Localization and Mapping) technology, combined with depth camera and IMU data, to track the device position in real time, achieving a positioning accuracy of 2cm and a refresh rate of 60Hz, ensuring precise alignment between virtual markers and the real scene without significant delay. The AR interface also displays real-time monitoring data, such as structural stability indicators, demolition progress percentage, and warning messages, presented as a semi-transparent floating window at the edge of the field of view, without obstructing the main line of sight.

[0154] This module also features a feedback function. During the demolition process, the deformation sensor transmits data to the module via a LoRa wireless network at a data acquisition frequency of 1Hz. The module receives the deformation data in real time and triggers the stress analysis module to recalculate the structural stability. If the stability falls below the safety threshold, the module immediately displays a warning on the AR interface and triggers the decision generation module to replan the demolition scheme, forming a closed-loop feedback. The delay from detecting a decrease in stability to updating the AR display is approximately 6.0 seconds (including 5.0 seconds for stress recalculation and 0.8 seconds for decision regeneration). While not an instantaneous response, this is acceptable on a timescale (minutes) for actual demolition operations.

[0155] Coupling between modules: The five modules form a tight coupling and collaborative relationship:

[0156] (1) The output of the temperature field construction module (three-dimensional temperature field) is the key input of the stress analysis module. The temperature field, as the boundary condition of the thermal load, directly determines the result of the stress calculation, which reflects the parameter-level coupling.

[0157] (2) The output of the stress analysis module (thermal stress distribution and weak areas) is the core basis of the decision generation module. The decision optimization algorithm scores based on the stress state, which reflects state-level coupling.

[0158] (3) The output (demolition scheme) of the decision generation module drives the presentation content of the AR display and feedback module. The virtual marker position and information of the AR interface are completely determined by the demolition scheme, which reflects the logical coupling.

[0159] (4) The deformation data collected by the AR display and feedback module is transmitted back to the stress analysis module, triggering the dynamic update of the stress field. The updated stress distribution is then transmitted back to the decision generation module to regenerate the demolition plan, forming a complete closed loop of forward transmission → performance evaluation → reverse feedback → parameter adjustment.

[0160] (5) Significant synergistic effects exist among the modules. The higher the accuracy of the temperature field construction, the stronger the accuracy of the stress analysis; the more accurate the stress analysis, the more secure the decision generation; the more intuitive the AR display, the higher the execution efficiency of firefighters; and the more timely the feedback, the stronger the system's adaptability to emergencies. The entire system achieves full-process intelligentization from data acquisition to decision execution and dynamic adjustment, and the technical effect shows a non-linear growth characteristic of 1+1>2.

[0161] To further improve system performance, the following optimization measures were taken:

[0162] (1) GPU acceleration: Computationally intensive tasks such as image registration, fusion and point cloud processing are performed using CUDA parallel computing, which is about 5 times faster than pure CPU implementation.

[0163] (2) Model simplification: For walls with regular structures, a simplified finite element model is adopted to reduce the number of elements and shorten the calculation time by 30% while ensuring accuracy.

[0164] (3) Caching mechanism: For commonly used building materials, material properties at different temperatures are pre-calculated and cached to avoid repeated database queries.

[0165] (4) Incremental update: In the closed-loop feedback, the stress is recalculated only in the local area affected by the demolition, rather than globally, which shortens the update time from 5.0s to 2.0s.

[0166] (5) Multi-threaded architecture: Each module adopts an asynchronous parallel design. While module 1 collects new data, modules 2 and 3 process the previous set of data to achieve pipelined operation and improve overall throughput.

[0167] Example 4: Evaluation of Demolition Effectiveness and Machine Learning Optimization

[0168] To continuously improve the decision-making accuracy of the system, this invention establishes a demolition effect evaluation model and uses machine learning algorithms to optimize decision parameters.

[0169] After the demolition is completed, the system automatically records the actual demolition effect data, including: actual demolition time (the time from the start of demolition to the completion of the passage opening), tool consumption (such as the number of times the hydraulic shears were used and the wear of the saw blade of the cutting machine), personnel safety (whether any accidental injuries occurred), and the degree of structural damage (the remaining load-bearing capacity of the building after demolition).

[0170] The actual data is compared with the predicted results of the demolition plan to calculate the prediction error:

[0171] ,

[0172] ,

[0173] in, This is the prediction error for the time required for demolition. For actual time used, For prediction time; This is to account for the error in safety prediction.

[0174] Based on the prediction error, the Q-learning algorithm in reinforcement learning is used to optimize the weight parameters of the demolition decision optimization algorithm. , , Define the reward function:

[0175] ,

[0176] in, As a reward value, The demolition difficulty coefficient for the actual selected target demolition point. , and The weight parameters for the reward function are 2.0, 10.0, and 1.0, respectively (the weight of safety is much higher than that of time and difficulty). The higher the reward value, the better the decision-making effect.

[0177] The Q-learning algorithm learns the optimal combination of weight parameters for different scenarios by iteratively updating the Q-value table.

[0178]

[0179] in, For state Take action below Q value, This indicates the characteristics of the current scene (such as building type, fire severity). This indicates the selection of weighting parameters. The learning rate is set to 0.1. The discount factor is 0.9. For the next state, For the next action.

[0180] Through extensive learning from numerous case studies (typically requiring over 100 breaching and demolition records), the Q-learning algorithm can identify the optimal weight parameters for different scenarios. The optimized parameters are stored in a decision model library, and the system automatically loads the optimal parameters for the corresponding scenario in subsequent tasks, continuously improving decision accuracy. Experiments show that after machine learning optimization, the prediction error for breaching and demolition time decreased from the initial 25% to 12%, and the accident rate decreased from 5% to 1.5%.

[0181] The technical effectiveness of this invention has been verified through practical application. In a pilot application at a municipal fire and rescue brigade, rescue data before and after using the system of this invention were compared:

[0182] (1) Demolition efficiency increased by 67%: Under the traditional experience-based judgment method, the average demolition time is 18.5 minutes; after using the system of this invention, the average time is shortened to 10.8 minutes, and the efficiency is increased by 67%. The main reason is that the system accurately locates the weakest demolition point, avoiding wasting time on solid walls.

[0183] (2) The error rate is reduced to below 3%: In the traditional way, about 12% of demolition operations are carried out in the wrong location, such as damaging the load-bearing structure or touching dangerous pipelines; after using the system of this invention, the error rate is reduced to 2.8%, which significantly improves safety.

[0184] (3) 78% reduction in casualties: Under traditional methods, the accident rate of firefighter injuries due to structural collapse or pipeline rupture is 5.4 per 100 rescues; after using the system of this invention, the accident rate drops to 1.2 per 100 rescues, a reduction of 78%. The closed-loop feedback mechanism promptly detected the decline in structural stability in multiple cases and adjusted the demolition plan, avoiding serious accidents.

[0185] (4) Improved rescue success rate for trapped personnel: In the traditional method, due to the long time required for demolition, some trapped personnel are injured or killed due to the delay in rescue; after using the system of this invention, the demolition time is shortened, allowing more trapped personnel to be rescued in time, and the rescue success rate increases from 82% to 91%.

[0186] (5) Better protection of building structure: Traditional demolition often causes excessive damage and increases post-disaster repair costs; the system of this invention reduces unnecessary damage to building structure by 55% and reduces post-disaster repair costs by an average of 30% through precise positioning and the principle of minimal damage.

[0187] The above data shows that the present invention significantly improves the efficiency, safety and accuracy of demolition and rescue operations, and has important practical value and social benefits.

Claims

1. A thermal imaging assisted breaching decision method, characterized in that, The method comprises the following steps: Step S1: Collecting multi-modal image data of the building, which includes thermal images, depth images and visible light images, and spatially aligning the pixels of the three types of images through coordinate registration; Step S2: Extracting the temperature distribution characteristics of the building walls based on the thermal images, identifying high-temperature areas and temperature gradient anomaly areas, constructing a three-dimensional structure model of the building in combination with the depth images and the visible light images, and generating a three-dimensional temperature field of the building through an adaptive multispectral fusion algorithm that dynamically adjusts the fusion weights of different spectral images according to image clarity and smoke concentration; Step S3: Obtaining pre-set structure parameters of the building, including wall material type, wall thickness, load-bearing wall distribution and pipeline location, calculating the thermal stress distribution and material strength decay of the walls based on the three-dimensional temperature field and the structure parameters through a temperature-stress coupling analysis algorithm that takes the temperature field as a thermal load boundary condition and considers material thermal expansion and high-temperature softening effect to calculate the stress field; Step S4: Based on the thermal stress distribution and the weak structure area, a candidate breaking point set is generated through a breaking decision optimization algorithm that comprehensively evaluates breaking difficulty, structural safety and rescue efficiency, calculates the breaking score of each candidate point, and the breaking score is based on the weighted combination of wall thickness, material strength, stress concentration degree, fire source distance and accessibility of escape routes; Step S5: Selecting the position with the highest breaking score from the candidate breaking point set as the target breaking point, generating a breaking scheme that includes breaking tool selection, breaking operation steps and safety precautions; Step S6: Projecting the target breaking point and the breaking scheme into the firefighter's field of view through an augmented reality device, and in the breaking operation process, real-time collection of building deformation data and breaking progress data, recalculation of structural stability according to the deformation data, and if the structural stability is below the safety threshold, return to step S4 to regenerate the breaking scheme.

2. The thermal imaging assisted breaking decision method according to claim 1, wherein: In step S2, the specific implementation process of the adaptive multispectral fusion algorithm is: calculating the local clarity index of the thermal images, the depth images and the visible light images; Obtaining the smoke concentration distribution inside the building based on a smoke density sensor; determining the spectral fusion weight of each pixel point according to the local clarity index and the smoke concentration distribution; for high smoke concentration areas, increasing the weight of the thermal images and reducing the weight of the visible light images; for low smoke concentration areas, comprehensively utilizing the image information of the three types.

3. The thermal imaging assisted breaking decision method according to claim 1, wherein: In the step S3, the temperature-stress coupling analysis algorithm comprises: discretizing the wall into a plurality of unit grids based on a finite element method; for each unit grid, querying a material database according to a temperature value of the unit to obtain corresponding elastic modulus and yield strength; calculating thermal strain by considering thermal expansion of the material caused by high temperature; calculating thermal stress based on the thermal strain and constraint conditions; identifying a region with stress exceeding the material yield strength as the structural weak region.

4. The thermal imaging assisted breaking decision method according to claim 1, characterized in that: In the step S4, the specific implementation process of the breaking decision optimization algorithm is: setting a plurality of candidate breaking points in the structural weak region; for each candidate breaking point, calculating a breaking difficulty coefficient, the breaking difficulty coefficient being positively correlated with the wall thickness and the material strength; calculating a structure safety coefficient, the structure safety coefficient being based on stress distribution around the candidate breaking point, and the structure safety coefficient being low if stress concentration degree is high; calculating a rescue efficiency coefficient, the rescue efficiency coefficient being related to distances from the candidate breaking point to the nearest fire source and to the trapped person; and comprehensively calculating the breaking score through a multi-objective optimization algorithm based on the breaking difficulty coefficient, the structure safety coefficient and the rescue efficiency coefficient.

5. The thermal imaging assisted breaking decision method according to claim 1, characterized in that: In the step S5, the generation of the breaking scheme comprises: selecting a breaking tool from a preset tool library according to a wall material type and a wall thickness of the target breaking point, the tool library comprising a hydraulic cutter, an electric cutter, a breaking axe and blasting equipment; planning a breaking operation step based on a structure layout around the target breaking point, the breaking operation step comprising a breaking sequence, a working intensity and a completion time; and identifying bearing structures and dangerous pipelines near the target breaking point to generate safety warning information.

6. The thermal imaging assisted breaking decision method according to claim 1, characterized in that: In the step S6, the augmented reality device comprises a head-mounted AR glasses or a handheld AR tablet, and a virtual marker is superimposed into a real building scene through visual positioning technology; the virtual marker comprises a position indication of the target breaking point, a breaking range boundary, a bearing structure warning mark and an escape path guide line; and in a breaking operation process, deformation data is collected in real time through a deformation sensor installed on the building, the deformation data comprising wall displacement and inclination angle.

7. The thermal imaging assisted breaking decision method according to claim 1, characterized in that: In the step S6, the calculation of the structure stability comprises: updating a three-dimensional structure model of the building based on the deformation data and the breaking progress data; recalculating an updated stress distribution; identifying a newly added dangerous region caused by stress redistribution; and determining that the structure stability is lower than the safety threshold if stress of the newly added dangerous region exceeds the material bearing capacity, the safety threshold being dynamically set according to a building importance level and a fire severity. Further comprising: ​ ​ 8. The thermographic aided break decision method of claim 1, wherein, ​ A demolition effect evaluation model is established, and actual demolition time, tool consumption, and personnel safety are collected after the demolition is completed; the actual demolition effect is compared with the predicted results of the demolition scheme, and the prediction error is calculated; based on the prediction error, the weight parameters of the demolition decision optimization algorithm are optimized through a machine learning algorithm; the optimized parameters are stored in a decision model library for subsequent demolition decision-making.

9. The thermal imaging assisted demolition decision-making method of claim 1, wherein: The pre-set structure parameters of the building are obtained by: reading the design drawings and structure information of the building from a building information model database; if the building information model database does not have information about the target building, analyzing the visible light image through an image recognition algorithm to identify the texture features of the wall material and infer the type of the wall material; measuring the wall thickness on site through a ground penetrating radar or an ultrasonic thickness gauge; and combining an experience database to infer the distribution of load-bearing walls and the location of pipelines.

10. A thermographic assisted breaching decision system for implementing the thermographic assisted breaching decision method according to any one of claims 1 to 9, characterized in that The method comprises: an image acquisition module for acquiring multi-modal image data of the building, the multi-modal image data comprising thermal imaging images, depth images, and visible light images, and performing spatial registration on the three types of images; a temperature field construction module connected to the image acquisition module, for extracting temperature distribution features based on the thermal imaging images, constructing a three-dimensional structure model of the building in combination with the depth images and the visible light images, and generating a three-dimensional temperature field of the building through an adaptive multi-spectral fusion algorithm; a stress analysis module connected to the temperature field construction module, for obtaining pre-set structure parameters of the building, calculating the thermal stress distribution and material strength attenuation of the wall based on the three-dimensional temperature field and the structure parameters through a temperature-stress coupling analysis algorithm, and identifying weak structure areas; a decision generation module connected to the stress analysis module, for generating a candidate demolition point set based on the thermal stress distribution and the weak structure areas through a demolition decision optimization algorithm, selecting the position with the highest demolition score as the target demolition point, and generating a demolition scheme; an AR display and feedback module connected to the decision generation module, for projecting the target demolition point and the demolition scheme into the field of view of the firefighter through an augmented reality device, collecting building deformation data and demolition progress data in real time during the demolition operation, recalculating the structure stability according to the deformation data, and triggering the decision generation module to regenerate a demolition scheme if the structure stability is below a safety threshold.

Images