A method for identifying and extracting building key parameter information based on multi-source images

Abstract

The application discloses a method for extracting building key parameter information based on multi-source image recognition, which obtains a building plan, a roof plan, a street facade plan and an infrared thermal imaging plan of the same building object, and extracts a building length-width ratio, an open space depth ratio and an outer facade window-wall ratio; a position conversion relationship is established based on corresponding features in the street facade plan and the infrared thermal imaging plan, corresponding positions of the outer window area, the wall area and the window-wall area boundary in the infrared plan are determined, and a temperature feature is combined to distinguish a thermal bridge area and a heat preservation defect area; corresponding temperature comparison areas are selected respectively, and an abnormal area, a maximum temperature difference and an average temperature difference are calculated; spatial parameters and thermal parameters are collected according to building object numbers, a structured data set is formed, and a building thermal risk index is calculated to determine a reconstruction priority.

Description

Technical Field

[0001] This invention belongs to the field of computer vision and building information processing technology, specifically relating to a method for extracting key building parameter information based on multi-source image recognition. Background Technology

[0002] In urban renewal and energy-saving renovation of existing buildings, it is necessary to obtain building spatial parameters and thermal parameters of the building envelope to provide basic data for building energy-saving diagnosis and priority determination. Existing methods for processing building floor plans, roof top views, and street view elevations can extract building outlines, spatial dimensions, window areas, and wall areas, but the obtained parameters are usually acquired and saved separately from infrared thermal detection results, making it difficult to form unified spatial and thermal parameter data for the same building object. Infrared thermal images can reflect the temperature distribution of building surfaces and extract abnormal areas based on temperature thresholds, local temperature differences, or temperature gradients, but infrared temperature information itself cannot accurately indicate whether the abnormal area is located at an external window, wall, or window-wall junction, which can easily lead to confusion between thermal bridges and insulation defects. Existing methods that combine visible light images with infrared thermal images are mostly used to establish image position correspondence, overlay infrared detection results, or exclude interfering areas such as external windows, without fully utilizing the external window areas, wall areas, and window-wall area boundaries in street view elevations for the final classification of infrared thermal anomalies. Furthermore, current infrared thermal parameter calculations typically use the average temperature of the entire building facade or the surrounding area of ​​anomalies as a uniform temperature comparison benchmark, without selecting normal comparison areas based on the differences in the facade areas where thermal bridges and insulation defects are located. Due to the different materials and surface conditions of exterior windows, window frames, and walls, uniformly selecting a temperature comparison benchmark can easily lead to deviations in the calculation of maximum and average temperature differences.

[0003] Therefore, a method is needed to extract key building parameters based on multi-source image recognition. This method would integrate building spatial parameter extraction, streetscape facade area identification, and infrared thermal detection into the same processing flow. This would allow the exterior window area, wall area, and window-wall area boundaries to participate in anomaly classification and temperature comparison area selection before the final determination of infrared thermal anomalies. Furthermore, the obtained spatial and thermal parameters would be aggregated according to the same building object, ultimately forming a typical building sample set and renovation priority levels. Summary of the Invention

[0004] The purpose of this invention is to overcome the problems in the prior art, such as the separate extraction of building spatial parameters and thermal parameters of the building envelope, the lack of clear facade area attribution for infrared temperature anomalies, the easy interference of temperature differences between different surface materials with the judgment of thermal anomalies, the lack of uniformity in temperature comparison benchmarks and the calculation range of anomaly areas, and the lack of clarity regarding typical buildings in urban renewal. The invention provides a method for extracting key building parameter information based on multi-source image recognition.

[0005] This invention is achieved using the following technical solution: Obtain and associate the building floor plan, roof top view, street view elevation and infrared thermal image of the same building object, and extract the building space parameters and the elevation area information including the exterior window area, wall area and window wall area boundary based on the building floor plan, roof top view and street view elevation; Establish a positional conversion relationship between street view facade images and infrared thermal images, determine the facade area information to the corresponding position in the infrared thermal images, and extract candidate areas for thermal anomalies; based on the positional relationship between the candidate areas for thermal anomalies and the boundaries of window and wall areas and wall areas, and combined with the temperature characteristics of the candidate areas for thermal anomalies, determine thermal bridge areas and insulation defect areas; determine the temperature comparison areas corresponding to the thermal bridge areas and insulation defect areas respectively, and extract the thermal parameters of the building envelope based on the corresponding temperature comparison areas; collect the building spatial parameters and building envelope thermal parameters according to the building object number to form a structured dataset of key building parameter information.

[0006] The advantages of this invention compared to the prior art are as follows: (1) The present invention determines the boundaries of the external window area, wall area and window-wall area identified in the street view facade image to the corresponding positions in the infrared thermal imaging image, so that the facade area information participates in the anomaly attribution and category judgment before the infrared thermal anomaly is finally determined, rather than just being used for the superimposed display of infrared detection results.

[0007] (2) Based on the positional relationship between the candidate thermal anomaly region and the boundary between the window and wall regions and the wall region, and combined with temperature characteristics and regional morphology, the present invention determines the thermal bridge region at the window-wall connection and the insulation defect region inside the wall, thereby reducing the confusion of anomaly types caused by relying solely on infrared temperature distribution.

[0008] (3) In this invention, normal areas with comparable temperatures are selected at the corresponding window-wall connection positions and within the same wall area for thermal bridge areas and insulation defect areas, respectively, as temperature comparison benchmarks. Based on this, abnormal areas, maximum temperature differences and average temperature differences are calculated to avoid interference from surface temperature differences of different materials, shading and reflection on the calculation results.

[0009] (4) The present invention further utilizes the window-to-wall ratio of the facade, the area ratio of thermal bridge area, the area ratio of insulation defect area and the relative temperature difference index to construct a building thermal risk index, and determines the priority of building renovation accordingly. Attached Figure Description

[0010] Figure 1 Here is a flowchart of a method for extracting key building parameters based on multi-source image recognition; Figure 2 A schematic diagram illustrating the process of identifying and determining street view facades from infrared thermal imaging images; Figure 3A diagram illustrating parameter aggregation, risk index, and renovation priority for the same building object; Figure 4 This is a comparison chart of traditional infrared detection methods and our proposed method. Detailed Implementation

[0011] The present invention will now be described in further detail with reference to the accompanying drawings: This invention identifies the building outline and calculates the building length-to-width ratio using architectural floor plans or roof top views, identifies interior space areas and calculates the bay-to-depth ratio using architectural floor plans, and identifies exterior window and wall areas and calculates the exterior window-to-wall ratio using street view elevations.

[0012] The boundaries of the exterior window area, wall area, and window-wall area in the street view facade image are determined to the corresponding positions in the infrared thermal image, so that the information of the above facade area and the temperature characteristics of infrared thermal anomalies can be used together to determine the thermal bridge area and the insulation defect area.

[0013] Areas without temperature anomalies are selected from the facade areas corresponding to the thermal bridge area and the insulation defect area, respectively, as temperature comparison benchmarks. The areas of the thermal bridge area and the insulation defect area, as well as the maximum and average temperature differences relative to their respective temperature comparison benchmarks, are calculated.

[0014] The building spatial parameters and thermal parameters of the building envelope are collected according to the building object number to form a structured dataset of key building parameter information, which is then used for building thermal risk assessment, renovation priority determination and selection of typical building samples.

[0015] Figure 1 This demonstrates a complete technical process from multi-source image acquisition, building spatial parameter extraction, street view facade area identification, infrared thermal anomaly classification, thermal parameter calculation to risk assessment. Its core lies in incorporating the external window area, wall area, and window-wall area boundaries identified in the street view facade image into the final determination process of infrared thermal anomalies. This allows facade area information to not only be used for image overlay but also directly participate in the classification of thermal bridges and insulation defects, the selection of temperature comparison areas, and the calculation of anomaly areas, thus forming a complete extraction chain that correlates spatial parameters with thermal parameters.

[0016] Figure 2 This paper focuses on the identification process of effective building facades, exterior window areas, wall areas, and window-wall area boundaries in street view elevation images. It establishes a positional conversion relationship between the street view elevation image and the infrared thermal image by using corresponding points or edges, thus determining the corresponding locations of the aforementioned facade areas in the infrared image. This process provides a clear spatial attribution for infrared anomalies, enabling subsequent detection to determine whether the anomaly is located at the window-wall connection or inside the wall. This is a fundamental step in accurately distinguishing between thermal bridges and insulation defects.

[0017] Figure 3 This demonstrates the unified collection process of building object numbers, building spatial parameters, and thermal parameters of the building envelope. Building aspect ratio, bay-to-depth ratio, and window-to-wall ratio, along with thermal bridge area, insulation defect area, area ratio, maximum temperature difference, and average temperature difference, are recorded as structured data under the same building object number, and a building thermal risk index is further constructed. Its innovative value lies in breaking away from the traditional method of separately acquiring and storing spatial parameters and infrared detection results, achieving aggregated processing of multi-dimensional parameters, and providing a unified data foundation for prioritizing building renovations.

[0018] Figure 4 The comparison demonstrates that traditional infrared detection primarily relies on temperature distribution to extract abnormal areas, lacking constraints from facade areas such as exterior windows, walls, and window-wall boundaries. This can easily lead to confusion between thermal bridges and insulation defects, and the calculated ranges for temperature comparison areas and abnormal areas are also unstable. This invention, however, first determines the streetscape facade area information onto an infrared thermal image, then classifies the abnormal areas based on their proximity to window-wall boundaries and the positional relationship with wall areas. Furthermore, it selects temperature comparison areas for different abnormal types, thereby improving the consistency and accuracy of abnormal category determination, temperature difference calculation, and area calculation.

[0019] Specific implementation steps: Step 1: Acquire raw architectural image data and number objects to construct the initial dataset. 1.1 Public Data Collection: By calling publicly authorized open interfaces, street view images and rooftop top views of buildings are filtered to extract images containing complete front elevations.

[0020] 1.2 On-site survey and data collection: Street view elevation images were collected using visible light imaging equipment capable of clearly identifying the exterior window areas, wall areas, and window-wall boundaries of the target building. In addition, relevant building elevation drawings can be obtained through authorized websites.

[0021] 1.3 Image acquisition requirements: Street view elevation images used to establish a quantitative positional correspondence with infrared thermal images are preferably acquired simultaneously using visible light imaging units and infrared imaging units; alternatively, street view elevation images and infrared thermal images can be acquired sequentially while keeping the position and shooting direction of the acquisition equipment unchanged.

[0022] 1.4 Building Object Numbering: A unified building object number is assigned to each building object. The building object number includes the area number, building number, and facade number, for example, P01-B015-F02, where P01 represents the area number, B015 represents the building number, and F02 represents the facade number. Architectural floor plans, roof top views, street view facades, infrared thermal images, and their subsequent output parameters are all associated through the building object number.

[0023] Step 2: Dataset Image Preprocessing and Region Segmentation 2.1 Color Space Conversion: Convert the input image from the BGR color space to the HSV color space, and use the hue, saturation, and brightness of the HSV space to separate image information.

[0024] 2.2 Threshold segmentation: The corresponding HSV threshold range is preset according to the labeled color, and the contour area filtering threshold is set according to the image resolution and target area scale.

[0025] 2.3 Morphological Closure Operation Optimization: Perform closure operation to eliminate small holes inside the target region and enhance region connectivity.

[0026] 2.4 Morphological opening operation optimization: Perform the opening operation to remove shot noise in the image while preserving the edge details of the target region.

[0027] 2.5 Contour Extraction and Filtering: Use contour detection functions to extract the contours of the target region, filter valid contours, and filter out noise regions with too small an area.

[0028] 2.6 Background Removal: Determine the effective facade area for the street view elevation drawing and exclude the sky, ground, vegetation, vehicles, equipment, doorways, and background areas unrelated to the target building.

[0029] 2.7 Infrared Image Preprocessing: Prioritize reading pixel-by-pixel temperature data from the infrared thermal image to obtain the temperature matrix. T (x, y) is calculated, and temperature correction, noise filtering, and invalid region elimination are performed. When only an infrared pseudo-color image is available, the temperature value can be approximately recovered from the temperature scale, but quantitative calculations are preferably performed using an infrared file with the original radiation temperature data.

[0030] Step 3: Calculation and analysis of geometric features and the window-to-wall ratio of the facade. 3.1 Determining the Main Facade Orientation: First, the minimum bounding rectangle of rotation is calculated. For each valid contour, the minimum bounding rectangle of rotation is calculated, obtaining the coordinates of its center point, width, height, and rotation angle. Finally, principal component analysis is performed. By analyzing the principal component orientations of the contour vertices, the spatial distribution characteristics of the contour are determined, aiding in the determination of the main orientation of the building facade.

[0031] 3.2 Direction determination: The horizontal and vertical directions are determined based on the width and height relationship of the rotated rectangle.

[0032] 3.3 Calculation of building space area parameters: Calculate the building length-to-width ratio and the span-to-depth ratio.

[0033] 3.4 Window-to-wall ratio ( WWR Parameter calculation:

[0034] When both the exterior window area and the wall area use the pixel area in the same calibrated image, the window-to-wall ratio of the facade can be calculated directly; when the actual dimensions of the facade are known, the pixel area can be converted into the actual area.

[0035] 3.5 Target Area Identification: Extract the window and wall area boundaries based on the adjacent edges of the exterior window area and the wall area, and output the building outline map, the main orientation map of the building, the interior space area map, the exterior window area map, the wall area map, and the window and wall area boundary map.

[0036] Step 4: Match the street view facade with the infrared thermal image. 4.1 Location Conversion: Select corresponding points or edges reflecting the same actual location from the street view elevation image and the infrared thermal image. These corresponding locations include exterior window corners, exterior window centers, doorway corners, intersections of floor boundary lines, and intersections of building facade edges. Location determination can be achieved using affine transformation, perspective transformation, feature point matching, edge matching, or manual correction.

[0037] 4.2 Manual Correction and Verification: When the reprojection error exceeds the preset allowable value, the corresponding position is reselected or manual correction is performed; images that fail the position verification are not directly included in subsequent thermal parameter calculations. Output the original street view facade image, infrared thermal image, corresponding position image, and verification image after overlaying the boundaries of the exterior window area, wall area, and window-wall area onto the infrared thermal image.

[0038] Step 5: Facade thermal guidance testing and parameter calculation 5.1 Region Identification: Candidate regions for thermal anomalies are extracted based on local temperature difference, temperature gradient, anomalous area, and regional continuity conditions. Oh i The candidate areas are then filled with holes, noise is removed, and adjacent areas are merged. A preset distance is extended outwards from the window-wall area boundary to both sides of the outer window area and the wall area, forming the adjacent range of the window-wall area boundary. M b The converted wall area is denoted as M w Calculate the overlap ratio between the i-th candidate thermal anomaly region and the adjacent area of ​​the window / wall region boundary. P bi : P bi =Area( Oh i ∩ M b ) / Area ( Oh i ) Calculate the overlap ratio of the i-th thermal anomaly candidate region within the wall region.P wi : P wi =Area( Oh i ∩ M w ) / Area ( Oh i ) when P bi When the preset boundary coincidence condition is met, and the candidate thermal anomaly region is continuously distributed along the boundary of the window wall region, and simultaneously meets the temperature difference or temperature gradient condition, it is identified as a thermal bridge region; when P wi To achieve the pre-set wall area overlap condition, P bi If the preset boundary overlap condition is not met, but the candidate area for thermal anomalies meets the temperature anomaly condition and the area morphology condition, it is identified as an insulation defect area. When the candidate area for thermal anomalies crosses the window area and the wall area, it is divided into different parts according to the window-wall area boundary, and judgment and parameter calculation are performed according to the corresponding positional relationship and temperature characteristics. When the candidate area only meets the positional condition but not the temperature anomaly condition, it is not directly identified as a thermal bridge or insulation defect; areas affected by strong reflection, shading, uneven sunlight, infrared image saturation, or positional correspondence errors are set as areas to be reviewed. The window area, wall area, and window-wall area boundary are used to determine the spatial affiliation and candidate category of the anomaly, while infrared temperature difference, temperature gradient, and area morphology are used to confirm whether the anomaly has the corresponding thermal anomaly characteristics. The area of ​​the thermal bridge area, the area of ​​the insulation defect area, and the effective building facade area should be calculated under the same calibration coordinate system; when using pixel area, it is expressed as the number of effective pixels within the coordinate range of the infrared thermal image.

[0039] 5.2 Thermal parameter calculation: For thermal bridge areas, within the boundary of the window and wall area corresponding to the thermal bridge area, the area that is not identified as thermal anomaly, is not at the edge of the image, and is not significantly obstructed or affected by reflection is selected as the thermal bridge temperature comparison area.

[0040] For areas with insulation defects, within the same wall area, areas that are not identified as thermal anomalies, whose distance from the abnormal area meets the preset requirements, and whose surface conditions are similar are selected as the temperature comparison areas for insulation defects.

[0041] Reference temperature of the temperature comparison area T ref,q for: T ref,q = (1 / N ref,q )ΣT (x, y) in, N ref,q This represents the number of valid pixels within the temperature comparison area.

[0042] For the q-th thermal anomaly region, its maximum temperature difference relative to the temperature comparison benchmark is: ΔT max,q =max| T (x, y) - T ref,q | Its average temperature difference is: ΔT ave,q = (1 / N q )Σ| T (x, y) - T ref,q | in, N q denoted as the number of effective pixels within the q-th thermal anomaly region.

[0043] Image area of ​​all thermal bridge regions A bridge,p for: A bridge,p =Σ N bridge,q in, N bridge,q This represents the number of effective pixels within the q-th thermal bridge region.

[0044] Image area of ​​all insulation defect areas A defect,p for: A defect,p =Σ N defect,r in, N defect,r Let be the number of effective pixels within the r-th insulation defect region.

[0045] When using image area calculation, the effective building facade area is: A facade,p The area proportions of thermal bridge regions and insulation defect regions are respectively: R bridge = A bridge,p / A facade,p R defect = A defect,p / A facade,p When the actual area of ​​the building facade A facade,actual Given that the actual areas of the thermal bridge region and the insulation defect region are respectively: A bridge = A bridge,p / A facade,p × A facade,actual A defect = A defect,p / A facade,p × A facade,actual When a building facade has multiple valid anomalous areas, the maximum temperature difference of the building facade is taken as the maximum value among the maximum temperature differences of all anomalous areas; the average temperature difference of the building facade is calculated by weighting the areas of each anomalous area. ΔT ave =Σ( A q · ΔT ave,q ) / Σ A q Output the ID, category, outline coordinates, area, area percentage, maximum temperature difference, average temperature difference, corresponding temperature comparison area, and data validity indicator for each abnormal region.

[0046] Step 6: Collection of key building parameters and calculation of building thermal risk index 6.1 Structured Dataset: The building's length-to-width ratio, bay-to-depth ratio, and window-to-wall ratio are used as building spatial parameters. The areas of thermal bridge zones, insulation defect zones, the proportion of thermal bridge zones, the proportion of insulation defect zones, the maximum temperature difference, and the average temperature difference are used as thermal parameters of the building envelope. These parameters are grouped into the same data record according to the building object number, forming a structured dataset of key building parameter information.

[0047] 6.2 Calculation of Building Thermal Risk Index: Relative temperature difference index ΔT norm for:

[0048] The building thermal risk index S is: S = a·WWR + β · R bridge + c · R defect + d · ΔT norm in, WWR For the ratio of windows to walls on the exterior facade, R bridge This represents the area percentage of the thermal bridge region. R defect This represents the percentage of the area with insulation defects. α , β , c and d is a non-negative weighting coefficient, and: α + β + c + d =1 The window-to-wall ratio of a building facade is used to characterize the sensitivity of the building facade to the thermal performance of the exterior windows and the window-to-wall connection area, and is not used as a direct basis for determining the existence of thermal defects.

[0049] The weighting coefficients are determined based on building type characteristics, sample data, and urban renewal objectives. The building thermal risk index is used for relative evaluation among buildings within the same evaluation batch and does not replace the final assessment stipulated in engineering testing standards.

[0050] Step 7: Screening of Typical Building Samples and Output of Results The purpose of this step is to select representative buildings from multiple buildings within the urban renewal area that can represent different spatial and thermal characteristics, and to output complete image and data results.

[0051] 7.1 Parameter Clustering and Typical Building Samples: From the structured dataset of key building parameters, the building aspect ratio, bay-to-depth ratio, window-to-wall ratio, thermal bridge area ratio, insulation defect area ratio, and relative temperature difference index were selected as building classification parameters. These parameters were then standardized in terms of dimensions. Based on these standardized parameters, building objects were clustered to determine each building category and its parameter centers. Within each building category, the building object with the smallest overall deviation value and all valid building classification parameters were selected as the typical building sample for that category.

[0052] 7.2 Supplementary Explanation: The building thermal risk index is not repeatedly used as a building classification parameter in clustering. Instead, it is used as the basis for thermal risk assessment and renovation priority after typical building samples are determined.

Claims

1. A method for extracting key building parameter information based on multi-source image recognition, characterized in that, include: Obtain and associate the building floor plan, roof top view, street view elevation and infrared thermal image of the same building object, and extract the building space parameters and the elevation area information including the exterior window area, wall area and window wall area boundary based on the building floor plan, roof top view and street view elevation; Establish the positional conversion relationship between street view facade image and infrared thermal image, determine the facade area information to the corresponding position in the infrared thermal image, and extract candidate areas of thermal anomalies; based on the positional relationship between the candidate areas of thermal anomalies and the boundaries of window and wall areas and wall areas, and combined with the temperature characteristics of the candidate areas of thermal anomalies, determine the thermal bridge area and the insulation defect area. Temperature comparison regions corresponding to thermal bridge areas and insulation defect areas are determined respectively, and thermal parameters of the building envelope are extracted based on the corresponding temperature comparison regions; building spatial parameters and thermal parameters of the building envelope are collected according to building object number to form a structured dataset of key building parameter information.

2. The method for extracting key building parameter information based on multi-source image recognition as described in claim 1, characterized in that, Thermal bridging areas and insulation defect areas are determined as follows: Extending a predetermined distance outwards from the boundary of the window-wall area to both sides of the window area and the wall area, a range adjacent to the boundary of the window-wall area is formed. M b The wall area is denoted as M w Let the i-th thermal anomaly candidate region be denoted as Ω i The overlap ratio between the candidate thermal anomaly region and the adjacent area of ​​the window / wall region, and the wall region, was calculated respectively. when P bi When the preset boundary coincidence condition is met, the candidate thermal anomaly region is continuously distributed along the boundary of the window wall region and meets the temperature anomaly condition, it is determined to be a thermal bridge region; when P wi To achieve the pre-set wall area overlap condition, P bi If the preset boundary coincidence condition is not met, and the candidate region of thermal anomaly meets the temperature anomaly condition and the region morphology condition, it is determined to be a thermal insulation defect region. The thermal bridge region and the thermal insulation defect region are collectively referred to as the thermal anomaly region. The thermal bridge temperature comparison region and the thermal insulation defect temperature comparison region are determined for the thermal bridge region and the thermal insulation defect region respectively, and the thermal bridge temperature comparison region and the thermal insulation defect temperature comparison region are collectively referred to as the temperature comparison region.

3. The method for extracting key building parameter information based on multi-source image recognition as described in claim 2, characterized in that: When a candidate area for thermal anomalies is located in both the window area and the wall area, it is divided according to the boundary of the window and wall areas. The anomaly category is determined based on the positional relationship and temperature characteristics between each part and the boundary of the window and wall areas and the wall area. If the preset boundary coincidence condition or the preset wall area coincidence condition is met, but the temperature anomaly condition is not met, it is not identified as a thermal bridge area or insulation defect area. Areas affected by reflection, shading, uneven sunlight, infrared image saturation, or positional correspondence error are identified as areas to be reviewed.

4. The method for extracting key building parameter information based on multi-source image recognition as described in claim 2, characterized in that: The thermal bridge temperature comparison area is selected from the adjacent range of the window and wall area corresponding to the thermal bridge area, and is not identified as a thermally abnormal area, is not at the edge of the image, and is not significantly obstructed or affected by reflection. The temperature comparison area for insulation defects is selected from the wall area where the insulation defect area is located. It is not identified as a thermally abnormal area, the distance between it and the corresponding insulation defect area meets the preset requirements, the surface conditions are similar and it is not significantly affected by obstruction or reflection.

5. The method for extracting key building parameter information based on multi-source image recognition as described in claim 4, characterized in that: The thermal anomaly regions are numbered, and the local reference temperature corresponding to the q-th thermal anomaly region is... T ref,q for: The summation range is the temperature comparison region corresponding to the qth thermal anomaly region; T (x, y) represents the temperature value at position (x, y) in the infrared thermal image; N ref,q This represents the number of effective pixels within the temperature comparison area. The maximum temperature difference between the qth thermal anomaly region and the local reference temperature ΔT max,q for: The maximum value ranges from the qth thermal anomaly region; The average temperature difference between the qth thermal anomaly region and the local reference temperature ΔT ave,q for: The summation range is the q-th thermal anomaly region. N q Let be the number of effective pixels within the q-th thermal anomaly region; the areas of thermal bridge regions, insulation defect regions, and effective building facade areas are calculated under the same calibrated coordinate system; based on the areas of thermal bridge regions, insulation defect regions, and effective building facade areas, the proportions of thermal bridge regions and insulation defect regions are obtained respectively; when there are multiple thermal anomaly regions on the building facade, the maximum temperature difference is... ΔT max Take the maximum value among the maximum temperature differences in each thermal anomaly region; average temperature difference Δ T ave The calculation is weighted according to the area of ​​each thermal anomaly region: in, A q Let q be the area of ​​the qth thermal anomaly region.

6. The method for extracting key building parameter information based on multi-source image recognition as described in claim 5, characterized in that: Thermal bridge area percentage R bridge for: Percentage of Insulation Defect Area R defect for: Relative temperature difference index ΔT norm for: in, A bridge This is the sum of the thermal bridge areas of all valid building facades of the same building object; A defect This is the sum of the areas of insulation defects on all valid building facades of the same building object; A facade It is the sum of the effective exterior facade areas of all buildings belonging to the same architectural object; Δ T ref For a predetermined temperature difference reference value, The area of ​​the exterior window area within the building facade is denoted as... A window The area of ​​the wall area inside the building facade is denoted as A wall The exterior window wall ratio of the building facade WWR for: Based on the window-to-wall ratio (WWR) of the building facade and the area ratio of thermal bridge zones R bridge Percentage of areas with insulation defects R defect and relative temperature difference index ΔT norm Calculating the building thermal risk index S : in, α , β , γ and δ The weighting coefficients are non-negative and sum to 1, according to the building thermal risk index. S Determine the priority of renovation for building objects from high to low.

7. The method for extracting key building parameter information based on multi-source image recognition as described in claim 1, characterized in that: Architectural spatial parameters include the building's length-to-width ratio, bay-to-depth ratio, and window-to-wall ratio on the exterior facade; The building outline is determined based on the building floor plan or roof plan, and the building aspect ratio is determined based on the building outline. The interior space areas are determined based on the building floor plan, and the span-to-depth ratio is determined based on the interior space areas; the boundaries of the window and wall areas are determined based on the adjacent edges of the exterior window areas and the wall areas.

8. The method for extracting key building parameter information based on multi-source image recognition as described in claim 1, characterized in that: Street view facade images and infrared thermal images correspond to the same building facade and are acquired simultaneously using visible light imaging units and infrared imaging units, or sequentially while keeping the acquisition position and shooting direction unchanged; various images and their output parameters are associated through building object numbers.

9. The method for extracting key building parameter information based on multi-source image recognition as described in claim 1, characterized in that: A positional transformation relationship is established by comparing corresponding points or edges at the same actual location in the street view elevation image and the infrared thermal image, and the positional transformation results are checked for errors. Images that fail the positional verification are not used for calculating the thermal parameters of the building envelope. Pixel-by-pixel temperature data from the infrared thermal image are read to form a temperature matrix. T (x, y) and extract candidate regions for thermal anomalies based on local temperature difference, temperature gradient, anomaly area and regional continuity.

10. The method for extracting key building parameter information based on multi-source image recognition as described in claim 1, characterized in that: Building spatial parameters, thermal bridge area ratio, insulation defect area ratio, and relative temperature difference index are selected from the structured dataset of key building parameter information as building classification parameters. The building classification parameters are processed to a unified dimension, and the building objects are clustered according to the processed building classification parameters to determine each building category and its category parameter center. Within each building category, the building object with the smallest overall deviation from the corresponding category parameter center and with all building classification parameters being valid is identified as the typical building sample.

Images