A green energy-saving building construction quality evaluation system and method

Abstract

The present application relates to the technical field of building energy-saving detection and quality evaluation, in particular to a green energy-saving building construction quality evaluation system and method, the present application synchronously obtains infrared thermograph, visible light image and three-dimensional point cloud data through building envelope structure, automatically identifies thermal defect candidate area through multi-modal fusion and image segmentation and maps to point cloud to obtain spatial coordinates, extracts defect area and actual thickness of thermal insulation layer from point cloud, compares with building information model in design stage to obtain design thermal resistance, calculates actual thermal resistance, calculates annual equivalent heat loss per unit area based on defect area, thermal resistance difference value and local heating / cooling degree days, further obtains annual equivalent carbon emission increment, finally determines green energy-saving construction quality grade deduction value according to preset threshold mapping table according to the percentage of the sum of carbon emission increment of each defect to building benchmark carbon emission, outputs evaluation result, realizes carbon emission driven energy-saving construction quality automatic rating.

Description

Technical Field

[0001] This invention relates to the field of building energy conservation testing and quality evaluation technology, specifically to a green energy-saving building construction quality evaluation system and method. Background Technology

[0002] Against the backdrop of the rapid development of green and energy-saving buildings, the thermal performance of the building envelope is directly related to the effectiveness of building energy consumption and carbon emission control. However, current methods for detecting and evaluating the quality of insulation layer defects during the construction phase still have significant shortcomings. On the one hand, traditional methods mainly rely on manual inspections or handheld thermal imagers for sampling, which is not only inefficient and has limited coverage, but also makes it difficult to obtain key quantitative parameters such as the accurate three-dimensional spatial location, actual area, and insulation layer thickness deviation of the defects. On the other hand, existing detection methods usually only focus on single-point temperature anomalies or apparent cracks, lacking the ability to integrate and analyze infrared thermal imaging temperature anomalies, visible light image texture features, and three-dimensional point cloud spatial data. This results in low identification rates and ambiguous positioning of thermal defects, and makes it impossible to accurately compare defect areas with the Building Information Model (BIM) of the design phase. In addition, current quality evaluation methods are mostly based on qualitative descriptions or simple deductions, without establishing a quantitative conversion model from thermal defect parameters to annual equivalent heat loss and carbon emission increments. It is also difficult to combine local climate characteristics (heating / cooling degree-days) to comprehensively evaluate the impact of defects on the long-term energy consumption and green rating of buildings. To solve this problem, we provide a green and energy-saving building construction quality evaluation system and method. Summary of the Invention

[0003] The purpose of this invention is to provide a green and energy-saving building construction quality evaluation system and method to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, a green and energy-saving building construction quality evaluation system is provided, comprising: The multi-mode data acquisition unit scans the building envelope, simultaneously acquiring infrared thermal images, visible light images, and 3D point cloud data; The defect region identification and localization unit performs multimodal fusion of the infrared thermal image and the visible light image, and uses an image segmentation algorithm to automatically identify thermal defect candidate regions from the infrared thermal image by combining temperature anomaly region clustering and temperature gradient abrupt change boundary extraction. The thermal defect candidate regions are then mapped to the three-dimensional point cloud data to determine the spatial coordinates of each thermal defect region. The defect thermal parameter extraction unit extracts the defect area and the actual thickness of the insulation layer at the defect location from the three-dimensional point cloud data for each thermal defect area. It compares the spatial coordinates of the thermal defect area with the building information model in the design stage to obtain the design thermal resistance value of the building envelope corresponding to the defect location. It also calculates the actual thermal resistance value of the defect location based on the deviation between the actual thickness of the insulation layer and the design thickness. The equivalent heat loss calculation unit calculates the annual equivalent heat loss per unit area of ​​the thermal defect area based on the defect area, the difference between the design thermal resistance value and the actual thermal resistance value of the building envelope, and the annual heating days and cooling days of the building location. The annual equivalent heat loss per unit area is multiplied by the defect area to obtain the total annual equivalent heat loss of the thermal defect area, and then multiplied by the carbon emission factor per unit energy consumption to obtain the annual equivalent carbon emission increment corresponding to the thermal defect area. The energy-saving construction quality rating unit determines the green energy-saving construction quality level deduction value corresponding to the thermal defects of the building envelope based on the percentage of the sum of the annual equivalent carbon emission increments of each thermal defect area to the building's annual baseline carbon emissions, according to a preset percentage threshold mapping table, and outputs the energy-saving construction quality evaluation result.

[0005] The second objective of this invention is to provide a method for implementing a green and energy-saving building construction quality evaluation system as described in any one of the above-mentioned claims, comprising the following steps: S1. Scan the building's external envelope and simultaneously acquire infrared thermal images, visible light images, and 3D point cloud data; S2. The infrared thermal image and the visible light image are fused in a multimodal manner. An image segmentation algorithm is used to automatically identify thermal defect candidate regions by combining temperature anomaly region clustering and temperature gradient abrupt change boundary extraction. The thermal defect candidate regions are then mapped to the three-dimensional point cloud data to determine the spatial coordinates of each thermal defect region. S3. For each thermal defect area, extract the defect area and the actual thickness of the insulation layer at the defect location from the three-dimensional point cloud data, compare the spatial coordinate position with the building information model in the design stage, obtain the corresponding design thermal resistance value of the building envelope, and calculate the actual thermal resistance value of the defect location based on the deviation between the actual thickness of the insulation layer and the design thickness. S4. Based on the defect area, the difference between the designed thermal resistance value and the actual thermal resistance value, and combined with the annual heating days and annual cooling days of the building location, calculate the annual equivalent heat loss per unit area of ​​the thermal defect area, and then calculate the total annual equivalent heat loss and the corresponding annual equivalent carbon emission increment. Finally, based on the percentage of the sum of the annual equivalent carbon emission increments of each defect area to the building's annual baseline carbon emissions, and according to the preset percentage threshold mapping table, determine the green energy-saving construction quality level deduction value, and output the evaluation result.

[0006] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention utilizes a drone equipped with an infrared thermal imager, a visible light camera, and a lidar to simultaneously acquire infrared thermal images, visible light images, and 3D point cloud data of the building envelope. It then employs multimodal fusion and temperature gradient enhancement image segmentation algorithms to automatically identify candidate areas for thermal defects. This solves the technical problems of low efficiency and incomplete coverage in traditional manual inspections, achieving accurate identification and 3D spatial positioning of insulation layer defects. By mapping defect areas to the building information model (BIM) during the design phase, and extracting the defect area and actual insulation layer thickness based on the point cloud, the difference between the actual thermal resistance and the design thermal resistance is calculated. Combined with the local heating and cooling degree-days, the annual equivalent heat loss per unit area and the increase in carbon emissions are quantified. This addresses the technical challenges of existing evaluation methods' qualitative description and lack of quantitative correlation between long-term energy consumption and carbon emissions. It achieves a refined technical effect of driving carbon emission increments in construction quality, providing a scientific basis for green building energy-saving acceptance and performance management. Attached Figure Description

[0007] Figure 1 This is an overall block diagram of the present invention; Figure 2 This is the overall flowchart of the present invention.

[0008] The meanings of the labels in the diagram are as follows: 1. Multi-mode data acquisition unit; 2. Defect area identification and location unit; 3. Defect thermal parameter extraction unit; 4. Equivalent heat loss calculation unit; 5. Energy-saving construction quality rating unit. Detailed Implementation

[0009] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0010] This invention provides a green and energy-saving building construction quality evaluation system. Please refer to [link / reference]. Figure 1As shown, it includes: Multi-mode data acquisition unit 1 scans the building envelope structure and simultaneously acquires infrared thermal images, visible light images, and three-dimensional point cloud data; The defect region identification and localization unit 2 performs multimodal fusion of the infrared thermal image and the visible light image, and uses an image segmentation algorithm to automatically identify thermal defect candidate regions from the infrared thermal image by combining temperature anomaly region clustering and temperature gradient abrupt change boundary extraction. The thermal defect candidate regions are then mapped to the three-dimensional point cloud data to determine the spatial coordinates of each thermal defect region. For each thermal defect area, the defect thermal parameter extraction unit 3 extracts the defect area and the actual thickness of the insulation layer at the defect location from the three-dimensional point cloud data, compares the spatial coordinates of the thermal defect area with the building information model in the design stage, obtains the design thermal resistance value of the enclosure structure corresponding to the defect location, and calculates the actual thermal resistance value of the defect location based on the deviation between the actual thickness of the insulation layer and the design thickness. The equivalent heat loss calculation unit 4 calculates the annual equivalent heat loss per unit area of ​​the thermal defect area based on the defect area, the difference between the design thermal resistance value of the building envelope and the actual thermal resistance value, and in combination with the annual heating days and annual cooling days of the building location. The annual equivalent heat loss per unit area is multiplied by the defect area to obtain the total annual equivalent heat loss of the thermal defect area, and then multiplied by the carbon emission factor per unit energy consumption to obtain the annual equivalent carbon emission increment corresponding to the thermal defect area. Energy-saving construction quality rating unit 5 determines the green energy-saving construction quality level deduction value corresponding to the thermal defects of the building envelope based on the percentage of the sum of the annual equivalent carbon emission increments of each thermal defect area to the building's annual baseline carbon emissions, according to a preset percentage threshold mapping table, and outputs the energy-saving construction quality evaluation result.

[0011] Preferably, the multi-mode data acquisition unit 1 takes the lead in carrying out the synchronous acquisition of multi-source data of the building's external envelope. This step relies on a drone equipped with various types of sensing devices. Through a preset flight path and synchronous triggering mechanism, it realizes the integrated and synchronized acquisition of infrared thermal images, visible light images, three-dimensional point cloud data and environmental auxiliary data. This provides spatiotemporally aligned and dimensionally complete basic data for subsequent thermal defect identification and parameter extraction, avoiding detection errors caused by the limitations of single-mode data and spatiotemporal misalignment, and ensuring the comprehensiveness, accuracy and synchronization of data acquisition.

[0012] In the specific implementation process, the first step is to plan and set the parameters of the drone's preset path. The planning scope covers all external envelope areas of the building, including all exterior walls, roofs, and the areas around door and window openings. The flight altitude is set to 10 to 15 meters above the surface of the envelope to avoid the risk of collision if the distance is too close and the data accuracy if the distance is too far. The flight speed is set to 2 to 3 meters per second, and the scanning interval is set to 0.5 meters. The flight path adopts a serpentine back-and-forth mode, starting from one side of the building wall and scanning line by line to the other side of the wall. After completing the wall scanning, it switches to the roof to ensure that there are no omissions or repeated scanning areas. After the path planning is completed, it is imported into the drone flight control system as the reference trajectory for autonomous flight. Subsequently, the hardware installation and synchronous trigger configuration of the sensing equipment were completed. The infrared thermal imager, visible light camera, and LiDAR were fixed to the bottom of the drone using a shock-absorbing gimbal. The gimbal was kept horizontal to ensure that the scanning lenses of the three types of sensors were all perpendicular to the surface of the building envelope, avoiding data distortion caused by angular deviation. A fixed distance was maintained between the sensors to prevent lenses from blocking each other or causing signal interference. At the same time, the signal interfaces of the three types of sensors were connected to the synchronous trigger controller. The synchronous trigger controller is an independent control module used to uniformly send acquisition trigger signals to achieve synchronous start and stop of the three types of sensors, eliminating timing deviations. Next, the drone was launched to perform flight scanning operations. The drone took off from the preset takeoff point and calibrated its attitude, and then cruised autonomously according to the preset path. During the flight, the synchronous trigger controller continuously sent trigger signals at a fixed sampling frequency of 10 Hz. Each signal trigger synchronously activated the infrared thermal imager, visible light camera, and LiDAR, enabling the three types of sensors to collect data synchronously at the same time. Each trigger corresponds to a set of multi-mode data bound with the same timestamp, ensuring complete timing consistency. The lidar emits a high-frequency laser beam under synchronous triggering. After scanning the surface of the building envelope, the laser beam reflects back to the radar receiver. The radar calculates the time difference between laser emission and echo to accurately calculate the three-dimensional spatial coordinates of each scan point, while simultaneously recording the laser reflection intensity to generate three-dimensional point cloud data. Each data point contains spatial coordinate values ​​in the X, Y, and Z dimensions and a reflection intensity value, completely restoring the three-dimensional shape of the envelope. The infrared thermal imager captures the infrared radiation energy on the surface of the envelope under synchronous triggering, converting the radiation energy into corresponding temperature values ​​to generate an infrared thermal image. Each pixel in the image corresponds to a temperature value, and all pixel temperature values ​​form a temperature matrix, accurately presenting the temperature distribution on the surface of the envelope and providing a basis for identifying temperature anomaly areas. The visible light camera collects the visible light reflection signal on the surface of the envelope under synchronous triggering, generating a visible light image. The image contains three independent RGB channels of data (red, green, and blue), clearly recording the surface texture, color, cracks, and apparent defects of the envelope, providing a visual reference for locating thermal defect areas.Simultaneously, a unified spatial reference is used for all collected data. The local coordinate system of the building is used as the benchmark, with the lower left corner of the building's bottom surface set as the origin. The X-axis extends horizontally along the wall, the Y-axis points vertically outward from the wall, and the Z-axis points vertically upward. The UAV flight control system records its own GPS coordinates, flight attitude angles, and sensor intrinsic parameters in real time. Through coordinate transformation algorithms, the pixel coordinates of the infrared thermal image, the pixel coordinates of the visible light image, and the spatial coordinates of the 3D point cloud are uniformly mapped to this local coordinate system. This ensures that the three types of data correspond one-to-one in spatial location, achieving spatial alignment of multimodal data and avoiding defective mapping deviations caused by spatial misalignment. Throughout the entire process of UAV scanning and collecting multi-mode data, environmental sensors mounted on the UAV simultaneously collect auxiliary data. An ambient temperature sensor records the air temperature in real time at the moment of collection, and a wind speed sensor records the horizontal airflow speed in real time. This auxiliary data is bound and stored with the corresponding infrared thermal images, visible light images, and 3D point cloud data, forming a complete dataset containing timestamps, spatial coordinates, multi-mode data, and auxiliary data. Subsequently, in the extraction of defect thermal parameters and calculation of equivalent heat loss, the ambient temperature and wind speed data are used to correct for the influence of environmental factors on the surface temperature and thermal resistance of the building envelope, effectively improving the accuracy of the detection results and reducing errors caused by environmental interference. After data collection is complete, the UAV returns to the preset landing point, and the collected multi-mode data and auxiliary data are automatically transmitted to the system storage module, completing the entire multi-mode data acquisition process. This process, through autonomous UAV scanning, synchronous sensor triggering, and unified spatiotemporal alignment, comprehensively acquires multi-dimensional raw data of the building envelope, providing a complete and reliable data foundation for subsequent defect identification, parameter extraction, heat loss calculation, and quality rating, ensuring the accuracy and comprehensiveness of the green and energy-saving building construction quality evaluation.

[0013] Preferably, in specific implementation, the acquired infrared thermal image and visible light image are first preprocessed. Preprocessing provides standardized input for subsequent feature extraction. The preprocessing steps include image denoising, size normalization, and channel normalization. Image denoising uses a Gaussian filtering algorithm to eliminate environmental interference and sensor noise during acquisition, avoiding noise affecting the accuracy of feature extraction. Size normalization adjusts the infrared thermal image and visible light image to the same resolution, for example, to 512 pixels by 512 pixels, ensuring that the pixel dimensions of the two types of images are completely matched, providing a basis for subsequent feature alignment and stitching. Channel normalization maps the temperature matrix values ​​of the infrared thermal image and the RGB channel values ​​of the visible light image to the range of 0 to 1, eliminating the difference in the numerical range of the two types of images and ensuring the stability of the deep learning network training and inference process. After preprocessing, the standardized infrared image and standardized visible light image are output and input into two independent branches of the deep learning network, respectively.

[0014] The deep learning network employs a dual-branch parallel architecture, with both branches maintaining consistent structural parameters and independent functions. The first branch specifically processes standardized infrared thermal images, while the second branch processes standardized visible light images. Each branch consists of consecutive convolutional layers, pooling layers, and activation function layers. Convolutional layers extract deep local features from the image, pooling layers compress feature dimensions and retain key features, and activation function layers introduce nonlinear transformations to enhance feature representation. After receiving the standardized infrared thermal image, the first branch extracts core information such as temperature distribution gradients and temperature abrupt change regions through multi-layer convolutional operations, generating a temperature feature map. High-response areas in the temperature feature map correspond to temperature anomaly areas in the infrared thermal image, accurately reflecting the potential locations of thermal anomalies on the surface of the building envelope. After receiving the standardized visible light image, the second branch extracts core information such as surface texture, crack edges, and component contours through multi-layer convolutional operations, generating a texture feature map. High-response areas in the texture feature map correspond to apparent edges and texture abrupt change locations in the visible light image, clearly presenting the surface details of the building envelope. The temperature feature map and texture feature map output by the two branches maintain the same size and number of channels, ensuring smooth execution of subsequent weighting and stitching operations.

[0015] After extracting the two types of feature maps, an attention mechanism is introduced to calculate feature weights. The core function of the attention mechanism is to automatically identify regions in the temperature and texture feature maps that are more important for thermal defect identification and assign them higher weights, avoiding interference from invalid features. In specific implementation, global feature pooling is first performed on the temperature and texture feature maps respectively to extract the global feature vector of each feature map. The global feature vector condenses the overall information of a single feature map. Then, the two global feature vectors are concatenated and input into a fully connected layer to calculate the importance score of each channel. The importance score is the initial feature weight. Subsequently, the initial feature weights are mapped back to each spatial position of the temperature and texture feature maps to generate a spatial dimension feature weight map. The higher the value in the weight map, the more important the feature at the corresponding position is for defect identification. Finally, the temperature feature map is multiplied pixel by pixel with its corresponding weight map to obtain a weighted temperature feature map, and the texture feature map is multiplied pixel by pixel with its corresponding weight map to obtain a weighted texture feature map. After the weighting operation is completed, important features are strengthened and secondary features are weakened, effectively improving the recognition of key features.

[0016] The weighted temperature feature map and the weighted texture feature map are concatenated along the channel dimension. Channel dimension concatenation refers to merging the two types of feature information along the channel direction of the feature map. The resulting concatenated feature map has the same number of channels as the temperature feature map and the texture feature map. This feature map completely preserves two types of core information: temperature anomaly information from the weighted temperature feature map and visual edge information from the weighted texture feature map. It includes both appearance details that infrared images cannot show and thermal anomaly signals that visible light images cannot capture, thus achieving complementarity between the two types of information. For example, in the case of cracks in the insulation layer of a building envelope, infrared images show temperature anomalies, while visible light images show crack edges. The concatenated feature map can highlight both types of features simultaneously, providing dual evidence for defect identification. After generating the fused feature map, it is directly used as input for subsequent image segmentation algorithms, enhancing the recognition effect of thermal defect candidate regions. Compared with single-modal feature maps, the fused feature map has more prominent defect features, clearer boundaries, and stronger anti-interference ability, which can effectively reduce the probability of missed and false detections of thermal defects. It provides a high-quality feature foundation for subsequent defect localization and thermal parameter extraction. The entire multimodal fusion process is automatically executed by a deep learning network without manual intervention, realizing the automation and intelligence of feature extraction and fusion, and ensuring the efficiency and accuracy of the defect recognition process.

[0017] Preferably, firstly, the image segmentation algorithm is an image processing algorithm that divides image pixels into target regions and background regions to achieve accurate extraction of the target region. The encoder-decoder structure is a classic symmetric network architecture in the field of image segmentation. The encoder is responsible for feature extraction and dimensionality compression, while the decoder is responsible for feature restoration and resolution recovery. The convolutional layer is a network layer that extracts deep semantic features of the image through local convolution operations. Downsampling is an operation that reduces the size of the feature map and condenses key feature information through pooling or stride convolution. The decoder is a network branch symmetrical to the encoder, responsible for feature decoding and resolution recovery. The deconvolutional layer is a network layer that expands the size of the feature map and restores image details through transposed convolution. Upsampling is an operation that restores the original resolution of the feature map and restores the spatial details of the image. Decoded features are the outputs of each layer of the decoder. The features map is processed by upsampling. Skip connections are network connections that connect the shallow layers of the encoder to the corresponding layers of the decoder and pass on the shallow layer details. The temperature gradient calculation module is a functional module specifically for infrared thermal images, which calculates the temperature change gradient and extracts abrupt change edges. Gradient operation is a mathematical operation that calculates the temperature change amplitude of adjacent pixels and identifies temperature abrupt changes. Temperature abrupt change edges are pixel boundaries in infrared thermal images where the temperature value changes rapidly. The temperature abrupt change edge map is a binary image that marks all temperature abrupt change edge pixels. Overlay is an operation that fuses the temperature abrupt change edge information with the decoding features at the pixel level. The binary segmentation map is an image containing only two types of pixels, representing thermal defect candidate regions and the background, respectively. Thermal defect candidate regions are pixel regions identified by the segmentation algorithm that are suspected of having thermal anomalies.

[0018] The first step in the algorithm execution is input data preparation. The generated fused feature map is used as the initial input of the encoder, while the original infrared thermal image is completely preserved and synchronously input to the temperature gradient calculation module to ensure that feature extraction and edge calculation use the same source data and avoid segmentation errors caused by data misalignment.

[0019] The second step is to perform encoder downsampling feature extraction. The encoder consists of multiple layers of convolutional layers and downsampling layers stacked alternately. Each convolutional layer performs local convolution operations on the input feature map, gradually extracting deep semantic features from the fused feature map, including the overall distribution of temperature anomaly regions and deep patterns of texture anomalies. After each convolutional operation, the feature map size is reduced by the downsampling layer. While reducing the size, key defect features are condensed and invalid background information is filtered out. The encoder has a total of 4 convolutional downsampling layers. The feature map output of each layer is cached and stored separately to prepare for the subsequent skip connections to pass detailed features. After multiple downsampling layers, the feature map size is reduced to one-sixteenth of the original fused feature map, completing the condensation and extraction of deep defect features.

[0020] The third step is to perform decoder upsampling resolution restoration. The decoder and encoder have a symmetrical structure, consisting of multiple layers of deconvolutional and convolutional layers stacked alternately. The initial input is the deep condensed feature map output by the last layer of the encoder. Each deconvolutional layer performs upsampling operation on the input feature map, gradually increasing the feature map size. After each upsampling, the features are refined by convolutional layers, gradually restoring the spatial details of the fused feature map. After 4 layers of deconvolutional upsampling processing, the feature map size is restored to be completely consistent with the original fused feature map. At this time, the features output by the decoder are the basic decoded features. These features have complete deep defect semantics, but the edge detail information is weak and the temperature change boundary recognition is insufficient.

[0021] The fourth step initiates the collaborative processing of the skip connection and temperature gradient calculation modules. Skip connection paths are constructed between the shallow features cached in each layer of the encoder and the basic decoding features of the corresponding layer of the decoder. At the same time, an independent temperature gradient calculation module is inserted into each skip connection path. The temperature gradient calculation module receives the original infrared thermal image and performs gradient calculations on the image pixel by pixel. Specifically, it calculates the temperature difference between each pixel and its neighboring pixels, identifies pixels whose temperature difference exceeds a preset threshold, and these pixels are the temperature change edge pixels. All temperature change edge pixels are marked as valid edges, and the remaining pixels are marked as invalid background, generating a temperature change edge map. This edge map accurately presents all boundaries of rapid temperature changes in the infrared thermal image, including the edge contours of thermal defects such as insulation layer cracks and hollow areas.

[0022] The fifth step is to perform the overlay operation of the temperature change edge map and the decoding feature. The temperature change edge map is overlaid with the current layer's basic decoding feature at the pixel level. During the overlay, the deep semantic information of the basic decoding feature is preserved, while the feature response at the temperature change edge position is enhanced, making the originally blurry defect boundary clear and sharp. This effectively makes up for the problem of missing edge details in the basic decoding feature. After the overlay is completed, a new decoding feature with enhanced edge details is output. This feature contains both deep defect semantics and accurate temperature change edge information.

[0023] The sixth step completes the decoding and overlay loop layer by layer. Each layer of the decoder repeats the process of deconvolution upsampling, skip connections, temperature gradient module operation, and edge overlay until all layers are processed. This ensures that the decoded features of each layer are enhanced by the temperature change edge information, and finally outputs a complete decoded feature map with the same resolution as the original fused feature map and clear edges.

[0024] The seventh step generates a binary segmentation map. A final convolutional operation and binarization activation are performed on the final, fully decoded feature map, mapping each pixel in the feature map to two types of values: value 1 represents a thermal defect candidate region, and value 0 represents the background region. The final output is a standard binary segmentation map. This image accurately divides the thermal defect candidate region and the background at the pixel level, with the defect region boundary closely matching the edge of the infrared temperature change, effectively avoiding the edge blurring problem of traditional segmentation algorithms. The generated binary segmentation map is directly used as the defect region identification result and input into the mapping stage of the subsequent defect region identification and localization unit 2. This provides pixel-level boundary information for the accurate mapping of the thermal defect candidate region to the 3D point cloud data. The entire segmentation algorithm is fully automated, relying on the encoder-decoder architecture to achieve deep feature extraction and resolution restoration. The temperature gradient calculation module enhances the details of the temperature change edge, accurately completing the pixel-level segmentation of the thermal defect candidate region and ensuring the accuracy of subsequent defect localization and thermal parameter extraction.

[0025] Preferably, the camera calibration parameters are a set of parameters describing the internal and external geometric relationships of the imaging camera, including intrinsic parameters (focal length, principal point, distortion coefficient) and extrinsic parameters (position and attitude of the camera in the lidar coordinate system), used to establish the mapping relationship between image pixels and three-dimensional space; pixel coordinates are the row and column position identifiers of each pixel in the binary segmentation image on the image plane; the lidar coordinate system is a reference coordinate system with the lidar as the origin, defining the three-dimensional spatial position; the point cloud index is the unique identifier number of each three-dimensional point in the three-dimensional point cloud, used to quickly locate the corresponding point; the defect point cloud cluster is the set of three-dimensional points corresponding to the thermal defect candidate region; the convex hull boundary is the smallest convex polygon (a convex polyhedron in three dimensions) that encloses all three-dimensional points of the point cloud cluster; the spatial coordinate position of the thermal defect region is the reference positioning point of the defect region in three-dimensional space; the boundary point set is the set of all three-dimensional points that constitute the convex hull boundary, used for subsequent area calculation.

[0026] The first step in the process is data preparation. The defect area identification and localization unit 2 simultaneously retrieves three types of core data: a binary segmentation map containing candidate thermal defect areas, complete 3D point cloud data of the building envelope structure collected by UAV, and camera calibration parameters pre-calibrated during the multi-mode data acquisition stage. All three types of data are spatiotemporally aligned to ensure that the binary segmentation map, 3D point cloud, and camera parameters belong to the same scanning time period and the same building area, avoiding data misalignment that could lead to mapping failure. At the same time, the binary segmentation map is preprocessed to extract the set of pixels marked as candidate thermal defect areas, record the row and column pixel coordinates of each pixel, and form a defect pixel coordinate list to provide basic data for coordinate transformation.

[0027] The second step in the process is to transform the pixel coordinates in the binary segmentation image to the LiDAR coordinate system using camera calibration parameters. This step is the core of the 2D to 3D mapping. The camera calibration parameters are obtained in advance using the standard checkerboard calibration method. The intrinsic parameters are used to correct image distortion and establish the relationship between pixels and the camera imaging plane, while the extrinsic parameters are used to establish the spatial transformation relationship between the camera coordinate system and the LiDAR coordinate system. During the transformation, the image distortion of each defect pixel coordinate is first eliminated using the camera intrinsic parameters to obtain the corrected normalized image coordinates. Then, combined with the rotation matrix and translation vector of the camera extrinsic parameters, the normalized image coordinates are transformed into a 3D spatial ray in the LiDAR coordinate system. This ray represents the viewing direction of the pixel in 3D space. Each defect pixel corresponds to a unique spatial ray, completing the transformation from pixel coordinates to 3D rays and providing spatial orientation for point cloud matching.

[0028] The third step of the process is to find the corresponding 3D points through point cloud indexing to form a defect point cloud cluster. Each 3D point in the 3D point cloud data stores a unique index number, 3D coordinates, and reflection intensity. The point cloud index is a fast retrieval mechanism that can match the corresponding 3D points according to spatial location. In specific execution, for the spatial ray corresponding to each defect pixel, the 3D point cloud data is traversed to find 3D points located within a preset distance threshold near the ray. The distance threshold is set to 0.1 meters to ensure matching accuracy. All matched 3D points are summarized, and duplicate points and outliers (outliers refer to points that are more than the threshold away from the ray and significantly deviate from the defect area) are removed. Finally, a continuous and complete defect point cloud cluster is formed. This point cloud cluster accurately corresponds to the thermal defect candidate area in the binary segmentation map, realizing the mapping from the 2D defect area to the 3D point set.

[0029] The fourth step in the process is to calculate the convex hull boundary of the defect point cloud cluster. The convex hull boundary is the smallest closed convex contour that can completely enclose all three-dimensional points of the defect point cloud cluster. During the calculation, the defect point cloud cluster is first projected onto the plane of the building's external envelope (e.g., the wall is a vertical plane and the roof is a horizontal plane), simplifying the three-dimensional points into two-dimensional plane points. Then, the planar convex hull algorithm is used to connect the outermost points in sequence to form a closed convex polygon. This convex polygon is the convex hull boundary of the defect point cloud cluster. The convex hull boundary can completely surround all valid points in the defect area, eliminate interference from scattered internal points, and accurately define the outer contour range of the defect.

[0030] The fifth step of the process is to use the center point of the convex hull boundary as the spatial coordinate position of the thermal defect area. The center point is calculated using the geometric center method. First, the two-dimensional coordinates of all vertices of the convex hull boundary are extracted. The average value of the x-coordinate and the average value of the y-coordinate of all vertices are calculated to obtain the geometric center point of the convex hull. Then, the center point is mapped back to three-dimensional space to obtain the corresponding three-dimensional coordinates. These three-dimensional coordinates are the reference spatial coordinate positions of the thermal defect area, which are used to accurately identify the specific location of the defect on the building envelope and provide a precise positioning basis for subsequent comparison with the building information model.

[0031] The sixth step in the process is to store the convex hull boundary point set. All the three-dimensional vertices that constitute the convex hull boundary are organized into point set data in sequence. Each point contains complete three-dimensional coordinates. This point set completely records the three-dimensional boundary information of the outer contour of the defect and can be directly used for subsequent defect area calculation without reprocessing the point cloud data, thus improving the efficiency of subsequent parameter extraction. At the same time, the defect point cloud cluster, convex hull boundary, center point coordinates, boundary point set, corresponding binary segmentation map, and scanning time period information are bound and stored to form a complete defect location data archive, ensuring that the data is traceable and reusable. The entire mapping and positioning process starts from the two-dimensional defect pixels, and through coordinate transformation, point cloud matching, and convex hull calculation, finally obtains the three-dimensional defect location and boundary. The entire process is data-driven and automated, accurately realizing the transformation of thermal defects from image recognition to three-dimensional spatial positioning. This lays a solid foundation for the subsequent extraction of thermal parameters such as defect area and insulation layer thickness, ensuring the spatial accuracy of thermal defect detection in green and energy-saving buildings.

[0032] Preferably, the defect thermal parameter extraction unit 3 is the core processing unit in the green energy-saving building construction quality evaluation system responsible for extracting key physical parameters of defects from three-dimensional point cloud data and quantifying the degree of thermal defects. The three-dimensional point cloud data is a discrete three-dimensional point set collected by lidar and containing spatial location information of the building envelope structure. The defect point cloud cluster is a set of three-dimensional points that accurately corresponds to the thermal defect area and is obtained by screening through the defect mapping step. The best fitting plane is a two-dimensional plane generated by mathematical fitting that best fits the spatial distribution of the defect point cloud cluster and minimizes the sum of the distances from all points to the plane. The projected polygon is a closed two-dimensional polygon formed by vertically projecting all three-dimensional points of the defect point cloud cluster onto the best fitting plane. The polygon and defect area are the areas of the planar regions enclosed by the projected polygons. The point cloud of the insulation layer surface is the set of three-dimensional points in the defect area corresponding to the outer surface of the insulation layer and the side away from the building interior. The point cloud of the base layer surface is the set of three-dimensional points in the defect area corresponding to the base layer of the building below the insulation layer and the side close to the building interior. The normal direction is the spatial direction perpendicular to the best-fit plane and pointing out of the building. The average distance is the arithmetic mean of the distances along the normal direction between the point cloud of the insulation layer surface and the point cloud of the base layer surface. The design thickness is the standard thickness of the insulation layer specified in the building construction design drawings. The thickness deviation is the difference between the actual thickness of the insulation layer and the design thickness, which is used to directly quantify the degree of construction deviation of the insulation layer.

[0033] In the defect area extraction stage, the basic data retrieval is completed first. The defect thermal parameter extraction unit 3 simultaneously retrieves the defect point cloud cluster, original three-dimensional point cloud data and convex hull boundary data output by the defect area identification and positioning unit 2 to ensure that the data time sequence is consistent and the spatial range is matched. Then, the best fitting plane calculation is performed. For all three-dimensional points in the defect point cloud cluster, the least squares method is used to perform plane fitting. This fitting process iteratively adjusts the plane parameters to make the sum of the distances from each three-dimensional point in the defect point cloud cluster to the fitting plane reach the minimum value, and finally generates a unique best fitting plane. This plane has the highest degree of fit with the actual surface of the building envelope structure, which can effectively eliminate the projection error caused by the small curvature or concavity of the envelope structure, and provide an accurate reference plane for subsequent point cloud projection.

[0034] After generating the best-fit plane, a point cloud projection operation is performed. Each 3D point within the defect point cloud cluster is projected perpendicularly onto the best-fit plane along a direction perpendicular to it. Each 3D point generates a corresponding 2D projection point. All projection points form a discrete set of points on the plane. Then, a convex hull boundary calculation is performed on this discrete projection point set to generate a closed projection polygon. The boundary of this polygon perfectly matches the actual outer contour of the thermal defect, with no redundant or missing areas. Finally, the area of ​​the projection polygon is calculated. By statistically analyzing the coordinates of the vertices of the projection polygon boundary, the planar area of ​​the closed region is obtained using the polygon area calculation method. This area is the defect area, with the unit set to square meters. After the calculation is completed, the defect area value is bound and stored with the defect point cloud cluster and projection polygon data. It will then be directly used in the equivalent heat loss calculation stage. The larger the defect area, the wider the influence range of the thermal defect and the higher the potential heat loss.

[0035] In the actual thickness extraction stage of the insulation layer, the defect point cloud cluster is first subjected to planar segmentation. The segmentation benchmark is the generated best-fit plane. A distance threshold is set according to the conventional thickness range of the building envelope insulation layer. The default threshold value is 60% of the design thickness of the insulation layer. Using the best-fit plane as the benchmark, the defect point cloud cluster is divided into two parts. The set of three-dimensional points with a distance less than the threshold is defined as the base surface point cloud, and the set of three-dimensional points with a distance greater than the threshold is defined as the insulation layer surface point cloud. During the segmentation process, outliers with abnormal distances are automatically removed, that is, three-dimensional points that deviate significantly from the distribution of the two types of surfaces and exceed the reasonable distance range, so as to avoid outliers interfering with the thickness calculation results. Then, the normal direction is determined. Using the best-fit plane as the benchmark, the direction perpendicular to the plane and pointing outward of the building is taken as the normal direction. This direction is completely consistent with the actual physical direction of the insulation layer thickness measurement, ensuring no deviation in the thickness measurement direction. Next, the distance between the point cloud on the surface of the insulation layer and the point cloud on the surface of the base layer along the normal direction is calculated. The two sets of point clouds are traversed, and the distance from each three-dimensional point on the surface of the insulation layer to the corresponding three-dimensional point on the surface of the base layer along the normal direction is calculated point by point. All effective distance data are collected, and abnormal data with excessively large or small distances are eliminated. Finally, the arithmetic mean of the remaining effective distance data is calculated. This average value is the actual thickness of the insulation layer, with the unit set to millimeters. After the calculation is completed, the actual thickness value of the insulation layer is bound and stored with the two types of point cloud data and the normal direction data. It will be directly used in the subsequent calculation of the actual thermal resistance value of the defective parts. The actual thickness value directly reflects the true thickness of the insulation layer after construction and is the core parameter for evaluating the construction quality of the insulation layer.

[0036] In the thickness deviation calculation stage, the insulation layer design thickness data stored in the system database during the building design phase is first retrieved. The design thickness unit is consistent with the actual thickness. Then, the difference calculation is performed, subtracting the design thickness from the actual insulation layer thickness to obtain the thickness deviation. The sign of the thickness deviation directly indicates the direction of construction deviation. A positive deviation indicates that the actual insulation layer thickness is greater than the design thickness, while a negative deviation indicates that the actual insulation layer thickness is less than the design thickness. The larger the absolute value of the deviation, the higher the degree of construction deviation and the more serious the deviation of thermal performance from the design standard. After the calculation is completed, the thickness deviation value is compared with the defect area, The actual thickness data of the insulation layer is bound and stored, and then directly substituted into the calculation of the actual thermal resistance value. The thickness deviation is a key intermediate parameter connecting construction quality and thermal performance, providing a direct quantitative basis for construction quality rating. The entire process of extracting defect thermal parameters, from 3D point cloud to 2D area, from spatial distance to actual thickness, from comparison between actual and design to deviation quantification, is data-driven and logically closed-loop. All parameter calculation results clearly correspond to subsequent use stages, realizing the accurate quantification of thermal defects. This provides reliable and traceable quantitative data support for the evaluation of green and energy-saving building construction quality, ensuring the scientificity and accuracy of the evaluation results.

[0037] Preferably, the iterative nearest point algorithm is a three-dimensional registration algorithm used for spatial coordinate matching between two sets of three-dimensional point sets, solving for the optimal spatial transformation parameters through iterative optimization, and achieving precise alignment of the two sets of point sets. Component alignment is the operation of aligning the measured defect point cloud cluster with the corresponding enclosure structure component in the building information model in three-dimensional space. The design thermal resistance value is the standard thermal resistance parameter preset in the building information model when the thickness of the insulation layer of the enclosure structure meets the design standard. The thermal conductivity of the material is an inherent physical property of the insulation material and a constant parameter characterizing the ease with which heat is conducted inside the material. Thermal resistance is a physical quantity that hinders the transfer of heat through the enclosure structure. Its magnitude is directly proportional to the thickness of the insulation material and inversely proportional to the thermal conductivity of the material. The actual thermal resistance value is the real thermal resistance parameter calculated based on the measured thickness and the thermal conductivity of the material when there is a construction thickness deviation in the insulation layer of the enclosure structure.

[0038] The first step in the process is data retrieval and preprocessing. The defect thermal parameter extraction unit 3 simultaneously retrieves three types of core data: the spatial coordinates of the thermal defect area output by the defect area identification and positioning unit 2, the complete defect point cloud cluster, the building information model stored in the system database during the design phase, and the actual thickness and thickness deviation data of the insulation layer calculated above. The building information model has been pre-divided into the outer wall, roof, and door and window perimeter and other enclosure structural components. Each component is bound to independent three-dimensional geometric data, material property data, and thermal parameter data. After retrieval, the defect point cloud cluster is preprocessed to remove a small number of outlier points and retain complete and valid point cloud data. At the same time, the surface three-dimensional point set of all enclosure structural components in the building information model is extracted to provide two sets of points to be aligned for the subsequent iterative nearest point algorithm registration. The two sets of point sets are the measured defect point cloud cluster and the surface point set of the enclosure structural components in the building information model.

[0039] The second step employs an iterative nearest-point algorithm to achieve precise alignment between the defect point cloud cluster and the corresponding components in the Building Information Model (BIM). The alignment operation focuses on spatial feature matching between the two point sets. First, an initial spatial transformation estimation is performed. By calculating the geometric center points of the defect point cloud cluster and the geometric center points of the surface point sets of each building envelope component in the BIM, the building envelope component with the smallest center point distance and the highest spatial overlap is selected as the target component to be matched. Simultaneously, based on the coordinate difference between the two sets of center points, an initial translation transformation matrix is ​​generated to achieve preliminary alignment between the defect point cloud cluster and the target component. After preliminary alignment, slight positional deviations still exist between the two point sets, requiring fine-tuning through the iterative nearest-point algorithm. This fine-tuning process involves an iterative loop, with each iteration containing three core steps. The first step is nearest-point search, which traverses every 3D point in the defect point cloud cluster and searches for the nearest 3D point in the surface point set of the target component. The first step is to establish a one-to-one correspondence between two sets of points. The second step is to solve for the transformation parameters. Based on the corresponding point pairs obtained by matching, the optimal rotation transformation matrix and translation transformation matrix are calculated using the least squares method to minimize the mean square error between the two sets of points. The third step is to update the point set. The defect point cloud cluster is transformed in spatial coordinates according to the optimal transformation matrix obtained by solving, and the spatial position of the defect point cloud cluster is updated. This completes one round of iteration. The above three steps are repeated continuously during the iteration process. After each round of iteration, the mean square error of the two sets of points is calculated. When the mean square error value is less than the preset accuracy threshold (set to 0.001 meters), it is determined that the two sets of points have been accurately aligned, and the iteration loop is terminated. At this time, the defect point cloud cluster and the corresponding enclosure structure component in the building information model are completely superimposed in three-dimensional space, achieving accurate alignment of the components. The alignment result ensures the spatial matching of subsequent parameter readings and avoids errors in reading parameters across components.

[0040] The third step involves retrieving the design thermal resistance and material thermal conductivity of the components from the building information model (BIM). After precise alignment of the components, the defect thermal parameter extraction unit 3 locates the enclosure structure component in the BIM that matches the defect point cloud cluster, and retrieves the design thermal parameters bound to the component. The design thermal resistance is the standard thermal resistance value when the thickness of the component's insulation layer meets the design standard, and the material thermal conductivity is the inherent thermal conductivity value of the insulation material used in the component's insulation layer. Both parameters are fixed parameters preset in the design stage and stored in the component attribute column. The units corresponding to the parameters are recorded synchronously during the reading process to ensure consistency with the units of subsequent calculation parameters. The unit of the design thermal resistance value is square meters per degree Celsius per watt, and the unit of the material thermal conductivity is watts per meter per degree Celsius. After the reading is completed, the two parameters are bound and stored with the defect information to provide design benchmark parameters for the calculation of the actual thermal resistance value. The fourth step involves calculating the actual thermal resistance of the defective area based on the deviation between the actual and designed thickness of the insulation layer. This calculation utilizes the relationship that thermal resistance equals thickness divided by thermal conductivity. The core basis for this calculation is the physical relationship that thermal resistance is directly proportional to thickness and inversely proportional to the material's thermal conductivity. Given the two key parameters—the actual thickness of the insulation layer and the thermal conductivity of the insulation material—the actual thermal resistance of the defective area is obtained by directly dividing the actual thickness by the thermal conductivity. The unit of the actual thermal resistance value should be consistent with the design thermal resistance value. After calculating the actual thermal resistance value, the difference between the design and actual thermal resistance values ​​is further calculated. This difference directly reflects the degree of influence of the insulation layer construction deviation on the thermal resistance. A positive difference indicates that the actual thermal resistance is greater than the design thermal resistance, and the heat transfer resistance is greater; a negative difference indicates… The actual thermal resistance is less than the design thermal resistance, the heat transfer resistance is smaller, and the risk of heat loss is higher. This difference data will be directly input into the equivalent heat loss calculation unit 4 for the subsequent calculation of the annual equivalent heat loss per unit area. This realizes the transformation of measured thermal parameters into quantitative heat loss data. The entire process, from spatial registration to parameter reading and then to the calculation of actual thermal resistance value, relies on the precise correlation between measured data and design model. It eliminates spatial position deviation through the iterative nearest point algorithm and uses the thermal physical relationship of materials to complete the derivation of thermal resistance value. All calculation steps are clearly directed to the subsequent equivalent heat loss calculation stage, realizing the transformation of defective thermal parameters from geometric data to thermal performance data, and ensuring the accuracy, scientificity, and traceability of thermal defect detection and quality evaluation of green and energy-saving buildings.

[0041] Preferably, the defect area is the planar projected area of ​​the thermal defect region, characterizing the extent of the defect's influence; the design thermal resistance value of the building envelope is the standard thermal resistance when the insulation layer meets the design standards, serving as the benchmark value for heat transfer resistance; the actual thermal resistance value is the true thermal resistance when the insulation layer has thickness deviations, reflecting the actual heat transfer resistance; the annual heating degree days are the sum of cumulative temperature differences when the average daily temperature of the building's location during the heating season is lower than the benchmark temperature (usually 18 degrees Celsius), quantifying the degree of heating or cooling during the heating season; the annual cooling degree days are the sum of cumulative temperature differences when the average daily temperature of the building's location during the cooling season is higher than the benchmark temperature (usually 26 degrees Celsius), quantifying the degree of heat during the cooling season; and the annual equivalent heat loss per unit area. Heat loss is the equivalent heat loss per unit area caused by thermal defects throughout the year, characterizing the energy consumption impact per unit area of ​​the defect; additional heat flux is the extra heat transfer flux generated due to the actual thermal resistance being lower than the design thermal resistance; indoor-outdoor design temperature difference is the difference between the indoor comfort temperature specified in the building design phase and the outdoor reference temperature, serving as the reference temperature difference for heat transfer calculation; equivalent hours is the equivalent duration of heat transfer throughout the year obtained by converting the annual heating degree-days and annual cooling degree-days, realizing the conversion from degree-days to time units; annual equivalent carbon emission increment is the annual increase in carbon emissions corresponding to the extra energy consumption caused by thermal defects; unit energy consumption carbon emission factor is the carbon emission conversion coefficient corresponding to unit heat loss, which is a fixed industry standard value.

[0042] The first step in the process is basic data retrieval and preprocessing. The equivalent heat loss calculation unit 4 simultaneously retrieves multiple types of key data, including the defect area, the design thermal resistance value of the building envelope, and the actual thermal resistance value output by the defect thermal parameter extraction unit 3. It also retrieves the annual heating degree-days and annual cooling degree-days of the building's location from the system's climate database, and the indoor and outdoor design temperature difference and carbon emission factor per unit energy consumption from the building design code database. All data must undergo unified unit verification to ensure that the thermal resistance value is in square meters per degree Celsius per watt, the area is in square meters, the degree-days are in degrees Celsius per day, the temperature difference is in degrees Celsius, and the carbon emission factor is in kilograms per watt-hour. The preprocessing process removes abnormal climate data and invalid parameters to ensure the validity and consistency of the calculation input data, laying the foundation for subsequent accurate calculations.

[0043] The second step in the process is to calculate the additional heat flow per unit area caused by the reduction in thermal resistance. The core reason for this additional heat flow is that the actual thermal resistance is less than the design thermal resistance, resulting in insufficient heat transfer resistance in the building envelope and causing the heat transfer to exceed the design standard. The specific calculation logic is carried out in three steps. The first step is to calculate the difference between the reciprocal of the design thermal resistance and the reciprocal of the actual thermal resistance. The reciprocal of the thermal resistance represents the heat transfer coefficient, and the magnitude of the difference directly reflects the increase in heat transfer capacity caused by insufficient thermal resistance. The larger the difference, the more significant the improvement in heat transfer capacity and the greater the additional heat flow. The second step is to multiply this difference by the indoor and outdoor design temperature difference to obtain the instantaneous additional heat flow per unit area. The instantaneous additional heat flow represents the extra heat transfer per unit area per unit time due to the defect under the standard temperature difference. The heat transfer rate directly reflects the instantaneous heat transfer impact of the defect. The third step multiplies the instantaneous additional heat flow by the equivalent hours converted from the annual heating and cooling degree days to complete the conversion from instantaneous heat flow to annual cumulative heat loss. The conversion logic of the equivalent hours is to convert the annual heating degree days into heating equivalent hours based on 24 hours per day, and the annual cooling degree days into cooling equivalent hours based on 24 hours per day. The two types of equivalent hours are added together to obtain the annual equivalent hours. This conversion realizes the correlation between the climatological degree days and the heat transfer duration, and finally obtains the annual equivalent heat loss per unit area. This parameter accurately quantifies the additional heat loss generated by thermal defects per unit area throughout the year and is the core quantitative indicator of energy consumption impact.

[0044] The third step in the process is to calculate the total annual equivalent heat loss of the thermal defect area. This is achieved by multiplying the annual equivalent heat loss per unit area obtained in the second step by the defect area, resulting in the total additional heat loss generated throughout the year for the entire thermal defect area. A larger defect area and higher heat loss per unit area lead to a greater total annual heat loss. This parameter directly reflects the overall energy consumption impact of the thermal defect, providing a magnified basis for subsequent carbon emission calculations in the area dimension. After calculation, the total annual equivalent heat loss data is bound to the defect information for storage, ensuring data traceability and reusability. The fourth step in the process is to calculate the annual equivalent carbon emission increase corresponding to the thermal defect area. The amount is calculated by multiplying the total annual equivalent heat loss by the carbon emission factor per unit energy consumption. This yields the annual increase in carbon emissions corresponding to the additional energy consumption caused by thermal defects. The carbon emission factor per unit energy consumption is a fixed conversion factor published by the state or industry. Its value is based on the carbon emission intensity measurement of conventional energy combustion. Different energy types correspond to different factors. In the building sector, the carbon emission factor corresponding to electricity or natural gas is used by default. This conversion achieves a direct transformation from heat loss energy consumption to carbon emission impact. The larger the annual equivalent carbon emission increase, the more serious the negative impact of thermal defects on building carbon emissions. It is a key negative indicator for energy-saving construction quality rating.

[0045] After completing all calculations, the equivalent heat loss calculation unit 4 transmits three core data points—annual equivalent heat loss per unit area, total annual equivalent heat loss, and annual equivalent carbon emission increment—to the energy-saving construction quality rating unit 5. This provides accurate quantitative data for subsequent calculations of carbon emission ratios, matching quality grade thresholds, and determining construction quality deduction values. The entire calculation process strictly follows the principles of building thermal heat transfer and climate parameter conversion logic. It is data-driven and automated throughout, without human intervention, avoiding human calculation errors. This achieves accurate quantification of the impact of thermal defects on energy consumption and carbon emissions, ensuring the scientific, accurate, and authoritative nature of green and energy-saving building construction quality evaluation.

[0046] Preferably, the sum of annual equivalent carbon emission increments is the arithmetic sum of the annual equivalent carbon emission increments of all thermal defect areas in the building, representing the total annual additional carbon emissions caused by all thermal defects; the building's annual baseline carbon emission is the total carbon emission corresponding to the annual heating and cooling energy consumption when the building envelope has no thermal defects and is constructed entirely according to design standards, serving as the benchmark reference value for carbon emission impact assessment; the percentage is the proportion of the sum of annual equivalent carbon emission increments to the building's annual baseline carbon emission, quantifying the impact of thermal defect carbon emissions on the building's total carbon emissions; the percentage threshold mapping table is pre-developed based on green building evaluation standards, building energy conservation codes, and a large amount of historical project data, used to establish the percentages. A standardized table showing the correspondence between intervals and quality deduction scores; deduction scores are points deducted from the green building baseline score based on percentage matching, reflecting the negative impact of thermal defects on construction energy-saving quality; the green building score is the initial baseline score for building energy-saving performance, with a maximum score usually set at 100 points; the final construction quality score is the final score obtained after deducting deduction scores from the green building baseline score, directly representing the building's energy-saving construction quality level; the defect list is a structured list summarizing detailed parameter information of all thermal defects; the visualized 3D report is a visual document integrating the building's 3D model, defect spatial distribution, quantitative parameters, and scoring results, used to intuitively present the evaluation results.

[0047] The first step in the process is to summarize the sum of the annual equivalent carbon emission increments for each thermal defect area. The energy-saving construction quality rating unit 5 retrieves the annual equivalent carbon emission increment data corresponding to each thermal defect area output by the equivalent heat loss calculation unit 4, checks the completeness and validity of the data one by one, and removes abnormal and invalid data. Then, the arithmetic sum of all valid annual equivalent carbon emission increment values ​​is obtained to obtain the sum of annual equivalent carbon emission increments. This sum represents the total amount of additional carbon emissions generated by all thermal defects in the building throughout the year. It is the core quantitative indicator for measuring the overall carbon emission impact of thermal defects. The larger the value, the more serious the negative impact of thermal defects on building carbon emissions. After the summation is completed, the value is bound and stored with the number of defects and the defect details to provide basic data for subsequent percentage calculations.

[0048] The second step in the process is to calculate the building's annual baseline carbon emissions. First, the design energy consumption simulation results are extracted from the building information model (BIM) during the design phase. The design energy consumption simulation is the sum of the building's annual heating and cooling energy consumption calculated by the BIM during the design phase based on local climate parameters, building envelope design parameters, indoor design temperature and humidity, building usage functions, and other standard conditions. During extraction, it is ensured that the energy consumption data unit is uniformly kilowatt-hours. After extraction, the rationality of the energy consumption data is verified to avoid simulation abnormal data. Then, the unit energy consumption carbon emission factor, consistent with the previous step, is retrieved. The extracted annual design energy consumption value is multiplied by the unit energy consumption carbon emission factor to obtain the building's annual baseline carbon emissions. This baseline carbon emissions represent the total annual carbon emissions of a building without thermal defects and fully complying with design standards. It serves as a benchmark for assessing the impact of thermal defects on carbon emissions. The value is fixed and unique. After calculation, the baseline carbon emission value is stored and archived to provide benchmark denominator data for subsequent percentage calculations.

[0049] The third step in the process is to calculate the percentage of the sum of annual equivalent carbon emission increments to the building's annual baseline carbon emissions. This is done by dividing the sum of annual equivalent carbon emission increments obtained in the first step by the building's annual baseline carbon emissions obtained in the second step. This percentage accurately quantifies the proportion of additional carbon emissions caused by thermal defects to the building's standard carbon emissions, and intuitively reflects the weight of the impact of thermal defects on the building's overall carbon emissions. For example, a percentage of 5% means that additional carbon emissions from thermal defects account for 5% of the building's standard carbon emissions. The higher the percentage, the greater the negative impact of thermal defects on the building's energy-saving performance. After the calculation, the percentage is kept to two decimal places to ensure calculation accuracy and provide a precise numerical basis for subsequent threshold mapping.

[0050] The fourth step in the process is to determine the deduction value for the green and energy-saving construction quality level based on a preset percentage threshold mapping table. The percentage threshold mapping table is pre-defined by the system and is based on the current green building evaluation standards, building energy-saving construction quality acceptance specifications, and a large amount of historical green building testing data. The mapping table is divided into multiple continuous percentage intervals, each interval corresponding to a unique deduction score. The interval division follows the principle that the greater the impact, the higher the deduction score. For example, 0% to 3% corresponds to a deduction of 0 points, 3% to 6% corresponds to a deduction of 8 points, 6% to 10% corresponds to a deduction of 18 points, and above 10% corresponds to a deduction of 30 points. The mapping table is permanently stored in the system database and cannot be modified arbitrarily to ensure the uniformity and authority of the rating standards. The energy-saving construction quality rating unit 5 reads the preset percentage threshold mapping table, matches the percentage value calculated in the third step with the percentage interval in the mapping table, determines the interval to which the percentage value belongs, and reads the deduction score corresponding to that interval. This deduction score is the quantitative deduction value for thermal defects on the construction energy-saving quality. The matching process is fully automated without human intervention to avoid subjective judgment errors.

[0051] The fifth step in the process is to calculate the final construction quality score. First, the system's preset green building basic score is retrieved. The maximum score for the basic score is set to 100 points, which means that the construction energy-saving quality fully meets the highest green building standards. Then, the deduction points obtained in the fourth step are subtracted from the green building basic score to obtain the final construction quality score. For example, when the basic score is 100 points and the deduction points are 8 points, the final score is 92 points. The final score directly corresponds to the green energy-saving construction quality level. The higher the score, the better the construction energy-saving quality. The lower the score, the more serious the thermal defects and the worse the construction energy-saving quality. After the calculation is completed, the final score is bound and stored with the deduction points and percentage values ​​to provide core scoring data for the evaluation results output.

[0052] The sixth step in the process is to output the energy-saving construction quality evaluation results. The evaluation results include three parts: the final construction quality score, the defect list, and the visualized 3D report. The defect list is a structured text list that records the key parameters of each thermal defect, including the 3D spatial coordinates of the defect, the defect area, the actual thickness of the insulation layer, the thickness deviation, the thermal resistance deviation, the annual equivalent carbon emission increment, and the severity level of the defect, ensuring that the defect information is complete and traceable. The visualized 3D report is based on the building information model in the design phase, overlaying the 3D spatial locations of all thermal defects, and using different colors to mark the severity level of the defects. For example, red marks defects with high carbon emission increment, yellow marks defects with medium carbon emission increment, and green marks defects with low carbon emission increment. At the same time, pop-up windows for key defect parameters are embedded in the corresponding positions of the model, supporting interactive viewing. The report is available in both PDF and 3D model file formats. The PDF file includes the scoring results, defect list, model screenshots, and text descriptions. The 3D model file can be opened and interactively viewed with professional software, facilitating archiving, retrieval, and reporting. It provides a scientific basis for the acceptance of green energy-saving building construction quality, energy-saving performance evaluation, and defect rectification, ensuring that the energy-saving quality of green buildings meets design standards and industry specifications.

[0053] The second objective of this invention is to provide a method for implementing a green and energy-saving building construction quality evaluation system that includes any of the above-mentioned features, comprising the following steps: S1. Scan the building's external envelope and simultaneously acquire infrared thermal images, visible light images, and 3D point cloud data; S2. The infrared thermal image and the visible light image are fused in a multimodal manner. An image segmentation algorithm is used to automatically identify thermal defect candidate regions by combining temperature anomaly region clustering and temperature gradient abrupt change boundary extraction. The thermal defect candidate regions are then mapped to the three-dimensional point cloud data to determine the spatial coordinates of each thermal defect region. S3. For each thermal defect area, extract the defect area and the actual thickness of the insulation layer at the defect location from the three-dimensional point cloud data, compare the spatial coordinate position with the building information model in the design stage, obtain the corresponding design thermal resistance value of the building envelope, and calculate the actual thermal resistance value of the defect location based on the deviation between the actual thickness of the insulation layer and the design thickness. S4. Based on the defect area, the difference between the designed thermal resistance value and the actual thermal resistance value, and combined with the annual heating days and annual cooling days of the building location, calculate the annual equivalent heat loss per unit area of ​​the thermal defect area, and then calculate the total annual equivalent heat loss and the corresponding annual equivalent carbon emission increment. Finally, based on the percentage of the sum of the annual equivalent carbon emission increments of each defect area to the building's annual baseline carbon emissions, and according to the preset percentage threshold mapping table, determine the green energy-saving construction quality level deduction value, and output the evaluation result.

[0054] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A green and energy-saving building construction quality evaluation system, characterized in that, include: The multi-mode data acquisition unit (1) scans the building envelope and simultaneously acquires infrared thermal images, visible light images and three-dimensional point cloud data. The defect area identification and localization unit (2) performs multimodal fusion of the infrared thermal image and the visible light image, adopts an image segmentation algorithm, and automatically identifies thermal defect candidate areas from the infrared thermal image by combining temperature anomaly area clustering and temperature gradient abrupt boundary extraction, maps the thermal defect candidate areas to the three-dimensional point cloud data, and determines the spatial coordinate position of each thermal defect area. The defect thermal parameter extraction unit (3) extracts the defect area and the actual thickness of the insulation layer at the defect location from the three-dimensional point cloud data for each thermal defect area, compares the spatial coordinate position of the thermal defect area with the building information model in the design stage, obtains the design thermal resistance value of the enclosure structure corresponding to the defect location, and calculates the actual thermal resistance value of the defect location based on the deviation between the actual thickness of the insulation layer and the design thickness. The equivalent heat loss calculation unit (4) calculates the annual equivalent heat loss per unit area of ​​the thermal defect area based on the defect area, the difference between the design thermal resistance value of the building envelope and the actual thermal resistance value, combined with the annual heating days and annual cooling days of the building location. The annual equivalent heat loss per unit area is multiplied by the defect area to obtain the total annual equivalent heat loss of the thermal defect area, and then multiplied by the carbon emission factor per unit energy consumption to obtain the annual equivalent carbon emission increment corresponding to the thermal defect area. The energy-saving construction quality rating unit (5) determines the green energy-saving construction quality level deduction value corresponding to the thermal defects of the building envelope based on the percentage of the sum of the annual equivalent carbon emission increments of each thermal defect area to the building's annual baseline carbon emissions, and outputs the energy-saving construction quality evaluation result.

2. The green and energy-saving building construction quality evaluation system according to claim 1, characterized in that: When the multi-mode data acquisition unit (1) scans the building envelope, it uses a drone equipped with an infrared thermal imager, a visible light camera and a lidar. The drone flies along a preset path and simultaneously triggers the three sensors to collect data. This is used to ensure that the infrared thermal image, visible light image and three-dimensional point cloud data have the same timestamp and spatial reference. The three-dimensional point cloud data collected by the lidar includes three-dimensional coordinates and reflection intensity. The infrared thermal image includes a temperature matrix and the visible light image includes RGB channels. During the acquisition, the ambient temperature and wind speed are also recorded as auxiliary data for subsequent correction.

3. The green and energy-saving building construction quality evaluation system according to claim 1, characterized in that: When the defect region identification and localization unit (2) performs multimodal fusion of the infrared thermal image and the visible light image, it adopts a deep learning network. The deep learning network contains two branches, which are used to process the infrared thermal image and the visible light image respectively. It extracts the temperature feature map of the infrared thermal image and the texture feature map of the visible light image. It calculates feature weights through an attention mechanism to weight the temperature feature map and the texture feature map. The weighted temperature feature map and the texture feature map are then spliced ​​together to form a fusion feature map. The fusion feature map retains both temperature anomaly and visual edge information to enhance the identification of thermal defect candidate regions.

4. The green and energy-saving building construction quality evaluation system according to claim 3, characterized in that: The image segmentation algorithm adopts an encoder-decoder structure. The encoder extracts features from the fused feature map by downsampling through a convolutional layer, and the decoder restores the resolution by upsampling the features through a deconvolutional layer to obtain the decoded features. A temperature gradient calculation module is introduced into the skip connection. The temperature gradient calculation module performs gradient calculation on the infrared thermal image, detects temperature change edges, obtains a temperature change edge map, and superimposes the temperature change edge map with the decoded features to output a binary segmentation map, which is used to identify candidate regions for thermal defects.

5. The green and energy-saving building construction quality evaluation system according to claim 4, characterized in that: When the defect region identification and localization unit (2) maps the thermal defect candidate region to the three-dimensional point cloud data, it uses camera calibration parameters to convert the pixel coordinates in the binary segmentation map to the lidar coordinate system, finds the corresponding three-dimensional points through the point cloud index, forms a defect point cloud cluster, calculates the convex hull boundary of the defect point cloud cluster, uses the center point of the convex hull boundary as the spatial coordinate position of the thermal defect region, and stores the convex hull boundary point set for subsequent area calculation.

6. The green and energy-saving building construction quality evaluation system according to claim 5, characterized in that: When the defect thermal parameter extraction unit (3) extracts the defect area from the three-dimensional point cloud data, it projects the defect point cloud cluster onto the best fitting plane, obtains the projected polygon, calculates the area of ​​the projected polygon as the defect area, and extracts the actual thickness of the insulation layer at the defect location. It then performs planar segmentation on the defect point cloud cluster, identifies the point cloud on the surface of the insulation layer and the point cloud on the surface of the base layer, calculates the average distance along the normal direction between the point cloud on the surface of the insulation layer and the point cloud on the surface of the base layer as the actual thickness of the insulation layer, and obtains the deviation between the actual thickness of the insulation layer and the design thickness by subtraction.

7. The green and energy-saving building construction quality evaluation system according to claim 6, characterized in that: When the defect thermal parameter extraction unit (3) compares the spatial coordinates of the thermal defect area with the building information model in the design stage, it uses the iterative nearest point algorithm to align the defect point cloud cluster with the corresponding component in the building information model, reads the design thermal resistance value and material thermal conductivity of the component from the building information model, and calculates the actual thermal resistance value of the defect location based on the deviation between the actual thickness and the design thickness of the insulation layer and the relationship that thermal resistance equals thickness divided by thermal conductivity.

8. The green and energy-saving building construction quality evaluation system according to claim 1, characterized in that: The equivalent heat loss calculation unit (4) calculates the annual equivalent heat loss per unit area of ​​the thermal defect area based on the defect area, the difference between the design thermal resistance value and the actual thermal resistance value of the building envelope, and the annual heating and cooling days of the building location. First, it calculates the additional heat flow per unit area caused by the reduction of thermal resistance. The additional heat flow is equal to the indoor and outdoor design temperature difference multiplied by the difference between the reciprocal of the design thermal resistance value and the reciprocal of the actual thermal resistance value, and multiplied by the equivalent hours converted from the annual heating and cooling days to obtain the annual equivalent heat loss per unit area. Then, it multiplies by the defect area and the carbon emission factor per unit energy consumption to obtain the annual equivalent carbon emission increment.

9. A green and energy-saving building construction quality evaluation system according to claim 8, characterized in that: The energy-saving construction quality rating unit (5) determines the green energy-saving construction quality level deduction value based on the percentage of the sum of the annual equivalent carbon emission increments of each thermal defect area to the building's annual baseline carbon emission, according to the preset percentage threshold mapping table. The building's annual baseline carbon emission is obtained by extracting the energy consumption value from the design energy consumption simulation results of the building information model and multiplying it by the unit energy consumption carbon emission factor. The percentage threshold mapping table maps the percentage range to the deduction score. The deduction score is subtracted from the green building score to obtain the final construction quality score. The output energy-saving construction quality evaluation results include the score, defect list and visual three-dimensional report.

10. A method for implementing a green and energy-saving building construction quality evaluation system according to any one of claims 1-9, characterized in that: Includes the following steps: S1. Scan the building's external envelope and simultaneously acquire infrared thermal images, visible light images, and 3D point cloud data; S2. The infrared thermal image and the visible light image are fused in a multimodal manner. An image segmentation algorithm is used to automatically identify thermal defect candidate regions by combining temperature anomaly region clustering and temperature gradient abrupt change boundary extraction. The thermal defect candidate regions are then mapped to the three-dimensional point cloud data to determine the spatial coordinates of each thermal defect region. S3. For each thermal defect area, extract the defect area and the actual thickness of the insulation layer at the defect location from the three-dimensional point cloud data, compare the spatial coordinate position with the building information model in the design stage, obtain the corresponding design thermal resistance value of the building envelope, and calculate the actual thermal resistance value of the defect location based on the deviation between the actual thickness of the insulation layer and the design thickness. S4. Based on the defect area, the difference between the designed thermal resistance value and the actual thermal resistance value, and combined with the annual heating days and annual cooling days of the building location, calculate the annual equivalent heat loss per unit area of ​​the thermal defect area, and then calculate the total annual equivalent heat loss and the corresponding annual equivalent carbon emission increment. Finally, based on the percentage of the sum of the annual equivalent carbon emission increments of each defect area to the building's annual baseline carbon emissions, and according to the preset percentage threshold mapping table, determine the green energy-saving construction quality level deduction value, and output the evaluation result.

Images