A high-rise construction building height measurement method and system

Abstract

The application provides a high-rise construction building height measurement method and system, and belongs to the technical field of building data processing, and comprises the following steps: acquiring facade video data of a building to be measured; inputting the facade video data into a real-time semantic segmentation model to perform segmentation prediction on the facade video data and obtain a video data semantic segmentation prediction result; collecting unmanned aerial vehicle barometer parameters, combining the video data semantic segmentation prediction result and the unmanned aerial vehicle barometer parameters to obtain building height information; and according to building drawing floor information, calibrating the building height information and the building drawing floor information by machine learning to obtain a floor measurement result of the building to be measured. The application obtains relatively accurate building height and floor information by constructing a key point data set in a high-rise construction building, acquiring actual measurement video data by an unmanned aerial vehicle, performing segmentation prediction by a real-time semantic segmentation model, and calibrating height information and floor information in combination with various height-related parameters.

Description

Technical Field

[0001] This invention relates to the field of building data processing technology, and in particular to a method and system for measuring the height of high-rise buildings under construction. Background Technology

[0002] In the construction of high-rise buildings, it is usually necessary to measure key nodes of the high-rise building, such as height and floors. Typically, three-dimensional point clouds collected by LiDAR are used to reconstruct the building project in three dimensions to build a three-dimensional model of the building project, and then the height of the corresponding key nodes can be read.

[0003] Currently, using LiDAR drones to scan point clouds and calculate point cloud parameters to reconstruct building heights requires significant financial and computational resources. However, construction is a dynamic process, and frequent interventions lead to wasted funds and computing power. Existing methods cannot simultaneously achieve both high measurement accuracy and low-cost resource investment.

[0004] Therefore, a new method for measuring the height of high-rise buildings under construction is needed. Summary of the Invention

[0005] This invention provides a method and system for measuring the height of high-rise construction buildings, which solves the problems of high implementation difficulty, low accuracy and high cost in the existing technology for measuring the height of high-rise construction buildings.

[0006] In a first aspect, the present invention provides a method for measuring the height of a high-rise building under construction, comprising:

[0007] Acquire video data of the exterior facade of the building to be measured;

[0008] The facade video data is input into a pre-trained real-time semantic segmentation model to perform segmentation prediction on the facade video data and obtain the semantic segmentation prediction result of the video data.

[0009] By collecting barometer parameters from the UAV and combining the semantic segmentation prediction results of the video data with the barometer parameters, the building height information is obtained.

[0010] Based on the floor information in the building drawings, machine learning is used to calibrate the building height information with the floor information in the building drawings to obtain the floor measurement results of the building to be measured.

[0011] Secondly, the present invention also provides a height measurement system for high-rise construction buildings, comprising:

[0012] The acquisition module is used to acquire video data of the exterior facade of the building to be measured.

[0013] The prediction module is used to input the facade video data into a pre-trained real-time semantic segmentation model, perform segmentation prediction on the facade video data, and obtain the semantic segmentation prediction result of the video data.

[0014] The processing module is used to collect the barometer parameters of the UAV and combine the semantic segmentation prediction results of the video data with the barometer parameters of the UAV to obtain the building height information;

[0015] The calibration module is used to calibrate the building height information with the building drawing floor information using machine learning, based on the building drawing floor information, to obtain the floor measurement results of the building to be measured.

[0016] Thirdly, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the height measurement method for high-rise construction buildings as described above.

[0017] The method and system for measuring the height of high-rise construction buildings provided by this invention constructs a dataset of key points in high-rise construction buildings, acquires measured video data by UAVs, uses a real-time semantic segmentation model for segmentation prediction, and combines multiple height-related parameters to calibrate height and floor information, thereby obtaining relatively accurate building height and floor information. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating the height measurement method for high-rise construction buildings provided by the present invention;

[0020] Figure 2 This is a diagram of the system application environment provided by the present invention;

[0021] Figure 3 This is a structural diagram of the real-time semantic segmentation model provided by the present invention;

[0022] Figure 4 This is a schematic diagram of the structure of the height measurement system for high-rise construction buildings provided by the present invention;

[0023] Figure 5 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0025] Figure 1 This is a flowchart illustrating the height measurement method for high-rise construction buildings provided in an embodiment of the present invention, as shown below. Figure 1 As shown, it includes:

[0026] Step 100: Acquire video data of the exterior facade of the building to be measured;

[0027] Step 200: Input the facade video data into a pre-trained real-time semantic segmentation model to perform segmentation prediction on the facade video data and obtain the semantic segmentation prediction result of the video data;

[0028] Step 300: Collect the barometer parameters of the UAV, and combine the semantic segmentation prediction results of the video data with the barometer parameters of the UAV to obtain the building height information;

[0029] Step 400: Based on the floor information in the building drawings, machine learning is used to calibrate the building height information with the floor information in the building drawings to obtain the floor measurement results of the building to be measured.

[0030] It should be noted that the height measurement method for high-rise construction buildings proposed in this embodiment of the invention can be applied to, for example... Figure 2 In the application environment shown, the video data and barometer parameter acquisition device communicates with the video data and barometer parameter processing platform. The data storage system can store the video data and the data that the barometer parameter processing platform needs to process. The data storage system is located in the mobile workstation.

[0031] Based on the collected measured video data, a pre-trained real-time semantic segmentation model is invoked to perform segmentation prediction on the measured video data, obtaining the semantic segmentation prediction results. The real-time semantic segmentation model includes an attention mechanism feature extraction layer and a graph convolutional feature extraction layer, as well as an attention mechanism segmentation network and a convolutional neural segmentation network. By fusing the results of the attention mechanism feature extraction layer and the graph convolutional feature extraction layer, a fused feature is obtained. This fused feature is input into the segmentation network for analysis, yielding the semantic segmentation prediction results for the measured video data. Based on the semantic segmentation prediction results, barometer parameters are read to obtain the height parameters of the currently segmented key points. Then, OCR (Optical Character Recognition) is used to recognize the building's floor plan, and a machine learning linear regression model is constructed to correct and reduce errors caused by barometer inaccuracy, ultimately determining the height of the key points. Here, the video data and barometer parameter processing platform can be, but is not limited to, various personal computers, laptops, and mobile workstations with high-performance NVIDIA graphics cards, or it can be integrated on a standalone server or a server cluster composed of multiple servers. The video data and barometer parameter acquisition device can be any drone containing a barometer and a camera.

[0032] This invention constructs a dataset of key points in high-rise construction buildings, acquires measured video data using drones, employs a real-time semantic segmentation model for segmentation prediction, and combines various height-related parameters to calibrate height and floor information, thereby obtaining relatively accurate building height and floor information.

[0033] Based on the above embodiments, acquiring the exterior video data of the building to be measured includes:

[0034] Determine the preset key nodes of the building to be measured;

[0035] The preset key nodes are scanned using a drone-borne video acquisition device to obtain video data of the facade;

[0036] The drone transmits the facade video data to a mobile workstation for processing via a preset transmission protocol.

[0037] Specifically, in this embodiment of the invention, video data of key nodes such as openings on the exterior facade of a building during construction is acquired. Key nodes are nodes that are easy to identify when observing a building during construction, including room windows, balcony doors, toilet windows, or openings. By observing the openings, the number of floors in the building can be roughly determined.

[0038] The measured video data of key nodes such as openings on the exterior facade of the building during construction can include video data of all windows, openings, balcony windows, and climbing scaffolding on the exterior facade of the building during construction. This data is obtained by scanning and collecting the exterior facade of the building during construction using a drone equipped with a camera video acquisition device.

[0039] Video data consists of a set of consecutive photos. Scanning the video data and transmitting it to the mobile workstation uses a video protocol, selected based on the drone model, such as RTSP, RTMP, HLS, or HTTP. During transmission to the mobile workstation, depending on the drone's performance and network conditions, the system may choose to use compressed streaming data, such as H.264-based encoding, or uncompressed raw streaming data.

[0040] Based on the above embodiments, the pre-trained real-time semantic segmentation model is obtained through the following steps:

[0041] Construct an initial real-time semantic segmentation model;

[0042] The initial dual-branch downsampling layer in the initial real-time semantic segmentation model is pre-trained based on a preset public dataset to obtain the weights of the training dual-branch downsampling layer.

[0043] Using a preset keypoint training dataset, the pre-trained dual-branch downsampling layer in the initial real-time semantic segmentation model is trained using the weights of the trained dual-branch downsampling layer to obtain training feature data;

[0044] The training feature data is fused through an attention mechanism feature extraction layer and a graph convolution feature extraction layer to obtain training fused features;

[0045] The training fusion features are respectively input into the attention mechanism segmentation network and the neural convolution segmentation network in the initial dual-branch upsampling network to obtain the initial segmentation prediction value;

[0046] The initial segmentation prediction value is input into the initial classification network for iterative learning until the maximum preset number of iterations is met, thus obtaining the pre-trained real-time semantic segmentation model.

[0047] The construction of the initial real-time semantic segmentation model includes:

[0048] The input layer is connected to the attention mechanism feature extraction layer and the graph convolution feature extraction layer, which are located side by side. The attention mechanism feature extraction layer and the graph convolution feature extraction layer are connected to the feature fusion layer, respectively.

[0049] The feature fusion layer is connected to the attention mechanism segmentation network and the neural convolutional segmentation network, respectively. The attention mechanism segmentation network and the neural convolutional segmentation network are connected to the decoded feature fusion layer, respectively. The decoded feature fusion layer is connected to the output layer.

[0050] The preset key point training dataset is obtained through the following steps:

[0051] Obtain images of door and window openings of different styles, crop the images of door and window openings to a preset size, save the cropped images of door and window openings using a preset encoding and a preset format, and determine the cropped images of door and window openings as the foreground of the dataset;

[0052] Obtain wall images of different styles, crop the wall images to a preset size, save the cropped wall images using preset encoding and preset format, and determine the cropped wall images as the dataset background;

[0053] The foreground and background of the dataset are merged using a preset database to generate label images in a preset annotation format. The labels and images are then stored in a label folder and an image folder, respectively.

[0054] Traverse the label folder and the image folder to generate relative address of the label and relative address of the image, and determine the labeled sample dataset based on the relative address of the label and relative address of the image;

[0055] The labeled sample dataset is divided into a training set and a validation set according to a preset ratio to obtain the preset key point training dataset.

[0056] Specifically, an initial real-time semantic segmentation model is first constructed by inputting video data. This model includes a parallel feature extraction layer based on an attention mechanism and a graph convolutional network feature extraction layer, as well as a parallel segmentation network based on an attention mechanism and a convolutional neural network. The results of the attention mechanism feature extraction layer and the graph convolutional network feature extraction layer are fused to obtain a feature fusion layer. The fused features from this layer are then input into the attention mechanism segmentation network and the convolutional neural network for analysis and decoding. After decoding, the features are fused to obtain the semantic segmentation prediction result for the measured video data. The model structure is as follows: Figure 3 As shown.

[0057] The feature extraction layer of the attention mechanism in the real-time semantic segmentation model is the Swim Transformer, which employs Shifted Window based Self-Attention. This restricts the calculation of the correlation between Q and K to within a window, thus reducing the complexity to O(n). One layer of the network uses W-MSA (Window Multihead Self-Attention), and the next layer uses SW-MSA (Shift Window Multihead Self-Attention). Since attention calculated within a window can cause the loss of global information, thus limiting the model's capabilities, the next layer achieves interaction between adjacent windows by moving the window. A problem with SW-MSA is that after moving and re-segmenting the window, the number of windows increases, and their sizes become inconsistent. To address this, a Cyclic-Shifting method is used. Through a masking mechanism, the number and size of the calculated windows remain unchanged compared to W-MSA. Specifically, Cyclic-Shifting merges non-pixel blocks caused by the movement into pixel blocks.

[0058] In real-time semantic segmentation models, the graph convolutional network feature extraction layer uses ViG, while using GNN can obtain a wider field of view. Before using the GNN network, image preprocessing is required, as follows:

[0059] To convert the image into a graph suitable for graph convolution calculations, the 2D image is first uniformly divided into 12 small patches. These patches are then mapped to feature vectors, denoted by X, i.e., X = [X1, X2, ..., X...]. 12 Each patch is a node, i.e., V = [V1, V2, ..., V]. 12 For each patch, compute its K nearest neighbors N(V). i (i∈[1,12]) and connect them to obtain the graph structure G=G(X). The GNN layer performs feature aggregation on adjacent patches and exchanges information between nodes, as follows:

[0060] G'=F(G,W)=Updata(Aggregata(G,W agg ),W updata ),

[0061] Where Aggregate represents the aggregation operation, Update represents updating the aggregated patch features, and W agg W represents the learnable weights of the aggregation operation. updataLet G represent the learnable weights for the update operation, G represent the graph structure, W represent the set of aggregated update operations, F represent the feature aggregation function, and G' represent the result after feature aggregation. Nodes are calculated by aggregating the features of adjacent patches, and then the Update operation is performed.

[0062] More specifically, the Aggregate operation computes the representation x of a node by aggregating the features of its neighboring nodes. i ":

[0063]

[0064] Where N(x) i ) is x i The set of neighboring nodes, x j It is divided by x i For nodes other than those in the graph, calculate the maximum relative graph convolution.

[0065] The update operation is a feature that further merges and aggregates:

[0066] x i '=x i "Wupdata,

[0067] The bias term is omitted here, and the graph convolution processing above can be represented as X' = GraphConv(X).

[0068] Because patches are similar to each other and GNN aggregation can easily lead to oversmoothing, ViG networks introduce a feedforward neural network (FFN) module and more linear transformations to alleviate this problem. A linear layer is added before and after the graph convolutional neural network (GNN), and a non-linear activation function is inserted after the GNN; this module is called the Grapher module.

[0069] Given the input feature vector D is the feature dimension:

[0070] Y = σ(GraphConv(XW) in ))W out +X,

[0071] In the formula W in W out σ represents the weights of the fully connected layer, σ is the activation function, and the bias term is omitted.

[0072] To further mitigate the over-smoothing problem, a feedforward network is used on each node:

[0073] Z=σ(YW1)W2+Y

[0074] In the formula W1 and W2 are the weights of the fully connected layer; the bias term is omitted.

[0075] The resulting graph is then reconstructed into a 2D image structure, that is, each patch is restored to its initial position. Then, an aggregation layer is used to fuse the features extracted by the attention mechanism feature network and the graph convolutional network feature network.

[0076] Then, the attention-based segmentation network and the convolutional neural segmentation network segment the fused features.

[0077] Specifically, the attention mechanism segmentation network uses the segmentation head of Segformer, employing four multilayer perceptron (MLP) layers. Features from different layers are passed through a linear layer to ensure that their channel dimensions are the same, upsampled to 56×56 resolution, and then the images are concatenated and fused through a linear layer.

[0078] The convolutional neural network segmentation network uses a convolutional neural network segmentation head, referencing the upsampling method of U-Net. Through four upsampling passes, the extracted features are upsampled to a resolution of 56×56, and then the images are concatenated and fused. This is then fused with the results obtained from the attention-based segmentation network. Finally, it is upsampled to a resolution of 224×224 to obtain the final result.

[0079] Before training the model, it is also necessary to create a dataset of key points for the construction buildings, which includes:

[0080] Images of door and window openings of different styles were obtained through on-site photography and online searches. Each image was cropped using Adobe Photoshop, and the door and window openings were extracted, ensuring a size smaller than 224×224 resolution. These images were then numbered and saved in PNG format as the foreground of the dataset.

[0081] Images of different wall styles were obtained through on-site photography and online searches. These images were cropped to a size of 224×224, numbered, and saved in PNG format as the background of the dataset.

[0082] Using Python in conjunction with libraries such as CV2 and OS, the foreground and background of the numbered images are merged, and PNG label images similar to Cityscapes format are automatically generated and saved to the image folder and label folder.

[0083] Iterate through the image folder and tag folder, generate relative addresses of images and tags, save them to a text file, and complete the creation of a dataset in a Cityscapes-like format.

[0084] Based on the labeled sample dataset, the dataset is first randomly divided into a training set and a validation set in an 8:2 ratio to train the initial real-time semantic segmentation model and obtain the real-time semantic segmentation model.

[0085] Furthermore, a real-time semantic segmentation network is trained using a dataset of key points of buildings under construction, specifically including:

[0086] The initial two-branch downsampling layer in the initial real-time semantic segmentation model is pre-trained. The graph neural network and attention mechanism network are first trained on the public dataset ImageNet to perform classification tasks on each image and the corresponding labeled image, and the training weights of the two-branch downsampling layer are saved.

[0087] Import the training weights of the dual-branch downsampling layer and continue training the pre-trained dual-branch downsampling layer in the real-time semantic segmentation model. Extract the training attention mechanism feature data and training graph convolutional feature data of the video data in the labeled sample dataset through the pre-trained attention mechanism feature extraction layer and the pre-trained graph convolutional feature extraction layer in the semantic segmentation model. Extract the training attention mechanism feature data of the video data in the labeled sample dataset through the pre-trained attention mechanism feature extraction layer in the pre-trained semantic segmentation model. Extract the training graph convolutional feature data of the video data in the labeled sample dataset through the pre-trained graph convolutional feature extraction layer on the parallel branch with the pre-trained attention mechanism feature extraction layer.

[0088] The data from the attention mechanism feature extraction layer and the graph convolution feature extraction layer are fused to obtain the training fused features. It can be understood that the fusion network is composed of neural network structures such as linear layers and convolutional layers.

[0089] The training fusion features are respectively input into the attention mechanism network and the neural convolutional network in the initial dual-branch upsampling network to obtain the initial segmentation prediction value;

[0090] The initial upsampled predicted values ​​are input into the initial classification network for iterative learning. The termination condition for this iterative learning can be set based on the maximum number of iterations, which is the upper limit threshold for expanding the receptive field. This upper limit threshold can be determined based on the MIoU performance of the semantic segmentation metric within the receptive field. The output of the feature extraction network is used as the input to the segmentation prediction network, and vice versa, allowing for mutual guidance and iterative learning. This further improves the segmentation performance of the real-time semantic segmentation network. Enhancing the interrelationships between networks improves the accuracy of the final real-time semantic segmentation model.

[0091] When the mutual guided iterative learning meets the iteration termination condition, the real-time semantic segmentation model at this point is determined as the trained real-time semantic segmentation model. By using the output of the feature extraction network as the input to the segmentation prediction network, and vice versa, mutual guided iterative learning is performed until the iteration termination condition is met, thus obtaining the real-time semantic segmentation model. The purpose of training the semantic segmentation model is to obtain the optimal computational parameters, thereby ensuring the accuracy of semantic segmentation in practical use and improving algorithm efficiency.

[0092] Based on the above embodiments, the step of inputting the facade video data into a pre-trained real-time semantic segmentation model to perform segmentation prediction on the facade video data and obtain the semantic segmentation prediction result of the video data includes:

[0093] Obtain each frame of video image in the facade video data and the pre-trained parameter weights of the real-time semantic segmentation model, and downsample each frame of video image based on the pre-trained parameter weights to obtain each downsampled frame of video image;

[0094] Each downsampled video frame is input into the attention mechanism feature extraction layer and the graph convolution feature extraction layer to obtain the feature map extracted by the attention mechanism and the feature map extracted by the graph convolution feature extraction layer.

[0095] The feature map extracted by the attention mechanism and the feature map extracted by the graph convolution are fused to obtain a fused feature map;

[0096] The fused feature map is input into the attention mechanism segmentation network and the neural convolutional segmentation network respectively for upsampling to obtain the feature map of the attention channel and the feature map of the neural convolutional network channel.

[0097] The feature maps of the attention channel and the neural convolutional network channel are fused together, and pixel classification is performed based on the target category in the real-time semantic segmentation model. Then, an upsampling operation is performed to obtain a mask image of a preset target size.

[0098] Each frame of video image is sequentially input into the real-time semantic segmentation model to obtain overlay mask video data.

[0099] Specifically, in this embodiment of the invention, the cv2.VideoCapture() function in the CV2 library is used to read the input video data to obtain each frame of the video; then, the real-time semantic segmentation model is imported and the model reads the pre-trained parameter weights; then, a downsampling operation is performed to obtain each frame of the downsampled video; finally, each frame of the video is passed into the real-time semantic segmentation model.

[0100] After each frame of the video is fed into the real-time semantic segmentation model, the image passes through the attention mechanism feature extraction layer and the graph convolution feature extraction layer to obtain the feature map extracted by the attention mechanism and the feature map extracted by the graph convolution feature extraction layer, respectively.

[0101] The feature maps extracted by the attention mechanism and the feature maps extracted by graph convolution are fused to obtain the fused feature map extracted by the dual-branch network. Then, the fused feature map extracted by the dual-branch network is upsampled through the attention mechanism and the neural convolutional network to obtain the feature maps of the attention channel and the neural convolutional network channel.

[0102] Finally, the feature maps of the attention channel and the neural convolutional network channel are fused, and the pixels are classified according to the target category in the real-time semantic segmentation model. After upsampling, a mask image of the target size is obtained. Each frame of the video is continuously input into the real-time semantic segmentation model, and the model continuously outputs the corresponding mask image, ultimately obtaining the video covered by the mask.

[0103] Based on the above embodiments, the step of collecting UAV barometer parameters, combining the semantic segmentation prediction results of the video data and the UAV barometer parameters to obtain building height information includes:

[0104] Convert the UAV barometer parameters into altitude parameters;

[0105] The semantic segmentation prediction results of the video data are mapped to the height parameters to obtain the building height information.

[0106] Specifically, the mobile workstation performs real-time semantic segmentation on the video input and sets a baseline at the same position in each frame of the image. When a complete mask passes through the baseline, the barometric altitude parameter of the UAV barometer is read in. At this time, the real-time semantic segmentation model of the mobile workstation and the barometric altitude data measured by the UAV barometer constitute a multimodal interaction, which can obtain the building height information.

[0107] Based on the above embodiments, the step of using machine learning to calibrate the building height information with the building drawing floor information to obtain the floor measurement results of the building to be measured includes:

[0108] The building floor information in the building drawings is obtained by OCR text recognition, and the theoretical height of the floor is obtained by modeling based on the building drawing floor information;

[0109] The actual floor height is obtained from the building height information using a linear regression function.

[0110] The final actual height and number of floors of the building to be measured are obtained by averaging the theoretical height of the floors and the actual height of the floors.

[0111] It is understood that, in this embodiment of the invention, data with a mapping relationship to building model data is also generated by combining pre-identified drawing data. Specifically, floor plans in the drawings are identified using OCR to obtain building model data, and coefficients in a machine learning linear regression model are constructed to provide the height parameters of key nodes. Barometer parameters are imported, converted into height, and the regression model is used to correct the results, reducing errors caused by barometer inaccuracy, and ultimately determining the height of key building nodes.

[0112] The height measurement system for high-rise construction buildings provided by the present invention is described below. The height measurement system for high-rise construction buildings described below can be referred to in correspondence with the height measurement method for high-rise construction buildings described above.

[0113] Figure 4 This is a schematic diagram of the structure of the high-rise building height measurement system provided in an embodiment of the present invention, as shown below. Figure 4 As shown, it includes: an acquisition module 41, a prediction module 42, a processing module 43, and a calibration module 44, wherein:

[0114] The acquisition module 41 is used to acquire video data of the exterior facade of the building to be measured; the prediction module 42 is used to input the video data of the exterior facade into a pre-trained real-time semantic segmentation model to perform segmentation prediction on the video data of the exterior facade and obtain the semantic segmentation prediction result of the video data; the processing module 43 is used to collect the barometer parameters of the UAV, and combine the semantic segmentation prediction result of the video data with the barometer parameters of the UAV to obtain the building height information; the calibration module 44 is used to calibrate the building height information with the floor information of the building drawings according to the floor information of the building drawings using machine learning to obtain the floor measurement result of the building to be measured.

[0115] Figure 5 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 5As shown, the electronic device may include a processor 510, a communication interface 520, a memory 530, and a communication bus 540, wherein the processor 510, the communication interface 520, and the memory 530 communicate with each other through the communication bus 540. The processor 510 can call logical instructions in the memory 530 to execute a method for measuring the height of a high-rise building under construction. This method includes: acquiring video data of the facade of the building to be measured; inputting the facade video data into a pre-trained real-time semantic segmentation model to perform segmentation prediction on the facade video data to obtain a semantic segmentation prediction result for the video data; collecting barometer parameters from a UAV, combining the semantic segmentation prediction result for the video data with the barometer parameters to obtain building height information; and calibrating the building height information with the floor information on the building drawings using machine learning based on the floor information on the building drawings to obtain the floor measurement result of the building to be measured.

[0116] Furthermore, the logical instructions in the aforementioned memory 530 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0117] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the high-rise construction building height measurement method provided by the above methods. The method includes: acquiring exterior facade video data of the building to be measured; inputting the exterior facade video data into a pre-trained real-time semantic segmentation model to perform segmentation prediction on the exterior facade video data to obtain a video data semantic segmentation prediction result; collecting UAV barometer parameters, combining the video data semantic segmentation prediction result and the UAV barometer parameters to obtain building height information; and using machine learning to calibrate the building height information with the building drawing floor information according to the building drawing floor information to obtain the floor measurement result of the building to be measured.

[0118] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements a method for measuring the height of a high-rise construction building provided by the methods described above. The method includes: acquiring video data of the facade of the building to be measured; inputting the video data of the facade into a pre-trained real-time semantic segmentation model to perform segmentation prediction on the video data of the facade to obtain a semantic segmentation prediction result of the video data; collecting barometer parameters from a UAV and combining the semantic segmentation prediction result of the video data with the barometer parameters of the UAV to obtain building height information; and calibrating the building height information with the floor information of the building drawings using machine learning based on the floor information of the building drawings to obtain the floor measurement result of the building to be measured.

[0119] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0120] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0121] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A high-rise construction building height measurement method characterized by, The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. To convert the image into a graph that can be calculated by graph convolution, first, the 2D image is uniformly divided into 12 small blocks, the small blocks are called patches, the patches are mapped into feature vectors, denoted as X, that is, X = [X1, X2, …, X 12 ], each patch is a node, that is, V = [V1, V2, …, V 12 ], for each patch, calculate its K-nearest neighbor N(V i ), i∈[1, 12] and connect, obtain the graph structure G = G(X), the GNN layer performs feature aggregation on adjacent patches and exchanges information between nodes, and the operation is as follows: wherein, The application relates to a building height measurement method and device. denotes a pooling operation, The application relates to a building height measurement method and device. denotes updating the pooled patch features, denotes a learnable weight for the pooling operation, denotes a learnable weight for the updating operation, denotes a graph structure, denotes a set of aggregation update operations, denotes a feature aggregation function, denotes the result after feature aggregation, the features of adjacent patches are aggregated, The application relates to a building height measurement method and device. the node is calculated, and then the The application relates to a building height measurement method and device. operation is performed.

2. The high-rise building height measurement method according to claim 1, wherein The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device.

3. The high-rise building height measurement method according to claim 1, wherein The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement method and device. The application relates to a building height measurement The preset key point training dataset is obtained by the following steps: Obtain door and window opening pictures of different styles, crop the door and window opening pictures to a preset size, save the cropped door and window opening pictures in a preset encoding and a preset format, and determine the cropped door and window opening pictures as dataset foregrounds; Obtain wall pictures of different styles, crop the wall pictures to a preset size, save the cropped wall pictures in a preset encoding and a preset format, and determine the cropped wall pictures as dataset backgrounds; Merge the dataset foregrounds and the dataset backgrounds through a preset database to generate label pictures in a preset annotation format, and store labels and pictures in a label folder and a picture folder respectively; 4. The high-rise building height measurement method according to claim 3, wherein Iterate through the label folder and the picture folder to generate label relative addresses and picture relative addresses, and determine an annotation sample dataset based on the label relative addresses and the picture relative addresses; Divide the annotation sample dataset into a training set and a validation set according to a preset ratio to obtain the preset key point training dataset. The unmanned aerial vehicle barometer parameters are collected, the video data semantic segmentation prediction result and the unmanned aerial vehicle barometer parameters are combined to obtain building height information, and the building height information is obtained by the following steps:

5. The high-rise building height measurement method according to claim 3, wherein The unmanned aerial vehicle barometer parameters are converted into height parameters; The video data semantic segmentation prediction result is corresponded with the height parameters to obtain the building height information. According to the building drawing floor information, the building height information is calibrated with the building drawing floor information by machine learning to obtain the floor measurement result of the building to be measured, and the building drawing floor information is obtained by the following steps: Building drawing floor information in a building drawing is obtained by OCR text recognition, and a floor theoretical height is obtained by modeling based on the building drawing floor information; ​ ​ 6. The high-rise building height measurement method according to claim 1, wherein ​ ​ ​ 7. The high-rise building height measurement method according to claim 1, wherein ​ ​ obtaining the actual floor height of the building based on the building height information by a linear regression function; averaging the theoretical floor height and the actual floor height to obtain the final actual height and the number of floors of the building to be measured.

8. A high-rise construction building height measurement system based on the high-rise construction building height measurement method according to any one of claims 1 to 7, characterized by, The method comprises the steps of: acquiring facade video data of a building to be measured; inputting the facade video data into a pre-trained real-time semantic segmentation model to perform segmentation prediction on the facade video data and obtain a video data semantic segmentation prediction result; processing drone barometer parameters in combination with the video data semantic segmentation prediction result and the drone barometer parameters to obtain building height information; calibrating the building height information with floor information of a building drawing by machine learning according to the floor information of the building drawing to obtain a floor measurement result of the building to be measured.

9. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor implements the high-rise building height measurement method according to any one of claims 1 to 7 when executing the program.

Images