Method, device and equipment for predicting building height through remote sensing image DOM and DSM and medium

Abstract

The invention provides a method, a device, equipment and a medium for predicting building height through DOM and DSM of remote sensing images, relates to the field of remote sensing image processing, and is used for solving the problems of low efficiency, high cost, complex operation and strong environmental dependence of building height extraction in the prior art. The method comprises the following steps: acquiring a remote sensing image DOM and digital surface model DSM data of a target area; according to the DOM, extracting a target contour of a target building in the target area; calculating an overlapping area of the target contour and the DSM data; and determining the maximum value and the minimum value of the DSM data in the overlapped area, and determining the difference value between the maximum value and the minimum value as the height of the target building. According to the method, the dependence on people is reduced, the building height is fully automatically and quickly extracted, the processing speed is higher, the requirement on imaging conditions is lower, the cost is lower, and the operation is simpler and more convenient.

Description

Technical Field

[0001] The present application relates to the field of remote sensing image processing, and provides a method, device, equipment and medium for predicting building height using remote sensing image DOM and DSM. Background Art

[0002] At present, the extraction of building height mainly relies on manual measurement and semi-automatic measurement methods. Manual measurement is usually done by surveyors using high-precision measuring equipment such as total stations and laser rangefinders to measure building heights. Semi-automatic measurement usually uses unmanned aerial vehicles (UAVs) equipped with high-precision cameras for aerial photography, and extracts building heights through image processing technology.

[0003] Manual measurement requires a lot of manpower and time, and the extraction efficiency is low. The purchase and maintenance costs of high-precision measurement equipment and drone systems are high, which increases the overall cost of the project. UAV flight missions and subsequent data processing require professional knowledge and skills, which increases the difficulty and uncertainty of operation. In addition, measurement and shooting are highly dependent on external environmental conditions. Bad weather and complex terrain will affect data quality and the accuracy of measurement results.

[0004] Therefore, traditional manual measurement and semi-automatic measurement methods have problems such as high cost, low efficiency, complex operation and strong environmental dependence in building height extraction, which limits their feasibility in large-scale applications. Summary of the invention

[0005] The present application provides a method, device, equipment and medium for predicting building height using remote sensing image DOM and DSM, which is used to solve the problems of low efficiency, high cost, complex operation and strong environmental dependence in extracting building height in the prior art.

[0006] In a first aspect, the present application provides a method for predicting building height using remote sensing images DOM and DSM, comprising:

[0007] Obtain remote sensing images DOM and digital surface model DSM data of the target area;

[0008] Extracting a target outline of a target building in the target area according to the DOM;

[0009] Calculating the overlap area between the target contour and the DSM data;

[0010] The maximum value and the minimum value of the DSM data in the overlapped area are determined, and the difference between the maximum value and the minimum value is determined as the height of the target building.

[0011] Optionally, the step of obtaining remote sensing images DOM and digital surface model DSM data of the target area includes:

[0012] Get the initial DOM and initial DSM data of the target area;

[0013] Using ArcGIS software to perform radiation correction and geometric correction on the initial DOM to obtain a corrected DOM;

[0014] The corrected DOM and the initial DSM data are registered to obtain the remote sensing image DOM and digital surface model DSM data of the target area.

[0015] Optionally, extracting the building outline of the target building in the target area according to the DOM includes:

[0016] The DOM is input into a trained neural network model, and a target outline of a target building in the target area is output; wherein the trained neural network model is obtained by training an initial neural network model based on a building dataset; the building dataset includes images of multiple sample buildings and a true outline and a true label of each sample building.

[0017] Optionally, before inputting the DOM into the trained neural network model and outputting the target outline of the target building in the target area, the method further includes:

[0018] Based on the contour loss and classification loss, the total loss function of the initial neural network model is established;

[0019] Inputting the image of each sample building into the initial neural network model, and outputting the predicted outline and predicted category of each sample building;

[0020] The predicted outline, true outline, predicted label and true label of each sample building are substituted into the total loss function, and the weight of the initial neural network model is updated according to the value of the total loss function until the value of the loss function is less than a preset error threshold or the maximum number of iterations is reached, thereby obtaining a trained neural network model.

[0021] Optionally, the total loss function of the initial neural network model is established based on the contour loss and the classification loss, including:

[0022] The contour loss and classification loss are weighted summed to establish the total loss function of the initial neural network model.

[0023] Optionally, the contour loss adopts Hausdorff loss, and the calculation formula of the Hausdorff loss is as follows:

[0024]

[0025] Among them, P irepresents the predicted outline of the i-th sample building, G i represents the true outline of the i-th sample building, N is the number of sample buildings, It represents the maximum distance from each point a in the predicted contour to the nearest point b in the true contour. It represents the maximum distance from each point b in the true contour to the nearest point a in the predicted contour.

[0026] Optionally, the calculation formula of the classification loss is as follows:

[0027]

[0028] Among them, y i is the true label of the i-th sample building, is the predicted probability of the initial neural network model for the i-th sample building, and N is the number of sample buildings.

[0029] In a second aspect, the present application provides a device for predicting building height using remote sensing images DOM and DSM, the device comprising:

[0030] Acquisition module, used to obtain remote sensing image DOM and digital surface model DSM data of the target area;

[0031] An extraction module, used for extracting a target outline of a target building in the target area according to the DOM;

[0032] A calculation module, used for calculating the overlap area between the target contour and the DSM data;

[0033] The determination module is used to determine the maximum value and the minimum value of the DSM data in the overlapped area, and determine the difference between the maximum value and the minimum value as the height of the target building.

[0034] In a third aspect, the present application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method for predicting building height using remote sensing image DOM and DSM as described in the first aspect.

[0035] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, and a processor executes the computer program to implement the method for predicting building height using remote sensing image DOM and DSM as described in the first aspect.

[0036] Compared with the prior art, the beneficial effects of this application are:

[0037] The present application proposes a method for predicting building height using remote sensing image DOM and DSM, the method comprising: obtaining remote sensing image DOM and digital surface model DSM data of the target area; extracting the target outline of the target building in the target area according to DOM; calculating the overlapping area between the target outline and the DSM data; determining the maximum and minimum values ​​of the DSM data in the overlapping area, and determining the difference between the maximum and minimum values ​​as the height of the target building. This method only requires DOM and DSM data to realize automatic recognition of building outlines and automatic extraction of building heights, reducing dependence on humans and having faster processing speed. In addition, the collection of DOM and DSM data in this method has no requirements for environmental conditions and has lower requirements for imaging conditions. This method does not require high-precision measurement equipment and drone systems, so it is less expensive and easier to operate. This method also has the following beneficial effects:

[0038] 1. Compared with traditional methods that take weeks or even months to process all the building outlines and heights of a city, the method provided by this application can complete the same task in a few minutes, greatly improving processing efficiency.

[0039] 2. By optimizing the neural network model and algorithm, the present application reduces the demand for computing resources, enabling it to run efficiently on ordinary computer equipment. The shortened processing time and reduced computing resource requirements make the present application more economical and efficient in large-scale applications.

[0040] 3. This application realizes a fully automated process from data collection, preprocessing, contour extraction to building height calculation, reducing dependence on manpower. Fully automated processing not only improves work efficiency, but also reduces the errors caused by manual operation and improves the reliability of data processing.

[0041] 4. The method provided in this application is suitable for fields such as large-scale urban planning and environmental monitoring. It can quickly provide accurate building height data, provide efficient and reliable data support for urban planners and environmental monitoring personnel, and assist in scientific decision-making and management. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 A schematic flow chart of a method for predicting building height using remote sensing images DOM and DSM provided in an embodiment of the present application;

[0043] Figure 2 A schematic diagram of the overall process of the method for predicting building height using remote sensing images DOM and DSM provided in an embodiment of the present application;

[0044] Figure 3 A comparison diagram of the building height before and after effects provided in the embodiment of the present application;

[0045] Figure 4A schematic diagram of the structure of a device for predicting building height using remote sensing images DOM and DSM provided in an embodiment of the present application. DETAILED DESCRIPTION

[0046] It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0047] The method of the embodiment of the present application can be executed in the form of a program by running a program on an existing computer device. The computer device can be a mobile phone, a tablet, a desktop computer, etc. The computer device may include a processor, a storage medium, etc. The storage medium is used to store the program for executing the method of the present embodiment, and the processor is used to load and execute the computer program so that the computer device executes the method for predicting building height using remote sensing image DOM and DSM as provided in the embodiment of the present application.

[0048] First, the background technology involved in the embodiments of the present application is introduced in detail.

[0049] At present, the extraction of building height mainly relies on manual measurement and semi-automatic measurement methods. Manual measurement is usually performed by surveyors using total stations, laser rangefinders and other tools. Although this method has high accuracy, it is complex, time-consuming and requires a lot of human resources, resulting in low measurement efficiency and unable to meet the needs of large-scale building height extraction.

[0050] Semi-automatic measurement usually uses unmanned aerial vehicles (UAVs) equipped with high-precision cameras to take aerial photos, and then extracts the building height through image processing technology. The flight path of the drone needs to be carefully designed to ensure that stereo pairs with sufficient overlap and appropriate baseline distance are obtained. Although this method improves measurement efficiency, there are still some problems and challenges:

[0051] 1. High cost: The cost of drone equipment and its maintenance is high, and flight missions require professional operators, which increases the overall operating cost.

[0052] 2. Complexity of data acquisition: In order to obtain high-quality data, the planning and execution of UAV flight missions are complex and time-consuming, and need to be carried out under suitable weather and lighting conditions to ensure image quality.

[0053] 3. High requirements for imaging conditions: UAV image processing has high requirements for the shooting environment. Factors such as weather, lighting, wind speed, etc. may affect data quality.

[0054] 4. Long processing time: Extracting building heights from stereo image pairs requires complex image matching and 3D reconstruction algorithms, which are computationally intensive and time-consuming, and have high demands on computing resources.

[0055] In summary, although manual measurement and semi-automatic measurement are widely used in building height extraction, they have problems such as high cost, low efficiency, complex operation and strong environmental dependence in building height extraction, which limits their feasibility in large-scale applications.

[0056] In view of this, the present application embodiment provides a method for predicting building height using remote sensing images DOM and DSM, which can be executed by a computer device. Figure 1 As shown, it is a flow chart of a method for predicting building height using remote sensing image DOM and DSM provided in an embodiment of the present application. The flow of the method is described as follows.

[0057] S101, obtaining remote sensing image DOM and digital surface model DSM data of the target area.

[0058] In the specific implementation process, remote sensing images, namely digital orthophoto maps (DOM), are digital orthophoto sets generated by digital differential correction and mosaicking of aerial (or aerospace) photos, cropped according to a certain map range, and have both map geometric accuracy and image features. The initial DOM of the target area can be obtained using high-resolution remote sensing satellites or drones and sent to computer equipment.

[0059] Furthermore, a digital surface model (DSM) is a three-dimensional geographic data model used to represent the height information of the earth's surface and objects above it. There are two ways to obtain DSM data of a target area, which are introduced below.

[0060] The first method is to use high-precision lidar to generate initial DSM data of the target area.

[0061] First, the target area is scanned by LiDAR to obtain point cloud data. Secondly, the point cloud data is processed, usually including denoising, point cloud classification (such as buildings, trees, ground, etc.) and point cloud filtering. Then, the processed point cloud data is converted into rasterized initial DSM data through interpolation algorithms (such as Kriging interpolation or nearest neighbor interpolation). Finally, the initial DSM data is sent to the computer device.

[0062] The second method is to use stereo image pairs to generate initial DSM data of the target area.

[0063] First, use a camera with good geometric accuracy to capture a stereo pair of images of the target area, ensuring that there is sufficient overlap and perspective difference between the images. Then, the stereo image pair is registered to ensure that the geometric relationship between the images is correct. Then, a stereo matching algorithm is used to identify and match the same features in the image pair, and calculate the disparity map of the pixels. Each pixel in the disparity map corresponds to the height information of a point on the surface. Finally, the disparity map is converted into initial DSM data and sent to the computer device.

[0064] In a possible embodiment, after acquiring the initial DOM and initial DSM data of the target area, the computer device may directly use the initial DOM of the target area as the DOM of the target area, and use the initial DSM data of the target area as the DSM data of the target area.

[0065] Considering that not all DOMs are correct source data, in a possible embodiment, the initial DOM and initial DSM data of the target area are obtained; the initial DOM is subjected to radiation correction and geometric correction using ArcGIS software to obtain the corrected DOM; the corrected DOM and the initial DSM data are registered to obtain the remote sensing image DOM and digital surface model DSM data of the target area.

[0066] In the specific implementation process, ArcGIS is a geographic information system (GIS) software developed by Esri for creating, managing, analyzing and sharing spatial data. In ArcGIS, the steps for performing radiation correction and geometric correction on DOM are as follows:

[0067] First, open ArcGIS, load the initial DOM of the target area, use the "Radiation Correction" function in the "Image Analysis" tool, select the appropriate radiation correction method (such as atmospheric correction or radiation linear conversion), enter the required radiation correction parameters (such as sensor characteristics, atmospheric conditions, etc.), and perform the correction operation.

[0068] Then use the "Geometric Correction" function in the "Image Processing" tool, load the ground control points (GCPs), select the transformation type (such as first-order polynomial transformation, second-order polynomial transformation, etc.), and perform image geometric correction based on the control points to generate the corrected DOM.

[0069] Finally, load the initial DSM data of the target area in ArcGIS, and use the "registration" or "alignment" tool to spatially align the corrected DOM with the initial DSM data to ensure that the two are completely aligned in space.

[0070] In the embodiment of the present application, the remote sensing image DOM and DSM data are preprocessed, including radiation correction, geometric correction and registration. Radiation correction is to adjust the radiation value of the image to eliminate the influence of factors such as sensor, lighting or atmospheric conditions on the image quality. Geometric correction is to adjust the geometric position of the image to align with the actual geographic coordinate system. Registration can ensure 100% accurate matching of the image and DSM data to ensure the accuracy of subsequent calculations.

[0071] S102: Extracting the target contour of the target building in the target area according to the DOM.

[0072] In a possible embodiment, the DOM is input into a trained neural network model, and a target outline of a target building in a target area is output. The trained neural network model is obtained by training an initial neural network model based on a building dataset; the building dataset includes images of multiple sample buildings and a true outline and a true label of each sample building.

[0073] In the specific implementation process, a variety of building images and building mask data are collected to construct a building data set for model training, where the mask is usually binary, that is, each pixel has only two possible values, usually 0 and 1, and the pixel value 1 represents the visible area (the real outline of the building), and the pixel value 0 represents the masked area (the background area except the building). The initial neural network model can be various neural networks, such as a convolutional neural network. The initial neural network model is trained based on the building data set to obtain a trained neural network model. The training process of the neural network model is introduced below.

[0074] In a possible embodiment, before the DOM is input into the trained neural network model and the target contour of the target building in the target area is output, a total loss function of the initial neural network model is established based on the contour loss and the classification loss; the image of each sample building is input into the initial neural network model, and the predicted contour and predicted category of each sample building are output; the predicted contour, true contour, predicted label and true label of each sample building are substituted into the total loss function, and the weight of the initial neural network model is updated according to the value of the total loss function until the value of the loss function is less than a preset error threshold or the maximum number of iterations is reached, thereby obtaining a trained neural network model.

[0075] In the specific implementation process, the image of each sample building is input into the initial neural network model, and the predicted contour and predicted label of each sample building are output. The contour loss is calculated based on the predicted contour and the true contour output by the model, and the classification loss is calculated based on the predicted label and the true label output by the model. The value of the total loss function is calculated based on the contour loss and the classification loss, and the model parameters are back-propagated based on the value of the total loss function, and the model weights are updated using the optimization algorithm. Repeat the forward propagation, loss calculation and optimization steps until the termination condition is met, such as the value of the total loss function is less than the preset error threshold, or the maximum number of iterations is reached, etc., to obtain the trained neural network model.

[0076] In a possible embodiment, based on the contour loss and the classification loss, a total loss function of the initial neural network model is established, including:

[0077] The contour loss and classification loss are weighted summed to establish the total loss function of the initial neural network model.

[0078] For example, the total loss function is calculated as follows:

[0079] L=α×L1+β×L2

[0080] Among them, L represents the total loss function, L1 represents the contour loss, α represents the weight of the contour loss, L2 represents the classification loss, and β represents the weight of the classification loss.

[0081] The determination of weights α and β usually depends on the specific task and experimental results. You can try different combinations of α and β, observe the performance of the model on the validation set, and select a combination that performs well in both contour segmentation and classification tasks.

[0082] In an embodiment of the present application, contour loss is used to measure the model's accuracy in extracting building contours, and classification loss is used to measure the model's accuracy in classifying whether an image contains building contours. Two weight parameters are used to balance the contribution of contour loss and classification loss to the total loss, so that the neural network model can maintain the quality of contour extraction while maintaining classification accuracy, thereby improving the overall task performance of the neural network model.

[0083] In a possible embodiment, the contour loss adopts Hausdorff loss, and the calculation formula of Hausdorff loss is as follows:

[0084]

[0085] Among them, L1 represents the contour loss, P i represents the predicted outline of the i-th sample building, G i represents the true outline of the i-th sample building, N is the number of sample buildings, It represents the maximum distance from each point a in the predicted contour to the nearest point b in the true contour. It represents the maximum distance from each point b in the true contour to the nearest point a in the predicted contour.

[0086] In an embodiment of the present application, the Hausdorff loss function is used as the contour loss and applied to building contour extraction. The contour of a building usually has a complex geometric shape. The Hausdorff loss function can process the geometric features of these complex shapes by focusing on the distance between the farthest point pairs, and accurately measure the maximum deviation between the predicted contour and the true contour, thereby improving the accuracy of building contour extraction and ensuring that the model can accurately identify the boundaries of the building.

[0087] In one possible embodiment, the classification loss is calculated as follows:

[0088]

[0089] Among them, L2 represents the classification loss, y i is the true label of the i-th sample building, is the predicted probability of the initial neural network model for the i-th sample building, and N is the number of sample buildings.

[0090] In the embodiment of the present application, cross-entropy loss is used as the classification loss, which directly measures the difference between the predicted probability distribution and the true label probability distribution. It encourages the probability distribution of the model output to be as close as possible to the distribution of the true label, thereby improving the classification accuracy.

[0091] S103: Calculate the overlap area between the target contour and the DSM data.

[0092] In the specific implementation process, the target contour of the target building is represented by a building template Mask. The building template Mask refers to a binary mask image that marks the building contour. The pixel value of the contour area is 1, and the pixel value of the background or other areas is 0. By multiplying the building template Mask with the DSM, the overlapping area can be obtained.

[0093] S104: Determine the maximum and minimum values ​​of the DSM data in the overlapping area, and determine the difference between the maximum and minimum values ​​as the height of the target building.

[0094] For example:

[0095]

[0096] H=H max -H min

[0097] Among them, R represents the overlapping area between the target contour and the DSM data, r represents each pixel value in the overlapping area, max() represents the maximum value function, min() represents the minimum value function, and H max represents the maximum value in the overlap area, H min represents the minimum value in the overlapping area, and H represents the height of the target building.

[0098] In the specific implementation process, because the prediction accuracy of the neural network model will not be 100%, the overlap area will not overlap 100% with the building on the DSM, but roughly overlap, for example, 80% overlap, but it is likely to overlap with the ground and the highest point of the building. Therefore, the maximum value of the DSM in the overlap area is most likely the roof height of the building, and the minimum value of the DSM in the overlap area is most likely the ground height. The height of the building can be obtained by subtracting the minimum value from the maximum value, that is, subtracting the ground height from the roof height of the building.

[0099] Please refer to Figure 2 , an overall flow chart of the method for predicting building height using remote sensing images DOM and DSM provided in an embodiment of the present application.

[0100] S201. Collect building dataset D.

[0101] S202. Train the model N according to the building dataset D, Hausdorff loss and classification loss.

[0102] S203, obtaining a building template M based on the remote sensing image DOM and the model N prediction.

[0103] S204: Calculate the overlapping area R according to the template M and the existing DSM data.

[0104] S205 , obtaining the maximum and minimum values ​​of the DSM data in the overlapping region R.

[0105] S206. Subtract the minimum value from the maximum value to obtain the height of each building.

[0106] Please refer to Figure 3 , which is a comparison diagram of the building height before and after effects provided in the embodiment of the present application. The left picture is the original remote sensing image, the middle picture is the building template Mask result after recognition, and the right picture is the DSM picture.

[0107] In summary, the embodiment of the present application provides a method for predicting building height using remote sensing image DOM and DSM. The building template of DOM is extracted through a neural network model, and the overlapping area is calculated based on the building template and the existing DSM data. The maximum value of the DSM data in the overlapping area is subtracted from the minimum value to obtain the building height. This method is a link between the entire 3D reconstruction technology, which can restore the building outline and building height more realistically, and improves the accuracy of outline restoration by 80% compared with the existing random building method.

[0108] Based on the same inventive concept, Figure 4 As shown, the embodiment of the present application also provides a device for predicting building height using remote sensing images DOM and DSM, including:

[0109] Acquisition module, used to obtain remote sensing image DOM and digital surface model DSM data of the target area;

[0110] An extraction module is used to extract the target outline of the target building in the target area according to the DOM;

[0111] A calculation module, used to calculate the overlap area between the target contour and the DSM data;

[0112] The determination module is used to determine the maximum and minimum values ​​of the DSM data in the overlapped area, and determine the difference between the maximum and minimum values ​​as the height of the target building.

[0113] Optionally, the acquisition module is specifically used to:

[0114] Get the initial DOM and initial DSM data of the target area;

[0115] ArcGIS software was used to perform radiation correction and geometric correction on the initial DOM to obtain the corrected DOM;

[0116] The corrected DOM and initial DSM data are registered to obtain the remote sensing image DOM and digital surface model DSM data of the target area.

[0117] Optionally, the extraction module is specifically used to:

[0118] The DOM is input into the trained neural network model, and the target outline of the target building in the target area is output; wherein the trained neural network model is obtained by training the initial neural network model based on the building dataset; the building dataset includes images of multiple sample buildings and the true outline and true label of each sample building.

[0119] Optionally, the device further comprises a training module, wherein the training module is used for:

[0120] Before the DOM is input into the trained neural network model and the target contour of the target building in the target area is output, a total loss function of the initial neural network model is established based on the contour loss and the classification loss;

[0121] Input the image of each sample building into the initial neural network model, and output the predicted outline and predicted category of each sample building;

[0122] The predicted outline, true outline, predicted label and true label of each sample building are substituted into the total loss function. According to the value of the total loss function, the weight of the initial neural network model is updated until the value of the loss function is less than the preset error threshold or the maximum number of iterations is reached, and the trained neural network model is obtained.

[0123] Optionally, the device further comprises an establishing module, the establishing module being used for:

[0124] The contour loss and classification loss are weighted summed to establish the total loss function of the initial neural network model.

[0125] Optionally, the contour loss adopts Hausdorff loss, and the calculation formula of Hausdorff loss is as follows:

[0126]

[0127] Among them, P i represents the predicted outline of the i-th sample building, G i represents the true outline of the i-th sample building, N is the number of sample buildings, It represents the maximum distance from each point a in the predicted contour to the nearest point b in the true contour. It represents the maximum distance from each point b in the true contour to the nearest point a in the predicted contour.

[0128] Optionally, the classification loss is calculated as follows:

[0129]

[0130] Among them, y i is the true label of the i-th sample building, is the predicted probability of the initial neural network model for the i-th sample building, and N is the number of sample buildings.

[0131] It should be noted that the modules in the device for predicting building height using remote sensing images DOM and DSM in this embodiment correspond one-to-one to the steps in the method for predicting building height using remote sensing images DOM and DSM in the aforementioned embodiment. Therefore, the specific implementation of this embodiment can refer to the implementation of the method for predicting building height using remote sensing images DOM and DSM in the aforementioned embodiment, which will not be repeated here.

[0132] In addition, in one embodiment, the present application also provides a computer device, which includes a processor, a memory, and a computer program stored in the memory. When the computer program is executed by the processor, the aforementioned method of predicting building height using remote sensing image DOM and DSM is implemented.

[0133] In addition, in one embodiment, the present application also provides a computer storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the aforementioned method of predicting building height using remote sensing image DOM and DSM is implemented.

[0134] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or system. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or system including the element.

[0135] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0136] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as a read-only memory / random access memory, a magnetic disk, or an optical disk), and includes a number of instructions for enabling a multimedia terminal device (which can be a mobile phone, a computer, a television receiver, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0137] The above are only preferred embodiments of the present application, and are not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A method for predicting building height using remote sensing image DOM and DSM. It is characterized in that include: Obtain remote sensing images DOM and digital surface model DSM data of the target area; Extracting a target outline of a target building in the target area according to the DOM; Calculating the overlap area between the target contour and the DSM data; The maximum value and the minimum value of the DSM data in the overlapped area are determined, and the difference between the maximum value and the minimum value is determined as the height of the target building.

2. The method for predicting building height using remote sensing images DOM and DSM as claimed in claim 1, It is characterized in that The step of obtaining the remote sensing image DOM and digital surface model DSM data of the target area includes: Get the initial DOM and initial DSM data of the target area; Using ArcGIS software to perform radiation correction and geometric correction on the initial DOM to obtain a corrected DOM; The corrected DOM and the initial DSM data are registered to obtain the remote sensing image DOM and digital surface model DSM data of the target area.

3. The method for predicting building height using remote sensing images DOM and DSM as claimed in claim 1, It is characterized in that Extracting the building outline of the target building in the target area according to the DOM includes: The DOM is input into a trained neural network model, and a target outline of a target building in the target area is output; wherein the trained neural network model is obtained by training an initial neural network model based on a building dataset; the building dataset includes images of multiple sample buildings and a true outline and a true label of each sample building.

4. The method for predicting building height using remote sensing images DOM and DSM as claimed in claim 3, It is characterized in that Before inputting the DOM into the trained neural network model and outputting the target outline of the target building in the target area, the method further includes: Based on the contour loss and classification loss, the total loss function of the initial neural network model is established; Inputting the image of each sample building into the initial neural network model, and outputting the predicted outline and predicted category of each sample building; The predicted outline, true outline, predicted label and true label of each sample building are substituted into the total loss function, and the weight of the initial neural network model is updated according to the value of the total loss function until the value of the loss function is less than a preset error threshold or the maximum number of iterations is reached, thereby obtaining a trained neural network model.

5. The method for predicting building height using remote sensing images DOM and DSM as claimed in claim 4, It is characterized in that The total loss function of the initial neural network model is established based on the contour loss and the classification loss, including: The contour loss and classification loss are weighted summed to establish the total loss function of the initial neural network model.

6. The method for predicting building height using remote sensing images DOM and DSM as claimed in claim 5, It is characterized in that The contour loss adopts Hausdorff loss, and the calculation formula of the Hausdorff loss is as follows: Among them, P i represents the predicted outline of the i-th sample building, G i represents the true outline of the i-th sample building, N is the number of sample buildings, It represents the maximum distance from each point a in the predicted contour to the nearest point b in the true contour. It represents the maximum distance from each point b in the true contour to the nearest point a in the predicted contour.

7. The method for predicting building height using remote sensing images DOM and DSM as claimed in claim 5, It is characterized in that The calculation formula of the classification loss is as follows: Among them, y i is the true label of the i-th sample building, is the predicted probability of the initial neural network model for the i-th sample building, and N is the number of sample buildings.

8. A device for predicting building height using remote sensing images DOM and DSM. It is characterized in that The device comprises: Acquisition module, used to obtain remote sensing image DOM and digital surface model DSM data of the target area; An extraction module, used for extracting a target outline of a target building in the target area according to the DOM; A calculation module, used for calculating the overlap area between the target contour and the DSM data; The determination module is used to determine the maximum value and the minimum value of the DSM data in the overlapped area, and determine the difference between the maximum value and the minimum value as the height of the target building.

9. A computer device, It is characterized in that The computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method for predicting building height using remote sensing image DOM and DSM according to any one of claims 1 to 7.

10. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores a computer program, and the processor executes the computer program to implement the method for predicting building height using remote sensing image DOM and DSM as described in any one of claims 1-7.

Images