Method for measuring and calculating newly-added construction space of high-resolution satellite remote sensing data area based on wide-area high frequency

Abstract

The invention relates to a method for measuring and calculating a newly-added construction space of a high-resolution satellite remote sensing data area based on wide-area high frequency, and relates to the technical field of measurement and calculation of the newly-added construction space of the area, and the method comprises the following steps: obtaining a wide-area high-resolution satellite remote sensing image; extracting a construction area under construction based on tower crane identification; building area measurement and calculation based on instance segmentation; building height measurement and calculation based on shadow restoration; and measuring and calculating the regional newly-added construction space of the wide-area and large-range project under construction. By means of the method for measuring and calculating the newly-added construction space of the high-resolution satellite remote sensing data area based on wide-area high frequency, the construction increase situation of a single project within a period of time can be determined, meanwhile, the construction space increase situation within the area range can be counted, and the investment situation and construction intensity of all areas can be effectively measured.

Description

Technical Field

[0001] The present invention relates to the technical field of regional newly added construction space calculation, and in particular to a regional newly added construction space calculation method based on wide-area high-frequency high-resolution satellite remote sensing data. Background Art

[0002] There are many fixed asset investment projects, which are complex and widely distributed. Traditional statistical methods usually rely on manual reporting of the system. This method not only has high operating and maintenance costs, but is also limited by spatial factors, making it difficult to effectively monitor a large number of projects and facing the challenge of difficult supervision. In addition, there are projects that are missed in investment approval and projects that have not been handled according to procedures, resulting in a small amount of data on the estimated new construction space in the region, which cannot accurately reflect the actual construction situation and investment amount.

[0003] In order to strengthen the whole process supervision of civil engineering fixed asset investment projects in the investment field and reduce statistical fraud and falsification, the newly added construction space of the investment statistical projects that have been reported in the database is calculated through building extraction and building height shadow recognition technology. At the same time, combined with tower crane recognition technology, projects that are not included in the statistical scope are monitored and the newly added construction space of the projects is calculated to strengthen data verification, improve the quality of statistical data, support the whole process supervision of civil engineering fixed asset investment projects, comprehensively monitor and count various investment projects, and provide more reliable data support for investment decisions and planning.

[0004] The parameters used in calculating the construction space of a project are the building area and height within the construction project. Therefore, when interpreting remote sensing images, it is particularly important to identify the construction site, extract the building outline and accurately measure the height.

[0005] At present, the identification of tower cranes and construction sites, building outline extraction and methods are mainly focused on the use of high-resolution satellite remote sensing images and deep learning technology applications, while there are many methods for height measurement. The extraction of building height by multi-source data fusion effectively improves the accuracy of building height extraction, but the multi-source data fusion process is complex, involving the preprocessing and registration of multiple data, and the processing algorithm and computing resources are required to be high. The time synchronization problem of the data source may also affect the accuracy of height extraction. Using high-resolution stereo images, the height information of the building is obtained through parallax calculation and stereo matching technology. The acquisition of stereo images is affected by weather and lighting conditions, and the acquisition time and cost of the images are high. In addition, the stereo matching algorithm may have errors when processing complex building structures. Summary of the invention

[0006] The present invention aims to solve the technical problems in the prior art and provide a method for calculating the newly added construction space in a region based on wide-area, high-frequency, high-resolution satellite remote sensing data.

[0007] In order to solve the above technical problems, the technical solutions of the present invention are as follows:

[0008] A method for calculating the spatial calculation of newly added construction in a region based on wide-area, high-frequency, high-resolution satellite remote sensing data comprises the following steps:

[0009] Step S1: Acquire wide-area high-resolution satellite remote sensing images;

[0010] Step S2: Extracting the construction area under construction based on tower crane identification;

[0011] Step S3: Building area estimation based on instance segmentation;

[0012] Step S4: Calculate building height based on shadow restoration;

[0013] Step S5: Calculate the newly added construction space in the region for projects under construction in a wide area.

[0014] In the above technical solution, step S2 specifically includes the steps of:

[0015] Step S21: creating a tower crane recognition data set;

[0016] Using wide-area high-resolution satellite remote sensing images as data sources, a high-resolution crane recognition sample set is constructed;

[0017] Step S22: constructing a tower crane target recognition model;

[0018] Constructing a tower crane target recognition model based on YOLOv8; Step S23: Identifying the area of ​​the project under construction;

[0019] Generate a buffer zone with each target as the center; fuse the intersecting targets to get the target area; extract the centroid of each target area to get the central area point of the project under construction, and outline and confirm the scope of the area under construction based on the extraction results.

[0020] In the above technical solution, in step S22, the tower crane target recognition process includes:

[0021] (1) Model training: Use the prepared training data set to train the tower crane target recognition model;

[0022] (2) Tower crane target detection: The image to be detected is input into the tower crane target recognition model to perform tower crane target detection;

[0023] (3) Post-processing of detection results: Perform NMS processing on the merged detection results, filter out redundant detection boxes and output the final extraction results.

[0024] In the above technical solution, step S3 specifically includes the steps of:

[0025] Step S31: creating a building extraction data set;

[0026] Using wide-area high-resolution satellite remote sensing images as data sources, a high-resolution building extraction sample set is constructed;

[0027] Step S32: constructing a building extraction and recognition model;

[0028] For building contour recognition, we first perform image instance segmentation on all individual buildings in the project area, and distinguish the buildings in the segmentation results by contour extraction to distinguish the roofs of each building. We use the Mask2former model to perform segmentation and extraction on the buildings based on the prepared building sample set.

[0029] Step S33: building extraction post-processing;

[0030] The accuracy and regularity of building extraction are improved through point simplification, morphological expansion and erosion, and hole filling.

[0031] In the above technical solution, step S4 specifically includes the steps of:

[0032] Step S41: creating a building shadow detection data set;

[0033] Using wide-area high-resolution satellite remote sensing images as the data source, the satellite images are preprocessed and shadow targets are marked to construct a training dataset for identifying building shadows.

[0034] Step S42: constructing a shadow recognition model;

[0035] Building shadows and their lengths are automatically acquired through deep learning algorithms;

[0036] Step S43: Calculate the length and height of the building's shadow.

[0037] In the above technical solution, step S5 specifically includes the steps of:

[0038] According to the formula:

[0039]

[0040] Where n is the total number of buildings, h k is the height of the kth building, s k is the area of ​​the kth building, i is the monitoring period number, V i Construct space on the ground for the monitoring period;

[0041] Calculate the area of ​​the extracted building outline and multiply it by the measured building height to obtain the construction space of a single building;

[0042] The sum of the ground construction space of all buildings in the project area within the statistical range in the late image and the early image is calculated respectively, and the ground construction space of all buildings is calculated according to the construction project area identified by the tower crane;

[0043] The construction space of the project area is obtained by subtracting the construction space of the early image from the construction space of the later image.

[0044] The present invention has the following beneficial effects:

[0045] By utilizing the method for calculating newly added construction space in a region based on wide-area, high-frequency, and high-resolution satellite remote sensing data of the present invention, the construction increase of a single project over a period of time can be clarified, and the construction space increase within the region can be counted, which can effectively measure the investment situation and construction intensity of each region. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0047] Figure 1 The present invention is a schematic flow chart of the steps of the newly constructed spatial measurement method based on wide-area, high-frequency, and high-resolution satellite remote sensing data areas.

[0048] Figure 2 Schematic diagram of the relationship between the satellite, the same side of the sun and the buildings.

[0049] Figure 3 Schematic diagram of the relationship between satellites, the opposite side of the sun and buildings. DETAILED DESCRIPTION

[0050] The inventive concept of the present invention is: the method for calculating the newly added construction space in a region based on wide-area, high-frequency, high-resolution satellite remote sensing data of the present invention uses high-resolution remote sensing data and geographic information technology to calculate the newly added construction space in the region. The construction space of ground buildings refers to the sum of the products of the areas and heights of all buildings in a construction project. The newly added construction space in a region is a concept proposed for statistically analyzing the construction intensity of large-scale, wide-area investment projects. The calculation formula for the newly added regional construction space is the difference between the construction space of the later image project and the construction space of the ground buildings of the earlier image project in each construction project area.

[0051] The method for calculating newly added construction space in a region based on wide-area, high-frequency, high-resolution satellite remote sensing data of the present invention realizes the monitoring of newly added construction space of construction projects based on high spatial resolution remote sensing technology and deep learning technology, and extracts and predicts the outlines and heights of buildings in investment projects under construction based on a variety of deep learning models. This method can realize fast and accurate calculation of newly added construction space for investment projects in a wide range, and provide data reference and method support for policy formulation and research directions related to urban development.

[0052] The present invention is described in detail below with reference to the accompanying drawings.

[0053] The method for calculating the spatial addition of construction in a region based on wide-area, high-frequency, high-resolution satellite remote sensing data of the present invention is as follows: Figure 1 As shown, the following steps are included:

[0054] Step S1: Acquire wide-area high-resolution satellite remote sensing images.

[0055] Step S2: Extracting the construction area under construction based on tower crane identification;

[0056] Monitor the project area under construction over a wide area using crane models.

[0057] Step S3: Building area estimation based on instance segmentation;

[0058] The building area within the project area is estimated based on the instance segmentation model.

[0059] Step S4: Calculate building height based on shadow restoration;

[0060] Get the length of the building's shadow and calculate the building's height based on the shadow restoration formula.

[0061] Step S5: Calculation of additional construction space in the region for projects under construction in a wide area;

[0062] Complete the calculation of additional construction space in a wide-area area with projects under construction.

[0063] The method for calculating the newly added construction space in a region based on wide-area, high-frequency, high-resolution satellite remote sensing data of the present invention can be widely used in the monitoring and management of construction projects to improve the management efficiency and accuracy of construction projects. Specifically:

[0064] Step S1: Acquire wide-area high-resolution satellite remote sensing images.

[0065] First, we select image data covering a wide area over two periods of one quarter. The selected data comes from the Jilin-1 satellite constellation, which is composed of 108 remote sensing satellites. In order to extract buildings more clearly and obtain higher-precision height calculation results, data resolutions of 0.75m and 0.5m are used.

[0066] Step S2: Extraction of construction areas under construction based on tower crane recognition.

[0067] The specific steps include:

[0068] Step S21: Create a tower crane recognition data set.

[0069] The project uses Jilin-1 high-resolution remote sensing data as the data source and constructs a high-resolution crane recognition sample set of more than 20,000 images. The satellite images are preprocessed and cropped into uniform-sized images of 512×512, and the image is randomly translated and rotated to expand the sample set capacity. Mosaic data enhancement is applied to randomly crop, scale, arrange and splice four random images, which can enrich the data set, improve the model's recognition ability for targets of different sizes, and enhance the robustness of the model.

[0070] Step S22: Constructing a tower crane target recognition model

[0071] A tower crane target recognition model based on YOLOv8 was constructed. The network architecture mainly consists of four parts: input, backbone network, neck network, and head output layer. By optimizing the network architecture, a deep learning model for tower crane targets was constructed and trained to achieve accurate recognition of tower crane targets. The tower crane target recognition process is divided into three parts: model training, tower crane target detection, and post-processing of detection results:

[0072] (1) Model training: Use the prepared training data set to train the tower crane target recognition model;

[0073] (2) Tower crane target detection: The image to be detected is input into the tower crane target recognition model in the form of sliding window slices to perform tower crane target detection;

[0074] (3) Post-processing of detection results: Perform NMS processing on the merged detection results, filter out redundant detection boxes and output the final extraction results.

[0075] Step S23: Identify the area of ​​the project under construction.

[0076] After obtaining the tower crane target frame of the missed statistical area, a buffer zone with a radius of 100m is generated with each target as the center; the intersecting targets are fused to obtain the target area; the centroid of each target area is extracted to obtain the central area point of the project under construction, and finally the scope of the area under construction is outlined and confirmed based on the extraction results.

[0077] Step S3: Building area estimation based on instance segmentation.

[0078] The specific steps include:

[0079] Step S31: Create a building extraction data set.

[0080] The project uses Jilin-1 high-resolution remote sensing data as the data source and constructs a high-resolution building extraction sample set of more than 15,000 images. By preprocessing satellite images, the data set size is unified to 512×512. In order to improve the performance of Mask2former, it is necessary to perform corresponding data enhancement before model training, and add the Copy-Paste method to the data enhancement process before model training. This method combines information from multiple images in a target object-aware manner, copies target object instances from one image and pastes them to another image. The core of the method includes two aspects: mixed pasting of target objects and large-scale jittering.

[0081] Step S32: construct a building extraction and recognition model.

[0082] For building contour recognition, all individual buildings in the project area are first segmented by image instance, and the buildings in the segmentation results are distinguished by contour extraction to distinguish the roof of each building. The algorithm for image instance segmentation in the present invention adopts the Mask2former model, which is trained based on the prepared building sample set to segment and extract the buildings.

[0083] Step S33: Post-processing of building extraction.

[0084] After the preliminary extraction of buildings, a large amount of point data will be generated. Too much point data will not only increase the computational burden, but may also contain redundant information. Due to interference factors such as image noise, the local boundaries of buildings will have irregularities such as burrs and depressions. Point simplification, morphological expansion and corrosion, and hole filling can be used to further improve the accuracy and regularity of building extraction.

[0085] Step S4: Calculate building height based on shadow restoration.

[0086] The specific steps include:

[0087] Step S41: Create a building shadow detection dataset.

[0088] The project uses Jilin-1 high-resolution remote sensing data as the data source. By preprocessing the satellite images, the data set size is unified to 512×512, and the shadow targets are marked to construct a training data set for identifying building shadows.

[0089] Step S42: construct a shadow recognition model.

[0090] The shadow of the building and its length are automatically obtained through a deep learning algorithm. The algorithm uses the RetinaNet network structure as the basic network, and changes the number of feature fusion layers and modifies the loss function to enable it to have superior building shadow recognition capabilities. The production data set is input into the trained RetinaNet model to extract the shadows in the remote sensing image. The pixel coordinates of the diagonal points of the bounding box in the shadow detection result are used to calculate the pixel length of the shadow, and then the pixel length of the shadow is converted into the actual length according to the actual distance corresponding to each pixel in the image. Using the target recognition network based on optimized YOLOv8, the Transformer module, CBAM module, SPP module and loss function are optimized to build a target recognition model for tower cranes and buildings under construction.

[0091] Step S43: Calculate the length and height of the building's shadow.

[0092] When the sun and the satellite are on the same side of the building, the building height calculation formula is constructed based on the geometric relationship between the sun altitude angle, satellite altitude angle, building shadow and building height. H is the building height, S is the actual shadow length, α is the satellite altitude angle, β is the sun altitude angle, A is the visible length of the partial shadow, and B is the shadow-blocked part. When the satellite and the sun are on the same side, the shadow cannot be fully displayed on the image. The shadow length is A. Figure 2 In the geometric relationship, the building height H is:

[0093]

[0094] H=Atanωtanθ / (tanθ-tanω)Formula 2

[0095] The satellite is on the opposite side of the sun. Figure 3 The shadow length S can be fully presented in the figure. According to the geometric relationship, the building height H is:

[0096] H=StanωFormula 3

[0097] Step S5: Calculate the newly added construction space in the region for projects under construction in a wide area.

[0098]

[0099] Where n is the total number of buildings, h k is the height of the kth building, s k is the area of ​​the kth building, i is the monitoring period number, V i Construct space on the ground for the monitoring period.

[0100] Calculate the area of ​​the extracted building outline and multiply it by the measured building height to get the construction space of a single building;

[0101] The sum of the ground construction space of all buildings in the project area within the statistical range in the late image and the early image is calculated respectively, and the ground construction space of all buildings is calculated according to the construction project area identified by the tower crane;

[0102] The construction space of the project area is obtained by subtracting the construction space of the early image from the construction space of the later image.

[0103] By utilizing the method for calculating newly added construction space in a region based on wide-area, high-frequency, and high-resolution satellite remote sensing data of the present invention, the construction increase of a single project over a period of time can be clarified, and the construction space increase within the region can be counted, which can effectively measure the investment situation and construction intensity of each region.

[0104] Obviously, the above embodiments are merely examples for the purpose of clear explanation, and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived therefrom are still within the scope of protection of the invention.

Claims

1. A method for calculating the spatial calculation of newly constructed areas based on wide-area, high-frequency, high-resolution satellite remote sensing data, characterized in that: The following steps are involved: Step S1: Acquire wide-area high-resolution satellite remote sensing images; Step S2: Extracting the construction area under construction based on tower crane identification; Step S3: Building area estimation based on instance segmentation; Step S4: Calculate building height based on shadow restoration; Step S5: Calculate the newly added construction space in the region for projects under construction in a wide area.

2. The method for calculating the spatial distribution of newly constructed buildings in a region based on wide-area, high-frequency, high-resolution satellite remote sensing data according to claim 1 is characterized in that: Step S2 specifically includes the following steps: Step S21: creating a tower crane recognition data set; Using wide-area high-resolution satellite remote sensing images as data sources, a high-resolution crane recognition sample set is constructed; Step S22: constructing a tower crane target recognition model; Build a tower crane target recognition model based on YOLOv8; Step S23: Identification of the area of ​​the project under construction; Generate a buffer zone with each target as the center; fuse the intersecting targets to get the target area; extract the centroid of each target area to get the central area point of the project under construction, and outline and confirm the scope of the area under construction based on the extraction results.

3. The method for calculating the spatial distribution of newly constructed buildings in a region based on wide-area, high-frequency, high-resolution satellite remote sensing data according to claim 2 is characterized in that: In step S22, the tower crane target recognition process includes: (1) Model training: Use the prepared training data set to train the tower crane target recognition model; (2) Tower crane target detection: The image to be detected is input into the tower crane target recognition model to perform tower crane target detection; (3) Post-processing of detection results: Perform NMS processing on the merged detection results, filter out redundant detection boxes and output the final extraction results.

4. The method for calculating the spatial distribution of newly constructed buildings in a region based on wide-area, high-frequency, high-resolution satellite remote sensing data according to claim 1, characterized in that: Step S3 specifically includes the following steps: Step S31: creating a building extraction data set; Using wide-area high-resolution satellite remote sensing images as data sources, a high-resolution building extraction sample set is constructed; Step S32: constructing a building extraction and recognition model; For building contour recognition, all individual buildings in the project area are first segmented, and the buildings in the segmentation results are distinguished by contour extraction to distinguish the roof of each building; Use the Mask2former model to train based on the prepared building sample set to segment and extract the buildings; Step S33: building extraction post-processing; The accuracy and regularity of building extraction are improved through point simplification, morphological expansion and erosion, and hole filling.

5. The method for calculating the spatial distribution of newly constructed buildings in a region based on wide-area, high-frequency, high-resolution satellite remote sensing data according to claim 1, characterized in that: Step S4 specifically includes the following steps: Step S41: creating a building shadow detection data set; Using wide-area high-resolution satellite remote sensing images as the data source, the satellite images are preprocessed and shadow targets are marked to construct a training dataset for identifying building shadows. Step S42: constructing a shadow recognition model; Building shadows and their lengths are automatically acquired through deep learning algorithms; Step S43: Calculate the length and height of the building's shadow.

6. The method for calculating the spatial addition of construction in a region based on wide-area, high-frequency, high-resolution satellite remote sensing data according to claim 1, characterized in that: Step S5 specifically includes the following steps: According to the formula: Where n is the total number of buildings, h k is the height of the kth building, s k is the area of ​​the kth building, i is the monitoring period number, V i Construct space on the ground for the monitoring period; Calculate the area of ​​the extracted building outline and multiply it by the measured building height to obtain the construction space of a single building; The sum of the ground construction space of all buildings in the project area within the statistical range in the late image and the early image is calculated respectively, and the ground construction space of all buildings is calculated according to the construction project area identified by the tower crane; The construction space of the project area is obtained by subtracting the construction space of the early image from the construction space of the later image.

Images