Urban forest extraction method

Abstract

The invention provides an urban forest extraction method. The method comprises the following steps: step 1, preprocessing data; step 2, finding a city boundary in the data image, then selecting a forest sample plot in the city boundary, segmenting the data into image blocks with the same size, and forming a data set I by all the image blocks; each image block comprises an urban forest land parcel and a non-urban forest land parcel; step 3, creating a label image corresponding to each image block through visual interpretation, wherein the label image comprises a polygonal region of the urban forest land parcel and a polygonal region of a non-urban forest land parcel; all label images form a data set II; 4, establishing an artificial intelligence model; 5, training an artificial intelligence model based on the data set I and the data set II; step 6, obtaining an urban forest plot probability graph; and step 7, obtaining the area occupied by the urban forest land parcel. According to the invention, the precision of urban forest extraction is improved.

Description

Technical Field

[0001] The invention belongs to the technical field of ecological environment monitoring, and in particular relates to an urban forest extraction method. Background Art

[0002] The world, especially China, is experiencing rapid urbanization. As an integral part of cities, urban forests are constantly being re-evaluated in terms of their definition, characteristics, functions, value, and significance. Urban forests currently provide at least the following ecosystem services: alleviating summer heat in cities; intercepting and absorbing water to improve urban precipitation and surface runoff; absorbing air pollutants to improve air quality; converting greenhouse gases into dry matter; and providing aesthetic and spiritual value. Therefore, timely, accurate, and comprehensive analysis of urban forests is crucial for the environmental assessment and management of urban green spaces. Improving the quality of urban forest analysis can help enhance a city's image and improve the quality of life for its citizens.

[0003] Currently, there are two main methods for extracting urban forests: field surveys and statistics, and extraction from satellite remote sensing data. The former, derived from traditional forestry survey methods, uses forest stands as the field survey area to capture forest types, locations, and parameters within that area. The data obtained from this method is typically stored in a tabular format, which most closely matches actual operational requirements. However, this method is time-consuming and labor-intensive, with a limited survey scope and a long cycle. The latter, based on satellite remote sensing data, uses statistical and machine learning methods to extract urban forests using pixels as the basic unit. Currently, satellite remote sensing data is experiencing an explosive growth in production, with development progressing towards high spatial, temporal, and spectral resolution. Although medium-resolution remote sensing data existed in the past, its resolution was insufficient for identifying forest resources in highly spatially heterogeneous urban environments, resulting in numerous under-identifications and misidentifications. Therefore, the accuracy of urban forest extraction methods based on satellite remote sensing data needs to be improved. Summary of the Invention

[0004] The purpose of this invention is to provide an urban forest extraction method that combines high-resolution multispectral remote sensing data with artificial intelligence technology to improve the accuracy of urban forest extraction. The technical solution adopted is: A method for extracting urban forests comprises the following steps: Step 1: Obtain high spatial resolution multispectral satellite remote sensing data of the city and preprocess the data; Step 2: Find the city boundary in the data, then select forest plots within the city boundary and split the data into image blocks of equal size. All image blocks constitute Dataset 1; each image block includes both urban forest plots and non-urban forest plots. Step 3: Create a labeled image corresponding to each image block through visual interpretation. The labeled image includes the polygonal area of urban forest blocks and the polygonal area of non-urban forest blocks. All labeled images constitute Dataset 2. Among them, the label values of the polygonal areas of urban forest plots and non-urban forest plots are different; Step 4: Build an artificial intelligence model; Step 5: Train the artificial intelligence model based on dataset 1 and dataset 2; Step 6: Input the image blocks formed after preprocessing the high-spatial-resolution multispectral satellite remote sensing data of the city into the trained artificial intelligence model to obtain the probability map of urban forest plots. If the probability is greater than the set threshold, the spatial distribution result of the urban forest can be obtained; Step 7: Count the number of pixels in the forest plots and, based on the actual spatial size of the pixels, obtain the area occupied by the urban forest plots.

[0005] Preferably, the acquired high spatial resolution multispectral satellite remote sensing data has a 0.5-meter resolution, which is the spatial resolution corresponding to a satellite shooting altitude of 535 kilometers from the earth's surface, and includes four bands: blue, green, red, and near-infrared.

[0006] Preferably, the image block size in step 2 is 1024 pixels × 1024 pixels, the actual ground range corresponding to each image block is 512 meters × 512 meters, and 500 image blocks are randomly generated in the entire data.

[0007] Preferably, step 3 specifically includes the following steps: for each image segmentation, professionals identify urban forest plots, outline and draw the vector polygon boundaries of the plots, and convert the vector polygons of the plots into labeled images corresponding to the image resolution, wherein the label value of non-urban forest plots is 0, and the label value of urban forest plots is 1.

[0008] Preferably, the programming language used in step 4 is Python 3.11, and the artificial intelligence model is written using the PyTorch library; the overall framework of the model is selected as U-Net to achieve multi-level image feature extraction.

[0009] Preferably, a wavelet transform module is added to the model with U-Net as the overall framework to extract image texture features.

[0010] Preferably, the parameters of model training are set as follows: learning rate 0.0001, number of training rounds 50, number of input channels 4, number of output channels 1, batch size 16, number of processes 8; Adam is selected as the model optimizer, and the two decay factors are 0.5 and 0.999 respectively.

[0011] Preferably, the actual space size corresponding to the pixel in step 7 is 0.5 meters × 0.5 meters.

[0012] Compared with the prior art, the advantages of the present invention are: The urban forest extraction method proposed in this paper combines high-resolution multispectral remote sensing data with artificial intelligence technology. Using multispectral data from the four most common bands in remote sensing imagery, the method achieves rapid, accurate, and comprehensive extraction of urban forest spatial distribution at a high spatial resolution of 0.5 meters, using a lightweight artificial intelligence model. This overcomes the time-consuming and labor-intensive, low-precision, and incomplete overall results of previous urban forest extraction methods.

[0013] 2. The method is highly reusable and can meet the future demand for extracting urban forests from high-spatial-resolution satellite multispectral data. The highly precise urban forest extraction results can help reveal the spatial distribution characteristics of urban forests, the interaction between urban forests and other urban functional land uses, monitor the interannual changes in urban forests, and provide high-quality basic data support for subsequent quantitative research on urban forests. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 Schematic diagram of image segmentation and label images used for model training; Figure 2 This is a schematic diagram of the artificial intelligence model architecture; Figure 3 Schematic diagram of the distribution of urban forest plots in an extracted image segment. DETAILED DESCRIPTION

[0015] The following is a more detailed description of an urban forest extraction method according to the present invention, with reference to schematic diagrams. These diagrams illustrate preferred embodiments of the present invention. It should be understood that those skilled in the art may modify the present invention described herein while still achieving its advantageous effects. Therefore, the following description should be understood as a general guideline for those skilled in the art and not as a limitation of the present invention.

[0016] A method for extracting urban forests comprises the following steps: Step 1: Obtain high spatial resolution multispectral satellite remote sensing data of the city and preprocess the data; The high-spatial-resolution multispectral satellite remote sensing data obtained has a 0.5-meter resolution, which corresponds to the spatial resolution when the satellite is shot at an altitude of 535 kilometers from the earth's surface. It includes four bands: blue, green, red, and near-infrared.

[0017] Preprocessing includes radiometric calibration, atmospheric correction, terrain correction, geometric correction, mosaicking and stitching of multispectral images.

[0018] Step 2: Find the city boundary in the data, then select forest plots within the city boundary and divide the data into image blocks of the same size. All image blocks constitute Dataset 1; each image block includes urban forest plots and non-urban forest plots.

[0019] In this embodiment, the image block size is 1024 pixels×1024 pixels, the actual ground range corresponding to each image block is 512 meters×512 meters, and 500 image blocks are randomly generated in the entire data.

[0020] Step 3: Create a labeled image corresponding to each image block through visual interpretation. The labeled image includes the polygonal area of urban forest blocks and the polygonal area of non-urban forest blocks. All labeled images constitute Dataset 2. Among them, the label values of the polygonal areas of urban forest plots and the polygonal areas of non-urban forest plots are different.

[0021] The specific steps include: For each image segment, professionals identify urban forest plots, outline and draw the vector polygon boundaries of the plots, and convert the vector polygons of the plots into labeled images corresponding to the image resolution, where the label value of non-urban forest plots is 0 and the label value of urban forest plots is 1.

[0022] Among them, image segmentation and label image are as follows: Figure 1 shown.

[0023] Step 4: Build an artificial intelligence model.

[0024] like Figure 2 As shown in the figure, the programming language used is Python 3.11, and the artificial intelligence model is written using the PyTorch library; the overall framework of the model is selected as U-Net to achieve multi-level image feature extraction.

[0025] A wavelet transform module is added to the model with U-Net as the overall framework to extract image texture features.

[0026] Step 5: Based on Dataset 1 and Dataset 2, train the artificial intelligence model to obtain the required model parameters.

[0027] The parameters for model training are set as follows: learning rate 0.0001, number of training rounds 50, number of input channels 4, number of output channels 1, batch size 16, and number of processes 8; Adam is selected as the model optimizer, and the two decay factors are 0.5 and 0.999 respectively.

[0028] Step 6: Input the image blocks formed after preprocessing the urban high spatial resolution multispectral satellite remote sensing data into the trained intelligent model, and input the image blocks formed after preprocessing the urban high spatial resolution multispectral satellite remote sensing data into the trained artificial intelligence model to obtain the urban forest plot probability map. If it is greater than the set threshold (such as 0.5), it is judged to be an urban forest plot, and then the urban forest spatial distribution result is obtained, such as Figure 3 shown, otherwise it is a non-urban forest plot.

[0029] The image blocks can be cropped into a size of 256 pixels × 256 pixels to fit the memory size of the graphics card device.

[0030] Step 7: Count the number of pixels in the forest plots and, based on the actual spatial size of the pixels, obtain the area occupied by the urban forest plots.

[0031] Among them, the actual spatial size corresponding to a pixel is the same as the spatial resolution of satellite remote sensing data, which is 0.5 meters × 0.5 meters.

[0032] The above description is merely a preferred embodiment of the present invention and does not limit the present invention in any way. Any person skilled in the art who, without departing from the scope of the present invention, makes any equivalent substitution, modification, or other changes to the technical solution and technical content disclosed in the present invention shall be deemed to be within the scope of the present invention and still fall within the scope of protection of the present invention.

Claims

1. A method for extracting urban forests, characterized in that: The following steps are involved: Step 1: Obtain high spatial resolution multispectral satellite remote sensing data of the city and preprocess the data; Step 2: Find the city boundary in the data, then select forest plots within the city boundary and split the data into image blocks of equal size. All image blocks constitute Dataset 1; each image block includes both urban forest plots and non-urban forest plots. Step 3: Create a labeled image corresponding to each image block through visual interpretation. The labeled image includes the polygonal area of urban forest blocks and the polygonal area of non-urban forest blocks. All labeled images constitute Dataset 2. Among them, the label values of the polygonal areas of urban forest plots and non-urban forest plots are different; Step 4: Build an artificial intelligence model; Step 5: Train the artificial intelligence model based on dataset 1 and dataset 2; Step 6: Input the image blocks formed after preprocessing the high-spatial-resolution multispectral satellite remote sensing data of the city into the trained artificial intelligence model to obtain the probability map of urban forest plots. If the probability is greater than the set threshold, the spatial distribution result of the urban forest can be obtained; Step 7: Count the number of pixels in the forest plots and, based on the actual spatial size of the pixels, obtain the area occupied by the urban forest plots.

2. The urban forest extraction method according to claim 1, characterized in that The high-spatial-resolution multispectral satellite remote sensing data obtained has a 0.5-meter resolution, which corresponds to the spatial resolution when the satellite is shot at an altitude of 535 kilometers from the earth's surface. It includes four bands: blue, green, red, and near-infrared.

3. The urban forest extraction method according to claim 1, characterized in that: The image block size in step 2 is 1024 pixels × 1024 pixels. The actual ground area corresponding to each image block is 512 meters × 512 meters. 500 image blocks are randomly generated in the entire data.

4. The urban forest extraction method according to claim 1, characterized in that: Step 3 specifically includes the following steps: for each image segment, professionals identify urban forest plots, outline and draw the vector polygon boundaries of the plots, and convert the vector polygons of the plots into labeled images corresponding to the image resolution, where the label value of non-urban forest plots is 0 and the label value of urban forest plots is 1.

5. The urban forest extraction method according to claim 1, characterized in that: The programming language used in step 4 is Python 3.11, and the artificial intelligence model is written using the PyTorch library; the overall framework of the model is selected as U-Net to achieve multi-level image feature extraction.

6. The urban forest extraction method according to claim 5, characterized in that: A wavelet transform module is added to the model with U-Net as the overall framework to extract image texture features.

7. The urban forest extraction method according to claim 5, characterized in that: The parameters for model training are set as follows: learning rate 0.0001, number of training rounds 50, number of input channels 4, number of output channels 1, batch size 16, and number of processes 8; Adam is selected as the model optimizer, and the two decay factors are 0.5 and 0.999 respectively.

8. The urban forest extraction method according to claim 1, characterized in that: In step 7, the actual space size corresponding to the pixel is 0.5 meters × 0.5 meters.

Images