Land space planning data enhancement method based on machine learning

Abstract

The invention relates to a territorial space planning data enhancement method based on machine learning, and belongs to the field of territorial space planning. The method comprises the following steps: data acquisition and processing: acquiring remote sensing image data from a plurality of data sources, and generating a low-resolution (LR) image pair and a high-resolution (HR) image pair; multi-scale feature extraction: extracting features of different scales, and outputting a multi-scale convolution feature image set; performing multi-scale feature fusion: performing weighted average fusion on the feature image set, and obtaining a fused feature image through residual connection; image enhancement: performing up-sampling of the feature map through a sub-pixel convolution layer, redistributing channel information to a spatial dimension, and obtaining an enhanced image through residual connection; training and optimizing the model: training the model by using the paired data set, fusing the mean square error loss and the perception loss by using a loss function, and optimizing the model by using an Adam optimizer. According to the method, more comprehensive data support is provided for space-time dynamic analysis in territorial space planning through data enhancement.

Description

Technical Field

[0001] The present invention belongs to the field of territorial spatial planning, and particularly relates to a method for enhancing territorial spatial planning data based on machine learning. Background Art

[0002] In modern society, territorial spatial planning is an important basis for the government and relevant institutions to make decisions in fields such as land use, urban development, and ecological protection. With the development of remote sensing technology, remote sensing images have become a key means to obtain geographical information in territorial spatial planning. However, due to the limitations of imaging equipment, transmission conditions, and natural factors (such as atmospheric conditions, lighting conditions, etc.), remote sensing images usually have problems such as low resolution and blurred details, making it difficult to meet the requirements of refined territorial spatial planning. Traditional image enhancement methods, such as interpolation algorithms or filtering techniques, although can improve the visual quality of images to a certain extent, often accompanied by information loss or increased artifacts, and cannot effectively retain the detailed information in the images. This makes the reliability of the enhanced image data poor in practical applications, especially in fields that require precise analysis such as urban planning and land management, affecting the accuracy and scientificity of planning. In recent years, machine learning, especially deep learning technology, has made remarkable progress in the field of computer vision. By constructing complex neural network models, the feature representations of images can be learned from massive data and applied to tasks such as image super-resolution and image restoration. Compared with traditional methods, the image enhancement technology based on machine learning shows superior performance in improving image resolution and enhancing the ability to retain image details. Based on this background, a method for enhancing territorial spatial planning data based on machine learning is proposed to solve the problems of insufficient resolution and missing details in existing remote sensing images. The present invention aims to provide higher-quality and more reliable remote sensing image data support for territorial spatial planning, improve the accuracy and scientificity of planning, and provide technical guarantees for the rational utilization and management of land resources by introducing multi-scale feature extraction and fusion technology, combined with advanced methods such as residual connection and sub-pixel convolution. Summary of the Invention

[0003] The present invention provides a method for enhancing remote sensing images in territorial spatial planning data based on machine learning, and its main purpose is to achieve image enhancement of remote sensing images and better provide data support for territorial spatial planning.

[0004] To achieve the above object, the present invention is realized through the following technical solutions: In the first aspect, the present invention provides a method for enhancing territorial spatial planning data based on machine learning, including the following steps: S1. Data acquisition and processing: Obtain remote sensing image data from multiple data sources to generate a pair of low-resolution LR and high-resolution HR images; S2. Construct a machine learning-based data augmentation model for territorial space planning, including a multi-scale feature extraction module, a multi-scale feature fusion module, and an image enhancement module; S3. Multi-scale feature extraction: Input an image, use convolutional kernels of different sizes to extract features of different scales, and output a multi-scale convolutional feature map set. The input image is a low-resolution LR image; S4. Multi-scale feature fusion: Perform weighted average fusion on the feature maps of different scales in the multi-scale convolutional feature map set, and obtain a fused feature map through residual connection; S5. Image enhancement: The fused feature map is upsampled through a sub-pixel convolutional layer to redistribute the channel information to the spatial dimension, and an enhanced image is obtained through residual connection; S6. Model training and optimization: Use a paired data set to train the model. The loss function combines the mean squared error loss and the perceptual loss, and uses the Adam optimizer to optimize the model to obtain the final machine learning-based data augmentation model for territorial space planning.

[0005] Furthermore, the specific steps of S1 include: S11. Data acquisition: The multiple data sources are denoted as , where n represents the number of data sources. The image data collected from the data source is denoted as ; S12. Data processing: Screen high-resolution images that meet the resolution requirements of territorial space planning from the image data , use downsampling technology to generate low-resolution images , and construct low-resolution LR and high-resolution HR image pairs , which is expressed by the formula as follows where represents the downsampling operation, and represents the scaling factor.

[0006] Furthermore, the specific multi-scale feature extraction in step S3 includes: Perform a convolution operation on the input image . The size of the image is , represents the height, represents the width, and the number of input channels is . Use multiple convolutional kernel sizes to extract features of different scales and obtain a multi-scale convolutional feature map set . The calculation formula is as follows: , Among them, represents a convolution operation, represents function, represents a convolution kernel, represents the corresponding bias, represents the feature map output by the i-th convolution kernel, , is the height of the feature map after the convolution operation, is the width of the feature map after the convolution operation, is the number of output channels of the feature map after the convolution operation. The three convolution kernel sizes are: , which respectively represent convolution kernels of 3x3, 5x5, and 7x7.

[0007] Furthermore, the specific steps of step S4 include: The multi-scale convolution feature map set fuses multi-scale features through weighted averaging, and then links the features obtained after weighted averaging with the input image to obtain a fused feature map , and the calculation formula is as follows: , , Among them, is the dot product operation, represents the feature obtained after weighted averaging, , represents the weight of the feature of the th scale, represents the number of multi-scale features.

[0008] Furthermore, the specific steps of step S5 include: The fused feature map performs a convolution operation through a preprocessing convolution layer to obtain a preprocessed feature map , and the preprocessed feature map passes through a sub-pixel convolution layer to expand the number of channels to to obtain an expanded feature map , and the expanded feature map redistributes the channel information to the spatial dimension through sub-pixel rearrangement to obtain a high-resolution feature map with a size of , and then the high-resolution feature map and the upsampled input image are linked by residuals to obtain an enhanced image , and the calculation process formula is as follows: ,​ , , , Among them, represents the convolution kernel of the preprocessing convolutional layer, represents the bias of the preprocessing convolutional layer, represents the convolution kernel of the sub-pixel convolutional layer, represents the bias of the sub-pixel convolutional layer, represents the operation of rearranging the channel information into the spatial dimension to achieve upsampling, represents the upsampling factor, represents the bilinear interpolation upsampling operation.

[0009] Furthermore, the specific steps of step S6 include: S51. Training data: The training set includes a paired data set of the input image and the corresponding labeled image . The input image is a low-resolution LR image , and the labeled image is a high-resolution HR image . The training set contains pairs of images denoted as , where j represents the number of image pairs; S52. Loss function: The mean square error loss is used to measure the pixel difference between the model output image and the labeled image . The output image is the enhanced image . The mean square error loss formula is expressed as follows: , The formula for the perceptual loss is expressed as follows: , Among them, represents the number of mini-batch samples in the training process, represents the feature map of the th layer, represents the selected number of layers, represents the th layer, represents the square of the L2 norm. The final loss function is a weighted combination of the above losses, and the formula is expressed as follows: , Among them, is the hyperparameter corresponding to ​ For the corresponding hyperparameters; S53. Model Optimization: The Adam optimizer is used to minimize the loss function. For the model parameters , the update rule formula for each time is as follows: , where is the learning rate, is the loss function with respect to the parameter gradient, and the cosine annealing learning rate decay strategy is used to control the learning rate during the training process. The formula is as follows: where is the number of iterations during the training process, is the initial learning rate, is the total number of training steps.

[0010] In a second aspect, a machine learning-based land spatial planning data enhancement device includes: Data Acquisition and Processing Unit: Used to obtain remote sensing image data from multiple data sources and generate low-resolution LR and high-resolution HR image pairs; Model Construction Unit: Used to construct a machine learning-based land spatial planning data enhancement model, including a multi-scale feature extraction module, a multi-scale feature fusion module, and an image enhancement module; Multi-scale Feature Extraction Unit: Used to input an image, extract features of different scales using convolution kernels of different sizes, and output a multi-scale convolutional feature map set. The input image is the low-resolution LR image; Multi-scale Feature Fusion Unit: Used to perform weighted average fusion on the feature maps of different scales in the multi-scale convolutional feature map set and obtain a fused feature map through residual connection; Image Enhancement Unit: Used to perform feature map upsampling on the fused feature map through a sub-pixel convolutional layer, redistribute the channel information to the spatial dimension, and obtain an enhanced image through residual connection; Model Training and Optimization Unit: Used to train the model using a paired data set, the loss function fuses the mean square error loss and the perceptual loss, and uses the Adam optimizer to optimize the model to obtain the final machine learning-based land spatial planning data enhancement model.

[0011] In a third aspect, an electronic device includes: a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device runs, the processor communicates with the memory through the bus. When the machine-readable instructions are executed by the processor, a machine learning-based method for enhancing national land spatial planning data is executed.

[0012] In a fourth aspect, a computer-readable storage medium stores a computer program. When the computer program is run by a processor, a machine learning-based method for enhancing national land spatial planning data is executed.

[0013] The advantages of the present invention are as follows: The machine learning-based method for enhancing national land spatial planning data proposed by the present invention can effectively improve the spatial resolution of remote sensing images through multi-scale feature extraction and fusion techniques, enabling the enhanced images to provide more detailed geographical information. By adopting the design of residual connection and sub-pixel convolution, the details and information of the original image are retained during the upsampling process, reducing the artifacts that may occur in traditional upsampling methods, thereby improving the visual quality and reliability of the images. The present invention supports the processing of remote sensing images at different time periods, can effectively capture the impacts brought by seasonal and weather changes, and provides more comprehensive data support for spatio-temporal dynamic analysis in national land spatial planning. Therefore, the present invention provides a machine learning method that can effectively improve the quality of remote sensing images and enhance national land spatial planning data, providing more accurate and efficient technical support for planning decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The drawings are used to provide a further understanding of the present invention and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the present invention and do not constitute a limitation to the present invention.

[0015] Figure 1 It is the overall flowchart of the method of the present invention; Figure 2 It is the comparison diagram of the data-enhanced images of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0016] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0017] Embodiment 1 In this embodiment, as Figure 1As shown in the figure, a machine learning-based method for enhancing land spatial planning data includes the following steps: Data acquisition and processing: Obtain remote sensing image data from multiple data sources to generate low-resolution (LR) and high-resolution (HR) image pairs. Specifically, data acquisition: The multiple data sources are denoted as , where n represents the number of data sources, and the image data collected from the data sources is denoted as ; Data processing: Screen out the high-resolution images that meet the resolution requirements of land spatial planning from the image data , and use the downsampling technique to generate low-resolution images , and construct low-resolution (LR) and high-resolution (HR) image pairs , which is expressed by the formula as follows: , where represents the downsampling operation, represents the scaling factor.

[0018] Construct a machine learning-based land spatial planning data enhancement model, including a multi-scale feature extraction module, a multi-scale feature fusion module, and an image enhancement module. Multi-scale feature extraction: Input the image, and use convolutional kernels of different sizes to extract features of different scales, and output a multi-scale convolutional feature map set. The input image is the low-resolution (LR) image. Specifically, perform a convolution operation on the input image . The size of the image is , represents the height, represents the width, and the number of input channels is . Use multiple convolutional kernel sizes to extract features of different scales, and obtain a multi-scale convolutional feature map set , and the calculation formula is as follows: , where represents the convolution operation, represents the function, represents the convolutional kernel, represents the corresponding bias, represents the feature map output by the i-th convolutional kernel, , is the height of the feature map after the convolution operation, is the width of the feature map after the convolution operation, is the number of output channels of the feature map after the convolution operation. The three convolution kernel sizes are as follows: represent convolution kernels of 3x3, 5x5, and 7x7 respectively.

[0019] Multi-scale feature fusion: Perform weighted average fusion on the feature maps of different scales in the multi-scale convolution feature map set, and obtain the fused feature map through residual connection; Specifically, the multi-scale convolution feature map set fuses multi-scale features in a weighted average manner, and then combines the features obtained after weighted average with the input image for residual connection to obtain the fused feature map The calculation formula is as follows: , , where, is the dot product operation, represents the feature obtained after weighted average, , represents the weight of the feature at the scale, and

[0020] Image enhancement: The fused feature map is upsampled through a sub-pixel convolutional layer to redistribute the channel information to the spatial dimension, and the enhanced image is obtained through residual connection; Specifically, the fused feature map undergoes a convolution operation through a preprocessing convolutional layer to obtain the preprocessed feature map , and the preprocessed feature map is expanded to channels through a sub-pixel convolutional layer to obtain the expanded feature map . The expanded feature map is sub-pixel rearranged to redistribute the channel information to the spatial dimension, obtaining a high-resolution feature map with a size of . Then, the high-resolution feature map and the upsampled input image are subjected to residual connection to obtain the enhanced image . The calculation process formula is as follows: , , , , where, represents the convolution kernel of the preprocessing convolutional layer, Represents the bias of the preprocessing convolutional layer, Represents the convolutional kernel of the sub-pixel convolutional layer, Represents the bias of the sub-pixel convolutional layer, Represents the operation of rearranging channel information into the spatial dimension to achieve upsampling, Represents the upsampling factor, Represents the bilinear interpolation upsampling operation.

[0021] Model training and optimization: The model is trained using a paired dataset. The loss function combines the mean squared error loss and the perceptual loss. The Adam optimizer is used to optimize the model to obtain the final machine learning-based land spatial planning data augmentation model.

[0022] Specifically, training data: The training set includes input images and the corresponding annotated images of a paired dataset. The input image is a low-resolution LR image , and the annotated image is a high-resolution HR image . The training set contains pairs of images denoted as , where j represents the number of image pairs; Specifically, loss function: The mean squared error loss is used to measure the pixel difference between the model output image and the annotated image . The output image is the enhanced image . The formula for the mean squared error loss is as follows: , The formula for the perceptual loss is as follows: , where, represents the number of mini-batch samples in the training process, represents the feature map of the th layer, represents the selected number of layers, represents the th layer, represents the square of the L2 norm. The final loss function is a weighted combination of the above losses, and the formula is as follows: , where, is the hyperparameter corresponding to , is the hyperparameter corresponding to ; Specifically, for model optimization, the Adam optimizer is used to minimize the loss function. For model parameters , the update rule formula for each time is as follows: , where, is the learning rate, is the loss function with respect to the gradient of the parameter , and the cosine annealing learning rate decay strategy is used to control the learning rate during the training process. The formula is as follows: , where, is the number of iterations during the training process, is the initial learning rate, is the total number of training steps.

[0023] Example 2 In this example, to evaluate the effect of the present invention, we select the publicly available high-resolution remote sensing image dataset as the basic data for evaluation, ensuring that the dataset covers a variety of terrains, seasons, and lighting conditions to ensure the extensiveness of the evaluation results. The corresponding low-resolution images are generated by downsampling for algorithm input. The downsampling factors are selected as three different scales of 2 times, 4 times, and 8 times. Table 1 shows the experimental comparison of the method of the present invention with three representative super-resolution image enhancement methods, namely Bicubic Interpolation, SRCNN, and FSRCNN, from two aspects of PSNR (Peak Signal-to-Noise Ratio) and SSIM (Structural Similarity). The experimental results show that the method of the present invention is superior to the traditional Bicubic Interpolation, SRCNN, and FSRCNN methods in improving image quality and structure retention, proving its effectiveness in data enhancement for national land space planning.

[0024] Table 1 Experimental comparison between the method of the present invention and existing methods In this example, as Figure 2 shown is the data enhancement image comparison diagram of the present invention. The left figure is the image before image enhancement, and the right figure is the image after image enhancement. It can be seen from the figure that the image enhanced by the method of the present invention has very high clarity, and the method of the present invention has a very good data enhancement effect.

[0025] Example 3 This example provides a national land space planning data enhancement device based on machine learning, including: Data acquisition and processing unit: used to obtain remote sensing image data from multiple data sources and generate low-resolution LR and high-resolution HR image pairs; Model construction unit: used to construct a machine learning-based land spatial planning data enhancement model, including a multi-scale feature extraction module, a multi-scale feature fusion module, and an image enhancement module; Multi-scale feature extraction unit: used to input an image, extract features of different scales using convolutional kernels of different sizes, and output a multi-scale convolutional feature map set. The input image is a low-resolution LR image; Multi-scale feature fusion unit: used to perform weighted average fusion on the feature maps of different scales in the multi-scale convolutional feature map set, and obtain a fused feature map through residual connection; Image enhancement unit: used to upsample the feature map of the fused feature map through a sub-pixel convolutional layer, redistribute the channel information to the spatial dimension, and obtain an enhanced image through residual connection; Model training and optimization unit: used to train the model using a paired data set, the loss function fuses the mean square error loss and the perceptual loss, and uses the Adam optimizer to optimize the model to obtain the final machine learning-based land spatial planning data enhancement model.

[0026] Embodiment 4 This embodiment provides an electronic device, including: a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device runs, communication between the processor and the memory is carried out through the bus. When the machine-readable instructions are executed by the processor, a machine learning-based land spatial planning data enhancement method is executed.

[0027] The memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. In addition, the memory can include high-speed random access memory, and can also include non-transitory memory, such as at least one disk storage device, a flash memory device, or other non-transitory solid-state storage devices. In some embodiments, the memory may optionally include a memory remotely set relative to the processor, and these remote memories can be connected to the processor through a network. Examples of the above networks include, but are not limited to, the Internet, an enterprise intranet, a local area network, a mobile communication network, and combinations thereof.

[0028] Embodiment 5 This embodiment provides a computer-readable storage medium, on which a computer program is stored. When the computer program is run by a processor, a machine learning-based land spatial planning data enhancement method is executed.

[0029] Storage media include: various media that can store programs, such as USB flash drives, external hard drives, read-only memory (ROM for short), random access memory (RAM for short), magnetic disks, or optical discs.

[0030] Finally, it should be noted that the above are only preferred embodiments of the present invention and are not used to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements for some of the technical features. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for enhancing land space planning data based on machine learning, characterized in that: The following steps are involved: S1. Data acquisition and processing: Acquire remote sensing image data from multiple data sources and generate low-resolution LR and high-resolution HR image pairs; S2. Construct a national land space planning data enhancement model based on machine learning, including a multi-scale feature extraction module, a multi-scale feature fusion module, and an image enhancement module; S3. Multi-scale feature extraction: input an image, use convolution kernels of different sizes to extract features of different scales, and output a multi-scale convolution feature atlas, wherein the input image is a low-resolution LR image; S4. Multi-scale feature fusion: performing weighted average fusion on the feature maps of different scales in the multi-scale convolutional feature map set, and obtaining a fused feature map through residual connection; S5. Image enhancement: The fused feature map is upsampled by a sub-pixel convolution layer, the channel information is redistributed to the spatial dimension, and an enhanced image is obtained by a residual connection; S6. Model training and optimization: Use paired data sets to train the model. The loss function integrates mean square error loss and perceptual loss. Use Adam optimizer to optimize the model and obtain the final land space planning data enhancement model based on machine learning.

2. According to the method for enhancing land space planning data based on machine learning in claim 1, it is characterized in that: The step S1 specifically includes: S11. Data acquisition: The multiple data sources are recorded as , n represents the number of data sources, from which The collected image data is recorded as ; S12. Data processing: from the image data Screening high-resolution images that meet the resolution requirements of national land space planning , using downsampling technology to generate low-resolution images , construct low-resolution LR and high-resolution HR image pairs , the formula is as follows ， in, represents the downsampling operation, Represents the scaling factor.

3. The method for enhancing land space planning data based on machine learning according to claim 1, characterized in that: The multi-scale feature extraction in step S3 specifically includes: For the input image Perform convolution operation, image The size is , Indicates height, represents the width, and the number of input channels is , using a variety of convolution kernel sizes Extract features of different scales to obtain multi-scale convolution feature atlas , the calculation formula is as follows: ， in, represents the convolution operation, express function, represents the convolution kernel, represents the corresponding bias, Represents the feature map output by the i-th convolution kernel, , is the height of the feature map after the convolution operation, is the width of the feature map after the convolution operation, is the number of output channels of the feature map after the convolution operation, and the three convolution kernel sizes are: , representing 3x3, 5x5, and 7x7 convolution kernels respectively.

4. The method for enhancing land space planning data based on machine learning according to claim 1, characterized in that: The step S4 specifically includes: The multi-scale convolutional feature atlas The multi-scale features are fused by weighted averaging, and then the features obtained after weighted averaging are combined with the input image. Perform residual connection to obtain fusion feature map , the calculation formula is as follows: ， ， in, is the dot multiplication operation, Represents the features obtained after weighted averaging. , Indicates The weight of the scale feature, Indicates the number of multi-scale features.

5. The method for enhancing land space planning data based on machine learning according to claim 1, characterized in that: The step S5 specifically includes: The fused feature map After the preprocessing convolution layer performs convolution operation, the preprocessed feature map is obtained The preprocessed feature map is expanded to a sub-pixel convolution layer with a channel number of , and get the expanded feature map The expanded feature map redistributes the channel information to the spatial dimension by rearranging sub-pixels, and the size is High-resolution feature map of , and then the high-resolution feature map And the upsampled input image Perform residual link to obtain enhanced image , the calculation process formula is as follows: ， ， ， ， in, represents the convolution kernel of the preprocessing convolution layer, represents the bias of the preprocessing convolutional layer, represents the convolution kernel of the sub-pixel convolution layer, represents the bias of the sub-pixel convolutional layer, Represents the rearrangement of channel information to the spatial dimension to achieve upsampling operation, represents the upsampling factor, Represents a bilinear interpolation upsampling operation.

6. The method for enhancing land space planning data based on machine learning according to claim 1, characterized in that: The step S6 specifically includes: S51. Training data: The training set includes input images and the corresponding annotated image Paired dataset, input image For low-resolution LR images , label the image For high-resolution HR images , the training set contains The image is recorded as , j represents the number of image pairs; S52. Loss function: The mean square error loss is used to measure the model output image and label images The pixel difference between the two is the output image after enhancement. , the mean square error loss formula is expressed as follows: ， The perceptual loss is expressed as follows: ， in, represents the number of mini-batch samples during training, Indicates The feature map of the layer, Indicates the number of selected layers, Indicates layer, represents the square of the L2 norm. The final loss function is a weighted combination of the above losses, and the formula is as follows: ， in, for The corresponding hyperparameters are, for The corresponding hyperparameters; S53. Model optimization: Adam optimizer is used to minimize the loss function. , each update rule formula is expressed as follows: ， in, is the learning rate, is the loss function Relative to the parameters The gradient of , using the cosine annealing learning rate decay strategy to control the learning rate during training, the formula is as follows: ， in, is the number of iterations in the training process, is the initial learning rate, is the total number of training steps.

7. A national land space planning data enhancement device based on machine learning, characterized in that: include: Data acquisition and processing unit: used to acquire remote sensing image data from multiple data sources and generate low-resolution LR and high-resolution HR image pairs; Model building unit: used to build a national land space planning data enhancement model based on machine learning, including a multi-scale feature extraction module, a multi-scale feature fusion module, and an image enhancement module; Multi-scale feature extraction unit: used to input an image, extract features of different scales using convolution kernels of different sizes, and output a multi-scale convolution feature atlas, wherein the input image is a low-resolution LR image; Multi-scale feature fusion unit: used to perform weighted average fusion on feature maps of different scales in the multi-scale convolution feature map set, and obtain a fused feature map through residual connection; Image enhancement unit: used to upsample the fused feature map through the sub-pixel convolution layer, redistribute the channel information to the spatial dimension, and obtain the enhanced image through the residual connection; Model training and optimization unit: used to train the model using paired data sets. The loss function integrates mean square error loss and perceptual loss. The Adam optimizer is used to optimize the model to obtain the final land space planning data enhancement model based on machine learning.

8. An electronic device, characterized in that: include: A processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor and the memory communicate through the bus. When the machine-readable instructions are executed by the processor, a land space planning data enhancement method based on machine learning is performed as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed by a processor, executes a land space planning data enhancement method based on machine learning as described in any one of claims 1 to 6.

Images