Satellite image vegetation extraction method combining deep learning and vegetation index
By combining deep learning and semantic segmentation networks with vegetation index, the problem of insufficient accuracy of traditional satellite image vegetation extraction methods under complex land objects and low spectral resolution is solved, and higher accuracy and robustness of vegetation extraction are achieved. Adapt to complex lands and suppress noise interference.
Patent Information
- Application Number
- CN202411041278.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-31
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2044-07-31
AI Technical Summary
Traditional satellite image vegetation extraction methods perform poorly in dealing with complex land objects and low spectral resolution, making it difficult to overcome noise and interference factors, resulting in insufficient extraction accuracy and robustness.
Using a semantic segmentation network method combining deep learning and vegetation index, the feature enhancement extraction module learns vegetation representation from different scales and feature levels, suppresses water, soil and building information, improves vegetation identification accuracy, and solves the problem with improved Focal loss function. The problem of positive and negative samples is unbalanced, and the quality of input data is improved through data preprocessing and normalization.
It significantly improves the accuracy and robustness of satellite image vegetation extraction, can more effectively adapt to complex landform conditions and resist noise interference, and improves the ability to identify vegetation.
Smart Images

Figure CN119027807A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical fields of deep learning, computer vision and satellite image vegetation extraction, and in particular to a satellite image vegetation extraction method combining deep learning and vegetation index. Background Art
[0002] With the continuous development of artificial intelligence and deep learning technology, the processing and analysis of satellite remote sensing images have become increasingly important in the fields of agriculture, environmental monitoring, urban planning, etc. Vegetation extraction, as a key task, is of great significance for monitoring crop growth, evaluating the utilization of natural resources, and monitoring environmental changes.
[0003] Traditional vegetation extraction methods are mainly based on vegetation indices, such as the Normalized Difference Vegetation Index and the Normalized Difference Vegetation Index, which reflect the degree of vegetation coverage by calculating the ratio or difference between the infrared band and the visible light band in satellite images. However, traditional methods often perform poorly when dealing with complex objects, poor quality remote sensing images, or low spectral resolution, and it is difficult to overcome the noise and other interference factors in remote sensing data.
[0004] The emergence of deep learning technology provides a new way to solve these problems. Deep learning models learn the characteristics and laws in images through a large amount of training data, can better adapt to complex ground conditions, and have strong generalization capabilities.
[0005] Satellite imaging technology can be used to obtain large-scale, high-resolution earth observation data, providing important support for environmental monitoring and resource management. Combining deep learning and vegetation index methods can make full use of the deep learning network's ability to learn image features, fully mine the information in satellite images, and improve the accuracy and robustness of vegetation extraction. Summary of the invention
[0006] The present invention proposes a satellite image vegetation extraction method combining deep learning and vegetation index, which can improve the accuracy and robustness of vegetation extraction.
[0007] The present invention adopts the following technical solutions.
[0008] A satellite image vegetation extraction method combining deep learning and vegetation index, the method uses vegetation index combined with a deep learning semantic segmentation network to optimize vegetation recognition in satellite images; the method takes vegetation index as input, introduces a feature enhancement extraction module, learns vegetation representation from different scales and different feature levels, so as to optimally highlight vegetation information and suppress other information of water bodies, soil, and buildings, thereby improving the recognition accuracy of vegetation.
[0009] The steps include:
[0010] Step S1: Data preprocessing: preprocessing the collected satellite images, including radiation correction, atmospheric correction and geographic correction, to eliminate noise and errors in the images, and manually annotating the vegetation in the satellite images at the pixel level to obtain labels and establish a satellite image vegetation extraction data set;
[0011] Step S2: construct vegetation index data. Use the preprocessed satellite image data to calculate the vegetation index. The selected vegetation index includes the normalized difference vegetation index NDWI, the simplified vegetation index VARI, and the visible light / infrared vegetation index VI, which form three-channel vegetation index data. These indices can be obtained by calculating the ratio or difference between different bands;
[0012] Step S3: Data normalization: The calculated vegetation index data is normalized so that its value falls within the range of 0 and 1, ensuring that the data input to the neural network has a similar scale and range, as follows: First, for each channel NDWI, VARI and VI, the minimum value min(NDWI), min(VARI) and min(VI), and the maximum value max(NDWI), max(VARI) and max(VI) are calculated respectively, and then the following formula is applied to the data of each channel to normalize the pixel value of each channel to the range of 0 and 1;
[0013] Step S4: constructing a vegetation semantic segmentation network: the vegetation semantic segmentation network includes a feature extraction module, a feature fusion module, and a feature output module; the normalized vegetation index data generated in step S3 is used as the input of the vegetation semantic segmentation network, and the deep features of the input image are extracted through the feature extraction module, and the features of different levels are fused using the feature fusion module, and then the vegetation segmentation result of the satellite image is output through the feature output module;
[0014] Step S5: constructing a loss function for the vegetation semantic segmentation network. In order to effectively solve the imbalance problem between positive and negative samples and between difficult and easy classification samples, an improved Focal loss function is used.
[0015] Step S6: training a vegetation semantic segmentation network, using the vegetation index data generated in step S3 as input of the vegetation semantic segmentation neural network, using the label marked in step S1 as output, and training the vegetation semantic segmentation network;
[0016] Step S7: Identify vegetation in satellite images. The specific method is: after the satellite images to be identified are processed in steps S2 and S3, they are input into the neural network trained in step S6 to obtain the result of vegetation identification by satellite images.
[0017] In step S2, the calculation method for calculating the ratio or difference between different bands to obtain vegetation index data is:
[0018] The Normalized Difference Vegetation Index (NDWI) is used to reflect the distribution of water bodies and vegetation. It is calculated as follows: NDWI = (Green-NIR) / (Green+NIR)
[0019] Among them: NIR represents the reflectance value of the near-infrared band, and Green represents the reflectance value of the green light band.
[0020] The simplified vegetation index VARI is suitable for scenes where soil and vegetation are separated, and has less impact on the soil background. It is calculated as follows:
[0021] VARI=(Green-Red) / (Green+Red-Blue)
[0022] Among them: Red represents the reflection value of the red light wave VARI band, Green represents the reflection value of the green light band, and Blue represents the reflection value of the green light band;
[0023] The Visible / Infrared Vegetation Index VI is used to estimate the green biomass and growth status of vegetation and to compare vegetation coverage in different regions or at different times. It is calculated as follows:
[0024] VI=(NIR-Red) / Red
[0025] Among them: NIR represents the reflectance value of the near-infrared band, and Red represents the reflectance value of the red light band.
[0026] Step S3: Data normalization. The specific method is: normalize the calculated vegetation index data so that their values fall within the range of 0 and 1, ensuring that the data input into the neural network have similar scales and ranges. The specific method is as follows: first, for each channel NDWI, VARI and VI, calculate its minimum value min(NDWI), min(VARI) and min(VI), and maximum value max(NDWI), max(VARI) and max(VI), respectively. Then, apply the following formula to the data of each channel to normalize the pixel value of each channel to the range of 0 and 1.
[0027]
[0028] Where: x i ′ is the eigenvalue after normalization; x represents the original value; min(X) and max(X) represent the minimum and maximum values of the three channels of NDWI, VARI and VI, respectively.
[0029] In step S4, the feature extraction module obtains data features of different scales from different levels to fully capture the spatial information of the input data. It sequentially uses three networks consisting of two convolutional layers with a convolution kernel size of 3×3, a feature enhancement extraction module, and a maximum pooling layer with a kernel size of 3×3. The output feature sizes of the three feature layers S1, S2, and S3 are 1 / 2, 1 / 4, and 1 / 8 of the input image, respectively.
[0030] The feature enhancement extraction module is used to fully capture the deep feature information of the data. The feature enhancement extraction module consists of a convolution layer that uses convolution kernel sizes of 1×1 and 3×3 in sequence, a convolution layer that uses convolution kernel sizes of 1×1, 3×3 and 3×3 in sequence, a convolution layer that uses convolution kernel sizes of 1×1, 3×1, 1×3 in sequence, and a global average pooling layer; the global average pooling is used to aggregate the features of each channel, and the sigmoid function is used to obtain the weights of different channels of the feature map; the channel weights are multiplied by the feature maps of different levels extracted by the non-layered convolution to obtain the features after channel attention enhancement, these enhanced features are spliced, and 1×1 convolution is used to perform feature fusion to obtain the final enhanced features.
[0031] The feature fusion module fuses the features of three different levels, S1, S2 and S3, extracted from the feature extraction module. The feature map S3 is fused with S2 after upsampling, and the fusion result is fused with S1 after upsampling. The upsampling operation adopts the bilinear interpolation method, which can provide better image features while maintaining computational efficiency.
[0032] The feature output module consists of a global average pooling layer, a convolution layer with a convolution kernel size of 3×3, a feature extraction enhancement module, and an upsampling layer, and finally obtains the same size as the input image.
[0033] The mathematical expression of the improved Focal loss function is:
[0034]
[0035] FL(P t )=(1-P t ) γ *lg(P t )*P t 2
[0036] Where: FL is the loss value, P is the model's estimated probability of the category; y = 1 is the true category; γ is the balance factor between difficult and easy classification samples, γ > 0; P t To control the weight of the sample on the total loss.
[0037] Step S6: When training the vegetation semantic segmentation network, pytorch is used as the deep neural network learning framework, and NVDIA TITAN XP is used as the GPU for model training. The Adam optimizer is used, combined with the momentum method and the adaptive learning rate method. The learning rate and momentum parameters are adjusted by maintaining the exponential decay average of the momentum and the exponential decay average of the gradient. The initial learning rate is set to 0.01, and the number of training times is 100 epochs (all training data complete 1 operation).
[0038] The present invention can improve the accuracy and robustness of vegetation extraction. BRIEF DESCRIPTION OF THE DRAWINGS
[0039] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:
[0040] Attached Figure 1 It is a schematic diagram of the method flow of the present invention;
[0041] Attached Figure 2 It is a schematic diagram of the vegetation semantic segmentation network of the present invention;
[0042] Attached Figure 3 It is a schematic diagram of a feature extraction module of the present invention;
[0043] Attached Figure 4 It is a schematic diagram of a feature enhancement extraction module of the present invention;
[0044] Attached Figure 5 It is a schematic diagram of a feature fusion module of the present invention;
[0045] Attached Figure 6 It is a schematic diagram of the feature output module of the present invention. DETAILED DESCRIPTION
[0046] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.
[0047] As shown in the figure, a satellite image vegetation extraction method combining deep learning and vegetation index is described. The method uses vegetation index combined with a deep learning semantic segmentation network to optimize vegetation recognition in satellite images. The method uses vegetation index as input and introduces a feature enhancement extraction module to learn vegetation representation from different scales and feature levels, so as to optimally highlight vegetation information and suppress other information such as water bodies, soil, and buildings, thereby improving the accuracy of vegetation recognition.
[0048] The steps include:
[0049] Step S1: Data preprocessing: preprocessing the collected satellite images, including radiation correction, atmospheric correction and geographic correction, to eliminate noise and errors in the images, and manually annotating the vegetation in the satellite images at the pixel level to obtain labels and establish a satellite image vegetation extraction data set;
[0050] Step S2: construct vegetation index data. Use the preprocessed satellite image data to calculate the vegetation index. The selected vegetation index includes the normalized difference vegetation index NDWI, the simplified vegetation index VARI, and the visible light / infrared vegetation index VI, which form three-channel vegetation index data. These indices can be obtained by calculating the ratio or difference between different bands;
[0051] Step S3: Data normalization: The calculated vegetation index data is normalized so that its value falls within the range of 0 and 1, ensuring that the data input to the neural network has a similar scale and range, as follows: First, for each channel NDWI, VARI and VI, the minimum value min(NDWI), min(VARI) and min(VI), and the maximum value max(NDWI), max(VARI) and max(VI) are calculated respectively, and then the following formula is applied to the data of each channel to normalize the pixel value of each channel to the range of 0 and 1;
[0052] Step S4: constructing a vegetation semantic segmentation network: the vegetation semantic segmentation network includes a feature extraction module, a feature fusion module, and a feature output module; the normalized vegetation index data generated in step S3 is used as the input of the vegetation semantic segmentation network, and the deep features of the input image are extracted through the feature extraction module, and the features of different levels are fused using the feature fusion module, and then the vegetation segmentation result of the satellite image is output through the feature output module;
[0053] Step S5: constructing a loss function for the vegetation semantic segmentation network. In order to effectively solve the imbalance problem between positive and negative samples and between difficult and easy classification samples, an improved Focal loss function is used.
[0054] Step S6: training a vegetation semantic segmentation network, using the vegetation index data generated in step S3 as input of the vegetation semantic segmentation neural network, using the label marked in step S1 as output, and training the vegetation semantic segmentation network;
[0055] Step S7: Identify vegetation in satellite images. The specific method is: after the satellite images to be identified are processed in steps S2 and S3, they are input into the neural network trained in step S6 to obtain the result of vegetation identification by satellite images.
[0056] In step S2, the calculation method for calculating the ratio or difference between different bands to obtain vegetation index data is:
[0057] The Normalized Difference Vegetation Index (NDWI) is used to reflect the distribution of water bodies and vegetation. It is calculated as follows: NDWI = (Green-NIR) / (Green+NIR)
[0058] Among them: NIR represents the reflectance value of the near-infrared band, and Green represents the reflectance value of the green light band.
[0059] The simplified vegetation index VARI is suitable for scenes where soil and vegetation are separated, and has less impact on the soil background. It is calculated as follows:
[0060] VARI=(Green-Red) / (Green+Red-Blue)
[0061] Among them: Red represents the reflection value of the red light wave VARI band, Green represents the reflection value of the green light band, and Blue represents the reflection value of the green light band;
[0062] The Visible / Infrared Vegetation Index VI is used to estimate the green biomass and growth status of vegetation and to compare vegetation coverage in different regions or at different times. It is calculated as follows:
[0063] VI=(NIR-Red) / Red
[0064] Among them: NIR represents the reflectance value of the near-infrared band, and Red represents the reflectance value of the red light band.
[0065] Step S3: Data normalization. The specific method is: normalize the calculated vegetation index data so that their values fall within the range of 0 and 1, ensuring that the data input into the neural network have similar scales and ranges. The specific method is as follows: first, for each channel NDWI, VARI and VI, calculate its minimum value min(NDWI), min(VARI) and min(VI), and maximum value max(NDWI), max(VARI) and max(VI), respectively. Then, apply the following formula to the data of each channel to normalize the pixel value of each channel to the range of 0 and 1.
[0066]
[0067] Where: x i ′ is the eigenvalue after normalization; x represents the original value; min(X) and max(X) represent the minimum and maximum values of the three channels of NDWI, VARI and VI, respectively.
[0068] In step S4, the feature extraction module obtains data features of different scales from different levels to fully capture the spatial information of the input data. It sequentially uses three networks consisting of two convolutional layers with a convolution kernel size of 3×3, a feature enhancement extraction module, and a maximum pooling layer with a kernel size of 3×3. The output feature sizes of the three feature layers S1, S2, and S3 are 1 / 2, 1 / 4, and 1 / 8 of the input image, respectively.
[0069] The feature enhancement extraction module is used to fully capture the deep feature information of the data. The feature enhancement extraction module consists of a convolution layer that uses convolution kernel sizes of 1×1 and 3×3 in sequence, a convolution layer that uses convolution kernel sizes of 1×1, 3×3 and 3×3 in sequence, a convolution layer that uses convolution kernel sizes of 1×1, 3×1, 1×3 in sequence, and a global average pooling layer; the global average pooling is used to aggregate the features of each channel, and the sigmoid function is used to obtain the weights of different channels of the feature map; the channel weights are multiplied by the feature maps of different levels extracted by the non-layered convolution to obtain the features after channel attention enhancement, these enhanced features are spliced, and 1×1 convolution is used to perform feature fusion to obtain the final enhanced features.
[0070] The feature fusion module fuses the features of three different levels, S1, S2 and S3, extracted from the feature extraction module. The feature map S3 is fused with S2 after upsampling, and the fusion result is fused with S1 after upsampling. The upsampling operation adopts the bilinear interpolation method, which can provide better image features while maintaining computational efficiency.
[0071] The feature output module consists of a global average pooling layer, a convolution layer with a convolution kernel size of 3×3, a feature extraction enhancement module, and an upsampling layer, and finally obtains the same size as the input image.
[0072] The mathematical expression of the improved Focal loss function is:
[0073]
[0074] FL(P t )=(1-P t ) γ *lg(P t )*P t 2
[0075] Where: FL is the loss value, P is the model's estimated probability of the category; y = 1 is the true category; γ is the balance factor between difficult and easy classification samples, γ > 0; P t To control the weight of the sample on the total loss.
[0076] Step S6: When training the vegetation semantic segmentation network, pytorch is used as the deep neural network learning framework, and NVDIA TITAN XP is used as the GPU for model training. The Adam optimizer is used, combined with the momentum method and the adaptive learning rate method. The learning rate and momentum parameters are adjusted by maintaining the exponential decay average of the momentum and the exponential decay average of the gradient. The initial learning rate is set to 0.01, and the number of training times is 100 epochs (all training data complete 1 operation).
Claims
1. A satellite image vegetation extraction method combining deep learning and vegetation index, characterized by: The method uses vegetation index combined with a deep learning semantic segmentation network to optimize vegetation recognition in satellite images. It takes vegetation index as input, introduces a feature enhancement extraction module, learns vegetation representation from different scales and different feature levels, and optimally highlights vegetation information and suppresses other information such as water bodies, soil, and buildings, thereby improving the accuracy of vegetation recognition.
2. The satellite image vegetation extraction method combining deep learning and vegetation index according to claim 1, characterized in that: The steps include: Step S1: Data preprocessing: preprocessing the collected satellite images, including radiation correction, atmospheric correction and geographic correction, to eliminate noise and errors in the images, and manually annotating the vegetation in the satellite images at the pixel level to obtain labels and establish a satellite image vegetation extraction data set; Step S2: construct vegetation index data. Use the preprocessed satellite image data to calculate the vegetation index. The selected vegetation index includes the normalized difference vegetation index NDWI, the simplified vegetation index VARI, and the visible light / infrared vegetation index VI, which form three-channel vegetation index data. These indices can be obtained by calculating the ratio or difference between different bands; Step S3: Data normalization: The calculated vegetation index data is normalized so that its value falls within the range of 0 and 1, ensuring that the data input to the neural network has a similar scale and range, as follows: First, for each channel NDWI, VARI and VI, the minimum value min(NDWI), min(VARI) and min(VI), and the maximum value max(NDWI), max(VARI) and max(VI) are calculated respectively, and then the following formula is applied to the data of each channel to normalize the pixel value of each channel to the range of 0 and 1; Step S4: constructing a vegetation semantic segmentation network: the vegetation semantic segmentation network includes a feature extraction module, a feature fusion module, and a feature output module; The normalized vegetation index data generated in step S3 is used as the input of the vegetation semantic segmentation network, and the deep features of the input image are extracted by the feature extraction module, and the features of different levels are fused by the feature fusion module, and the vegetation segmentation result of the satellite image is output by the feature output module; Step S5: constructing a loss function for the vegetation semantic segmentation network. In order to effectively solve the imbalance problem between positive and negative samples and between difficult and easy classification samples, an improved Focal loss function is used. Step S6: training a vegetation semantic segmentation network, using the vegetation index data generated in step S3 as input of the vegetation semantic segmentation neural network, using the label marked in step S1 as output, and training the vegetation semantic segmentation network; Step S7: Identify vegetation in satellite images. The specific method is: after the satellite images to be identified are processed in steps S2 and S3, they are input into the neural network trained in step S6 to obtain the result of vegetation identification by satellite images.
3. The satellite image vegetation extraction method combining deep learning and vegetation index according to claim 2, characterized in that: In step S2, the calculation method for calculating the ratio or difference between different bands to obtain vegetation index data is: The Normalized Difference Vegetation Index (NDWI) is used to reflect the distribution of water bodies and vegetation. It is calculated as follows: NDWI=(Green-NIR) / (Green+NIR) Among them: NIR represents the reflectance value of the near-infrared band, and Green represents the reflectance value of the green light band. The simplified vegetation index VARI is suitable for scenes where soil and vegetation are separated, and has less impact on the soil background. It is calculated as follows: VARI=(Green-Red) / (Green+Red-Blue) Among them: Red represents the reflection value of the red light wave VARI band, Green represents the reflection value of the green light band, and Blue represents the reflection value of the green light band; The Visible / Infrared Vegetation Index VI is used to estimate the green biomass and growth status of vegetation and to compare vegetation coverage in different regions or at different times. It is calculated as follows: VI=(NIR-Red) / Red Among them: NIR represents the reflectance value of the near-infrared band, and Red represents the reflectance value of the red light band.
4. The satellite image vegetation extraction method combining deep learning and vegetation index according to claim 2, characterized in that: Step S3: Data normalization. The specific method is: normalize the calculated vegetation index data so that their values fall within the range of 0 and 1, ensuring that the data input into the neural network have similar scales and ranges. The specific method is as follows: first, for each channel NDWI, VARI and VI, calculate its minimum value min(NDWI), min(VARI) and min(VI), and maximum value max(NDWI), max(VARI) and max(VI), respectively. Then, apply the following formula to the data of each channel to normalize the pixel value of each channel to the range of 0 and 1. Where: x i ′ is the eigenvalue after normalization; x represents the original value; min(X) and max(X) represent the minimum and maximum values of the three channels of NDWI, VARI and VI, respectively.
5. The satellite image vegetation extraction method combining deep learning and vegetation index according to claim 2, characterized in that: In step S4, the feature extraction module obtains data features of different scales from different levels to fully capture the spatial information of the input data. It sequentially uses three networks consisting of two convolutional layers with a convolution kernel size of 3×3, a feature enhancement extraction module, and a maximum pooling layer with a kernel size of 3×3. The output feature sizes of the three feature layers S1, S2, and S3 are 1 / 2, 1 / 4, and 1 / 8 of the input image, respectively.
6. The satellite image vegetation extraction method combining deep learning and vegetation index according to claim 5, characterized in that: The feature enhancement extraction module is used to fully capture the deep feature information of the data. The feature enhancement extraction module consists of a convolution layer that uses convolution kernel sizes of 1×1 and 3×3 in sequence, a convolution layer that uses convolution kernel sizes of 1×1, 3×3 and 3×3 in sequence, a convolution layer that uses convolution kernel sizes of 1×1, 3×1, 1×3 in sequence, and a global average pooling layer; the global average pooling is used to aggregate the features of each channel, and the sigmoid function is used to obtain the weights of different channels of the feature map; the channel weights are multiplied by the feature maps of different levels extracted by the non-layered convolution to obtain the features after channel attention enhancement, these enhanced features are spliced, and 1×1 convolution is used to perform feature fusion to obtain the final enhanced features.
7. The satellite image vegetation extraction method combining deep learning and vegetation index according to claim 5, characterized in that: The feature fusion module fuses the features of three different levels, S1, S2 and S3, extracted from the feature extraction module. The feature map S3 is fused with S2 after upsampling, and the fusion result is fused with S1 after upsampling. The upsampling operation adopts the bilinear interpolation method, which can provide better image features while maintaining computational efficiency.
8. The satellite image vegetation extraction method combining deep learning and vegetation index according to claim 2, characterized in that: The feature output module consists of a global average pooling layer, a convolution layer with a convolution kernel size of 3×3, a feature extraction enhancement module, and an upsampling layer, and finally obtains the same size as the input image.
9. The satellite image vegetation extraction method combining deep learning and vegetation index according to claim 2, characterized in that: The mathematical expression of the improved Focal loss function is: FL(P t )=(1-P t ) γ *lg(P t )*P t 2 Where: FL is the loss value, P is the model's estimated probability of the category; y = 1 is the true category; γ is the balance factor between difficult and easy classification samples, γ > 0; P t To control the weight of the sample on the total loss.
10. The satellite image vegetation extraction method combining deep learning and vegetation index according to claim 2, characterized in that: Step S6: When training the vegetation semantic segmentation network, pytorch is used as the deep neural network learning framework, and NVDIA TITAN XP is used as the GPU for model training. The Adam optimizer is used, combined with the momentum method and the adaptive learning rate method. The learning rate and momentum parameters are adjusted by maintaining the exponential decay average of the momentum and the exponential decay average of the gradient. The initial learning rate is set to 0.01, and the number of training epochs is 100.
Citation Information
Patent Citations
Remote sensing vegetation extraction method and device based on deep learning semantic segmentation
CN116385875A
Method for identifying turf in rice field by using multi-temporal high-resolution satellite image
CN116543316A
Remote sensing multispectral image water body identification method based on two-channel segmentation network
CN117726936A
System and method of feature detection in satellite images using neural networks
US20220301301A1
Cited By
Insect pest detection method and device
CN120032180A
