Method, system and computer program for simultaneous detection of forest canopy density and tree species type

Abstract

The application discloses a kind of forest canopy density and tree species type synchronous detection method, comprising the following steps: obtaining the unmanned aerial vehicle multi-source remote sensing data of test area, constructs sample data set;According to sample data set, based on multi-task learning constructs multi-task neural network model and trains, obtains forest canopy density and tree species type synchronous detection model;Obtain the satellite multispectral remote sensing image data of the region to be detected;The satellite multispectral remote sensing image data of the region to be detected is input into forest canopy density and tree species type synchronous detection model, and the synchronous detection result of forest canopy density and tree species type is obtained.The technical scheme provided by the application realizes the synchronous detection of forest canopy density and tree species type by multi-task neural network model;Fine-tune the network model to improve;Statistical analysis is carried out on the detection result;Construct the spatiotemporal distribution diagram of the change of forest canopy density and tree species type in the region to be detected.

Description

Technical Field

[0001] This invention relates to the field of forest image processing, and more particularly to a method for simultaneous detection of forest canopy closure and tree species type. Background Technology

[0002] Forest canopy closure typically refers to the ratio of the vertical projection area of ​​tree canopies to the total area of ​​the forest land. Canopy closure is an important factor reflecting forest structure and the forest environment, and it has wide applications in soil and water conservation, forest stand quality assessment, and forest landscape construction. It is also a crucial factor in forest management, including dividing forest stands into small plots, determining the intensity of thinning and harvesting, and identifying whether a forest is a forest.

[0003] The identification and classification of forest tree species are of great significance for the statistical analysis of forestry resources, the exploration of ecosystem diversity, and even the study of biological evolution. Therefore, the accurate identification and statistical analysis of tree species in forest areas is becoming increasingly important. With the rapid development of digital information technology in recent years, digital forestry based on remote sensing technology has greatly facilitated the sustainable management of forest resources.

[0004] Simultaneous detection of forest canopy closure and tree species type can obtain more complete and richer forest-related information and conditions for the area to be detected. By combining tree species type and canopy closure, the structure and density of the forest stand can be assessed more accurately, which facilitates forest management or ecological protection.

[0005] However, existing technical methods can only calculate forest canopy density or identify forest tree species types separately, without simultaneously detecting both. The detection results are singular and the application is greatly limited. Summary of the Invention

[0006] In view of the above-mentioned shortcomings of the current technology, the present invention provides a method for simultaneous detection of forest canopy closure and tree species type. By constructing a multi-task neural network model based on multi-task learning, it is possible to achieve simultaneous detection of forest canopy closure and tree species type.

[0007] To achieve the above objectives, the embodiments of the present invention adopt the following technical solutions:

[0008] A method for simultaneous detection of forest canopy density and tree species type includes the following steps:

[0009] Acquire multi-source remote sensing data from UAVs in the test area and construct a sample dataset;

[0010] Based on the sample dataset, a multi-task neural network model was constructed and trained using multi-task learning to obtain a synchronous detection model for forest canopy closure and tree species type.

[0011] Acquire satellite multispectral remote sensing image data of the area to be detected;

[0012] The satellite multispectral remote sensing image data of the area to be detected is input into the forest canopy closure and tree species type synchronous detection model to obtain the synchronous detection results of forest canopy closure and tree species type.

[0013] According to one aspect of the present invention, the UAV multi-source remote sensing data includes spatial image data based on CCD high-resolution remote sensing image data and distance image data based on LiDAR scan data.

[0014] According to one aspect of the present invention, when constructing a sample dataset, tree species type is labeled for spatial image data and forest canopy density is labeled for distance image data.

[0015] According to one aspect of the present invention, the method for simultaneous detection of forest canopy density and tree species type further includes: preprocessing the UAV multi-source remote sensing data of the test area and the satellite multispectral remote sensing image data of the area to be detected.

[0016] According to one aspect of the present invention, the preprocessing of the UAV multi-source remote sensing data of the test area and the satellite multispectral remote sensing image data of the area to be tested includes: noise removal, geometric correction, radiometric calibration, atmospheric correction and image registration.

[0017] According to one aspect of the present invention, the construction and training of a multi-task neural network model based on multi-task learning includes:

[0018] Based on the Swing Transformer model, a hierarchical feature representation and a shift-window-based self-attention mechanism are introduced to construct a deep shared feature extraction layer;

[0019] After the shared feature extraction layer, branch networks are constructed for the forest canopy closure detection task and the tree species type detection task, respectively. The branch networks adopt weight sharing, self-attention and shared attention strategies.

[0020] The branch network adopts a multi-task loss function, which integrates the weighted losses of the forest canopy closure detection task and the tree species type detection task, and introduces adjustable weight coefficients to dynamically adjust the weights according to the task characteristics and complexity.

[0021] According to one aspect of the present invention, the step of constructing and training a multi-task neural network model based on multi-task learning further includes:

[0022] When training a neural network model, a shared feature extraction layer is trained first, and then task-specific layers are gradually added for fine-tuning.

[0023] Early stopping mechanisms and model regularization measures are adopted to prevent overfitting, enhance the model's generalization ability, and ensure the stability of the training process.

[0024] According to one aspect of the present invention, the step of constructing and training a multi-task neural network model based on multi-task learning further includes:

[0025] The model's performance is optimized by employing cross-validation, learning rate adjustment, and weight allocation methods.

[0026] According to one aspect of the present invention, the step of constructing and training a multi-task neural network model based on multi-task learning further includes:

[0027] Multiple metrics, including at least precision, recall, F1 score, and overall accuracy, are used to comprehensively evaluate model performance.

[0028] Based on the comprehensive evaluation results, the network model was fine-tuned.

[0029] According to one aspect of the present invention, the method for simultaneous detection of forest canopy density and tree species type further includes the following steps:

[0030] Statistical analysis was performed on the results of forest canopy density and tree species type detection.

[0031] According to one aspect of the present invention, the method for simultaneous detection of forest canopy density and tree species type further includes the following steps:

[0032] Based on the detection results of forest canopy density and tree species type at different time points in the area to be detected, a spatiotemporal distribution map of forest canopy density and tree species type in the area to be detected is constructed.

[0033] A system for simultaneous detection of forest canopy closure and tree species type, based on the method for simultaneous detection of forest canopy closure and tree species type as described above, includes:

[0034] The sample construction module is used to acquire multi-source remote sensing data from UAVs in the test area and construct a sample dataset.

[0035] The model building module is used to build and train a multi-task neural network model based on multi-task learning using the sample dataset to obtain a simultaneous detection model for forest canopy closure and tree species type.

[0036] The data acquisition module is used to acquire satellite multispectral remote sensing image data of the area to be detected;

[0037] The detection module is used to input satellite multispectral remote sensing image data of the area to be detected into the forest canopy closure and tree species type synchronous detection model to obtain the synchronous detection results of forest canopy closure and tree species type.

[0038] A computer program, when executed, implements the steps of the method for simultaneous detection of forest canopy closure and tree species type as described above.

[0039] Advantages of implementing this invention:

[0040] This invention provides a method for simultaneous detection of forest canopy closure and tree species type. By constructing a multi-task neural network model based on multi-task learning, applying Swing Transformer for feature extraction, comprehensively extracting different characteristics of multi-source remote sensing data, and performing deep learning, the method achieves simultaneous detection of forest canopy closure and tree species type.

[0041] This method can also comprehensively evaluate the network model based on multiple indicators, and fine-tune and improve the network model according to the evaluation results. After obtaining the detection results of forest canopy density and tree species type in the detection area through the detection model, statistical analysis of the detection results can further obtain relevant information and conditions of the forest in the area, which is convenient for subsequent ecological protection. Based on the detection results of forest canopy density and tree species type, the data is integrated to construct a spatiotemporal distribution map of the changes in forest canopy density and tree species type in the detection area, which can provide relevant information for environmental monitoring and decision analysis. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1 This is a flowchart of a method for simultaneous detection of forest canopy density and tree species type according to Embodiment 1 of the present invention;

[0044] Figure 2 This is a structural diagram of the forest canopy density and tree species type synchronous detection model described in this invention;

[0045] Figure 3 This is a flowchart of a method for simultaneous detection of forest canopy density and tree species type according to Embodiment 2 of the present invention. Detailed Implementation

[0046] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0047] Example 1

[0048] like Figure 1 As shown, a method for simultaneous detection of forest canopy density and tree species type includes the following steps:

[0049] S1: Acquire multi-source remote sensing data from UAVs in the test area and construct a sample dataset.

[0050] The UAV multi-source remote sensing data includes spatial image data based on CCD high-resolution remote sensing image data and distance image data based on LiDAR scan data.

[0051] High-resolution CCD remote sensing imagery is a type of remote sensing imagery with high spatial resolution acquired using charge-coupled device (CCD) technology. With its high resolution, high temporal resolution, high spectral resolution, and rich texture features, CCD high-resolution remote sensing imagery plays a vital role in fields such as environmental monitoring, resource surveys, urban planning, and military reconnaissance.

[0052] LiDAR scan data is a type of 3D data acquired using a laser scanner (LiDAR scanner). LiDAR primarily measures distance information; the LiDAR scanner calculates the distance to an object by emitting laser pulses and measuring their return time.

[0053] The method for simultaneous detection of forest canopy density and tree species type also includes: preprocessing multi-source remote sensing data from UAVs in the experimental area.

[0054] The preprocessing of the UAV multi-source remote sensing data of the test area includes: noise removal, geometric correction, radiometric calibration, atmospheric correction, and image registration.

[0055] Radiometric calibration refers to the process of converting the brightness grayscale values ​​of an image into absolute radiance when calculating the spectral reflectance or spectral radiance of a ground object, or when comparing images acquired at different times and by different sensors.

[0056] Atmospheric correction refers to eliminating errors in radiation levels caused by atmospheric absorption, especially scattering, and retrieving the true surface reflectance of ground features.

[0057] Image registration refers to the matching and overlaying of two or more images acquired at different times, by different detectors, or under different conditions during image processing. Through image registration, changes in ground features at different times or under different conditions can be compared, or images acquired by different sensors can be fused to improve the spatial or spectral resolution of the data. Image registration includes feature-based registration methods, grayscale-based registration methods, and transform domain-based registration methods, among others.

[0058] Radiometric calibration, atmospheric correction, and image registration are three indispensable steps in remote sensing data processing. Radiometric calibration ensures the accuracy and reliability of remote sensing data; atmospheric correction eliminates the influence of the atmosphere on remote sensing data, making the data more reflective of the true situation of ground features; and image registration enables the comparison and fusion of remote sensing data from different times and under different conditions, providing a foundation for monitoring ground feature changes and data analysis.

[0059] When constructing the sample dataset, tree species types were labeled for spatial image data based on CCD high-resolution remote sensing image data, and forest canopy density was labeled for distance image data based on LiDAR scan data.

[0060] When constructing the sample dataset, data augmentation is performed on the image data to build a broader sample dataset.

[0061] Data augmentation methods such as random rotation, flipping, and cropping are employed to enhance the model's ability to generalize to diverse surface features.

[0062] In practical applications, this method aims to establish a diverse and spatially widely distributed large-scale sample database to support iterative training and parameter optimization of deep learning models. The goal is to improve the training accuracy of the model to over 85%. For models that do not reach this accuracy standard, detailed testing will be conducted, and their interpretation markers and variation type labels will be supplemented and improved. Training parameters will be adjusted, and iterative training will be repeated to achieve even higher accuracy.

[0063] S2: Based on the sample dataset, construct and train a multi-task neural network model based on multi-task learning to obtain a synchronous detection model for forest canopy closure and tree species type.

[0064] The construction and training of a multi-task neural network model based on multi-task learning includes:

[0065] Based on the Swing Transformer model, a hierarchical feature representation and a shift-window-based self-attention mechanism are introduced to construct a deep shared feature extraction layer;

[0066] After the shared feature extraction layer, branch networks are constructed for the forest canopy closure detection task and the tree species type detection task, respectively. The branch networks adopt weight sharing, self-attention and shared attention strategies.

[0067] The branch network adopts a multi-task loss function, which integrates the weighted losses of the forest canopy closure detection task and the tree species type detection task, and introduces adjustable weight coefficients to dynamically adjust the weights according to the task characteristics and complexity.

[0068] In practical applications, the multi-task neural network model can adopt the multi-task fully convolutional neural network model (MT-FCN).

[0069] Multi-task fully convolutional neural network (MT-FCN) is a deep learning framework capable of processing multiple related visual tasks simultaneously. It typically includes a shared feature extraction part and multiple branch networks designed for specific tasks. By sharing the feature extraction part, MT-FCN reduces redundant computation, improves efficiency, and enhances the performance of each task through the correlation between tasks.

[0070] In multi-task fully convolutional neural network models, the Swing Transformer can be used as a shared feature extraction layer. The hierarchical design and windowed self-attention mechanism of the Swing Transformer enable it to efficiently extract local and global features from images, providing rich feature representations for subsequent specific tasks.

[0071] Swin Transformer is an advanced deep learning model designed specifically for computer vision tasks. It combines the powerful capabilities of Transformer models in sequence data processing with the efficient local feature extraction advantages of Convolutional Neural Networks (CNNs) in image recognition. Swin Transformer addresses the computational complexity issue in high-resolution image processing by introducing hierarchical feature representation and a shift-window-based self-attention mechanism, while maintaining high performance.

[0072] After the shared feature extraction layer, independent branch networks are created for each task. These branch networks can be built based on the blocks of the Swing Transformer, or they can be combined with other convolutional layers, fully connected layers (fully connected layers are not usually used in FCN, but can be appropriately adjusted according to the specific task requirements), or upsampling layers, etc.

[0073] In practical applications, the shared feature extraction layer can extract features only from the satellite multispectral remote sensing image data of the area to be detected, or it can extract features from the image data of both the test area and the area to be detected.

[0074] like Figure 2 The diagram shown is a structural diagram of the forest canopy closure and tree species type simultaneous detection model in this method.

[0075] The process of constructing and training a multi-task neural network model based on multi-task learning also includes:

[0076] When training a neural network model, a shared feature extraction layer is trained first, and then task-specific layers are gradually added for fine-tuning.

[0077] Early stopping mechanisms and model regularization measures are adopted to prevent overfitting, enhance the model's generalization ability, and ensure the stability of the training process.

[0078] Preferably, the step of constructing and training a multi-task neural network model based on multi-task learning further includes:

[0079] The performance of the model is optimized by methods such as cross-validation, learning rate adjustment, and weight allocation.

[0080] S3: Acquire satellite multispectral remote sensing image data of the area to be detected.

[0081] In practical applications, this method utilizes satellite multispectral remote sensing imagery of the area to be detected from Sentinel-2. Sentinel-2 is a series of high-resolution multispectral imaging satellites launched by the European Space Agency (ESA). Sentinel-2 is equipped with a state-of-the-art Multispectral Instrument (MSI), a sensor capable of capturing images in 13 different spectral bands, covering the range from visible light to shortwave infrared. Sentinel-2 data is widely used in environmental monitoring, including vegetation cover, soil moisture, and water quality monitoring. Its multispectral data helps to accurately identify land cover types, providing a scientific basis for environmental protection.

[0082] The satellite multispectral remote sensing image data includes spatial image data and distance image data.

[0083] Spatial imagery data can be used to detect forest tree species types, while distance imagery data can be used to detect forest canopy closure.

[0084] The method for simultaneous detection of forest canopy density and tree species type also includes: preprocessing satellite multispectral remote sensing image data of the area to be detected.

[0085] The preprocessing of satellite multispectral remote sensing image data of the area to be detected includes: noise removal, geometric correction, radiometric calibration, atmospheric correction, and image registration.

[0086] S4: Input the satellite multispectral remote sensing image data of the area to be detected into the forest canopy closure and tree species type synchronous detection model to obtain the synchronous detection results of forest canopy closure and tree species type.

[0087] The output of this detection model consists of two parts: forest canopy closure detection results and tree species type detection results. These two parts together constitute a comprehensive analysis of satellite multispectral remote sensing image data, providing rich information for subsequent application scenarios.

[0088] The beneficial effects of this embodiment are as follows: This method constructs a multi-task neural network model based on multi-task learning, applies Swing Transformer for feature extraction, comprehensively extracts different characteristics of multi-source remote sensing data, performs deep learning, and achieves simultaneous detection of forest canopy density and tree species type.

[0089] Example 2

[0090] like Figure 3 As shown, a method for simultaneous detection of forest canopy density and tree species type includes the following steps:

[0091] S1: Acquire multi-source remote sensing data from UAVs in the test area and construct a sample dataset.

[0092] The UAV multi-source remote sensing data includes spatial image data based on CCD high-resolution remote sensing image data and distance image data based on LiDAR scan data.

[0093] High-resolution CCD remote sensing imagery is a type of remote sensing imagery with high spatial resolution acquired using charge-coupled device (CCD) technology. With its high resolution, high temporal resolution, high spectral resolution, and rich texture features, CCD high-resolution remote sensing imagery plays a vital role in fields such as environmental monitoring, resource surveys, urban planning, and military reconnaissance.

[0094] LiDAR scan data is a type of 3D data acquired using a laser scanner (LiDAR scanner). LiDAR primarily measures distance information; the LiDAR scanner calculates the distance to an object by emitting laser pulses and measuring their return time.

[0095] The method for simultaneous detection of forest canopy density and tree species type also includes: preprocessing multi-source remote sensing data from UAVs in the experimental area.

[0096] The preprocessing of the UAV multi-source remote sensing data of the test area includes: noise removal, geometric correction, radiometric calibration, atmospheric correction, and image registration.

[0097] Radiometric calibration refers to the process of converting the brightness grayscale values ​​of an image into absolute radiance when calculating the spectral reflectance or spectral radiance of a ground object, or when comparing images acquired at different times and by different sensors.

[0098] Atmospheric correction refers to eliminating errors in radiation levels caused by atmospheric absorption, especially scattering, and retrieving the true surface reflectance of ground features.

[0099] Image registration refers to the matching and overlaying of two or more images acquired at different times, by different detectors, or under different conditions during image processing. Through image registration, changes in ground features at different times or under different conditions can be compared, or images acquired by different sensors can be fused to improve the spatial or spectral resolution of the data. Image registration includes feature-based registration methods, grayscale-based registration methods, and transform domain-based registration methods, among others.

[0100] Radiometric calibration, atmospheric correction, and image registration are three indispensable steps in remote sensing data processing. Radiometric calibration ensures the accuracy and reliability of remote sensing data; atmospheric correction eliminates the influence of the atmosphere on remote sensing data, making the data more reflective of the true situation of ground features; and image registration enables the comparison and fusion of remote sensing data from different times and under different conditions, providing a foundation for monitoring ground feature changes and data analysis.

[0101] When constructing the sample dataset, tree species types were labeled for spatial image data based on CCD high-resolution remote sensing image data, and forest canopy density was labeled for distance image data based on LiDAR scan data.

[0102] When constructing the sample dataset, data augmentation is performed on the image data to build a broader sample dataset.

[0103] Data augmentation methods such as random rotation, flipping, and cropping are employed to enhance the model's ability to generalize to diverse surface features.

[0104] In practical applications, this method aims to establish a diverse and spatially widely distributed large-scale sample database to support iterative training and parameter optimization of deep learning models. The goal is to improve the training accuracy of the model to over 85%. For models that do not reach this accuracy standard, detailed testing will be conducted, and their interpretation markers and variation type labels will be supplemented and improved. Training parameters will be adjusted, and iterative training will be repeated to achieve even higher accuracy.

[0105] S2: Based on the sample dataset, construct and train a multi-task neural network model based on multi-task learning to obtain a synchronous detection model for forest canopy closure and tree species type.

[0106] The construction and training of a multi-task neural network model based on multi-task learning includes:

[0107] Based on the Swing Transformer model, a hierarchical feature representation and a shift-window-based self-attention mechanism are introduced to construct a deep shared feature extraction layer;

[0108] After the shared feature extraction layer, branch networks are constructed for the forest canopy closure detection task and the tree species type detection task, respectively. The branch networks adopt weight sharing, self-attention and shared attention strategies.

[0109] The branch network adopts a multi-task loss function, which integrates the weighted losses of the forest canopy closure detection task and the tree species type detection task, and introduces adjustable weight coefficients to dynamically adjust the weights according to the task characteristics and complexity.

[0110] In practical applications, the multi-task neural network model can adopt the multi-task fully convolutional neural network model (MT-FCN).

[0111] Multi-task fully convolutional neural network (MT-FCN) is a deep learning framework capable of processing multiple related visual tasks simultaneously. It typically includes a shared feature extraction part and multiple branch networks designed for specific tasks. By sharing the feature extraction part, MT-FCN reduces redundant computation, improves efficiency, and enhances the performance of each task through the correlation between tasks.

[0112] In multi-task fully convolutional neural network models, the Swing Transformer can be used as a shared feature extraction layer. The hierarchical design and windowed self-attention mechanism of the Swing Transformer enable it to efficiently extract local and global features from images, providing rich feature representations for subsequent specific tasks.

[0113] Swin Transformer is an advanced deep learning model designed specifically for computer vision tasks. It combines the powerful capabilities of Transformer models in sequence data processing with the efficient local feature extraction advantages of Convolutional Neural Networks (CNNs) in image recognition. Swin Transformer addresses the computational complexity issue in high-resolution image processing by introducing hierarchical feature representation and a shift-window-based self-attention mechanism, while maintaining high performance.

[0114] After the shared feature extraction layer, independent branch networks are created for each task. These branch networks can be built based on the blocks of the Swing Transformer, or they can be combined with other convolutional layers, fully connected layers (fully connected layers are not usually used in FCN, but can be appropriately adjusted according to the specific task requirements), or upsampling layers, etc.

[0115] In practical applications, the shared feature extraction layer can extract features only from the satellite multispectral remote sensing image data of the area to be detected, or it can extract features from the image data of both the test area and the area to be detected.

[0116] like Figure 2The diagram shown is a structural diagram of the forest canopy closure and tree species type simultaneous detection model in this method.

[0117] The process of constructing and training a multi-task neural network model based on multi-task learning also includes:

[0118] When training a neural network model, a shared feature extraction layer is trained first, and then task-specific layers are gradually added for fine-tuning.

[0119] Early stopping mechanisms and model regularization measures are adopted to prevent overfitting, enhance the model's generalization ability, and ensure the stability of the training process.

[0120] The process of constructing and training a multi-task neural network model based on multi-task learning also includes:

[0121] The performance of the model is optimized by methods such as cross-validation, learning rate adjustment, and weight allocation.

[0122] Preferably, the step of constructing and training a multi-task neural network model based on multi-task learning further includes:

[0123] Multiple metrics, including at least precision, recall, F1 score, and overall accuracy, are used to comprehensively evaluate model performance.

[0124] Based on the comprehensive evaluation results, the network model was fine-tuned.

[0125] S3: Acquire satellite multispectral remote sensing image data of the area to be detected.

[0126] In practical applications, this method utilizes satellite multispectral remote sensing imagery of the area to be detected from Sentinel-2. Sentinel-2 is a series of high-resolution multispectral imaging satellites launched by the European Space Agency (ESA). Sentinel-2 is equipped with a state-of-the-art Multispectral Instrument (MSI), a sensor capable of capturing images in 13 different spectral bands, covering the range from visible light to shortwave infrared. Sentinel-2 data is widely used in environmental monitoring, including vegetation cover, soil moisture, and water quality monitoring. Its multispectral data helps to accurately identify land cover types, providing a scientific basis for environmental protection.

[0127] The satellite multispectral remote sensing image data includes spatial image data and distance image data.

[0128] Spatial imagery data can be used to detect forest tree species types, while distance imagery data can be used to detect forest canopy closure.

[0129] The method for simultaneous detection of forest canopy density and tree species type also includes: preprocessing satellite multispectral remote sensing image data of the area to be detected.

[0130] The preprocessing of satellite multispectral remote sensing image data of the area to be detected includes: noise removal, geometric correction, radiometric calibration, atmospheric correction, and image registration.

[0131] S4: Input the satellite multispectral remote sensing image data of the area to be detected into the forest canopy closure and tree species type synchronous detection model to obtain the synchronous detection results of forest canopy closure and tree species type.

[0132] The output of this detection model consists of two parts: forest canopy closure detection results and tree species type detection results. These two parts together constitute a comprehensive analysis of satellite multispectral remote sensing image data, providing rich information for subsequent application scenarios.

[0133] S5: Statistical analysis of the results of forest canopy closure and tree species type detection.

[0134] After obtaining the forest canopy closure and tree species type detection results through neural network models, statistical analysis can be performed on these data.

[0135] For example, after obtaining the forest canopy closure in different regions, a heat map can be used to display the forest canopy closure in each region, thereby enabling an overall analysis of a large area or region.

[0136] For example, statistics can be collected on different tree species in the area to be tested to analyze the main tree species and dominant tree species in the area.

[0137] Preferably, the method further includes:

[0138] S6: Based on the detection results of forest canopy density and tree species type at different time points in the area to be detected, construct a spatiotemporal distribution map of forest canopy density and tree species type in the area to be detected.

[0139] Satellite multispectral remote sensing imagery data of the area to be monitored at different time points can be acquired and input into a synchronous detection model for forest canopy closure and tree species type to obtain synchronous detection results for forest canopy closure and tree species type at different time points in the area. This data can then be integrated to construct a spatiotemporal distribution map showing the changes in forest canopy closure and tree species type in the area and their temporal progression, providing relevant information for environmental monitoring and decision analysis.

[0140] The beneficial effects of this embodiment are as follows: This method can also comprehensively evaluate the network model based on multiple indicators, and fine-tune and improve the network model according to the evaluation results; after obtaining the detection results of forest canopy density and tree species type in the area to be detected through the detection model, statistical analysis of the detection results can further obtain forest-related information and conditions in the area, which is convenient for subsequent ecological protection; based on the detection results of forest canopy density and tree species type, the data is integrated to construct a spatiotemporal distribution map of changes in forest canopy density and tree species type in the area to be detected, which can provide relevant information for environmental monitoring and decision analysis.

[0141] Example 3

[0142] A system for simultaneous detection of forest canopy closure and tree species type, based on the method for simultaneous detection of forest canopy closure and tree species type as described in Embodiment 1 or 2, includes:

[0143] The sample construction module is used to acquire multi-source remote sensing data from UAVs in the test area and construct a sample dataset.

[0144] The model building module is used to build and train a multi-task neural network model based on multi-task learning using the sample dataset to obtain a simultaneous detection model for forest canopy closure and tree species type.

[0145] The data acquisition module is used to acquire satellite multispectral remote sensing image data of the area to be detected;

[0146] The detection module is used to input satellite multispectral remote sensing image data of the area to be detected into the forest canopy closure and tree species type synchronous detection model to obtain the synchronous detection results of forest canopy closure and tree species type.

[0147] Preferably, the system further includes:

[0148] The preprocessing module is used to preprocess the UAV multi-source remote sensing data of the test area and the satellite multispectral remote sensing image data of the area to be tested;

[0149] The statistical analysis module is used to perform statistical analysis on the detection results of forest canopy density and tree species type.

[0150] Preferably, the system further includes:

[0151] The distribution map construction module is used to construct a spatiotemporal distribution map of forest canopy density and tree species type in the area to be detected based on the detection results of forest canopy density and tree species type at different time points in the area to be detected.

[0152] Example 4

[0153] A computer program, when executed, implements the steps of the method for synchronous detection of forest canopy closure and tree species type as described in Embodiment 1 or 2.

[0154] Example 5

[0155] A readable storage medium storing a computer program as described in Embodiment 4, wherein when executed, the computer program implements the steps of the forest canopy closure and tree species type synchronous detection method as described in Embodiment 1 or 2.

[0156] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for simultaneous detection of forest canopy density and tree species type, characterized in that, The method comprises the following steps: obtaining unmanned aerial vehicle multi-source remote sensing data of a test area, and constructing a sample data set; based on the sample data set, constructing a multi-task neural network model based on multi-task learning and training, to obtain a forest canopy density and tree species synchronous detection model; wherein, the construction of the multi-task neural network model comprises: based on a SwinTransformer model, introducing hierarchical feature expression and a self-attention mechanism based on a shift window, to construct a deep shared feature extraction layer; after the shared feature extraction layer, branch networks are constructed for the forest canopy density detection task and the tree species detection task, and the branch networks adopt weight sharing, self-attention and shared attention strategies; the branch networks adopt a multi-task loss function, fuse the weighted loss of the forest canopy density detection task and the tree species detection task, introduce an adjustable weight coefficient, and dynamically adjust the weight according to the task characteristics and complexity; the training comprises: first training the shared feature extraction layer, and then gradually adding task-specific layers for fine-tuning; an early stopping mechanism and model regularization measures are adopted to prevent overfitting, enhance the generalization ability of the model, and ensure the stability of the training process; obtaining satellite multispectral remote sensing image data of a region to be detected; inputting the satellite multispectral remote sensing image data of the region to be detected into the forest canopy density and tree species synchronous detection model, to obtain forest canopy density and tree species synchronous detection results.

2. The method for simultaneous detection of forest canopy density and tree species type according to claim 1, characterized in that, The unmanned aerial vehicle multi-source remote sensing data comprises spatial image data based on CCD high-resolution remote sensing image data and distance image data based on LiDAR scanning data.

3. The method for simultaneous detection of forest canopy density and tree species type according to claim 2, characterized in that, When constructing the sample data set, the spatial image data is labeled for tree species, and the distance image data is labeled for forest canopy density.

4. The method of claim 1, wherein the method further comprises: The forest canopy density and tree species synchronous detection method further comprises: preprocessing the unmanned aerial vehicle multi-source remote sensing data of the test area and the satellite multispectral remote sensing image data of the region to be detected.

5. The method for simultaneous detection of forest canopy density and tree species type according to claim 4, characterized in that, The preprocessing of the unmanned aerial vehicle multi-source remote sensing data of the test area and the satellite multispectral remote sensing image data of the region to be detected comprises: removing noise, geometric correction, radiation calibration, atmospheric correction and image registration.

6. The method of claim 1, wherein the method further comprises: The construction of the multi-task neural network model based on multi-task learning and the training further comprise: adopting cross-validation, learning rate adjustment and weight distribution methods to optimize the performance of the model.

7. The method for simultaneous detection of forest canopy density and tree species type according to claim 6, characterized in that, The construction of the multi-task neural network model based on multi-task learning and the training further comprise: adopting multiple indicators, at least including accuracy, recall rate, F1 score and overall accuracy, to comprehensively evaluate the performance of the model; according to the comprehensive evaluation results, fine-tuning the network model.

8. The method of claim 1, wherein the method further comprises: The forest canopy density and tree species synchronous detection method further comprises the following steps: statistically analyzing the forest canopy density and tree species detection results.

9. The method of claim 1, wherein the method further comprises: The forest canopy density and tree species synchronous detection method further comprises the following steps: based on the forest canopy density and tree species detection results of the region to be detected at different time points, constructing a forest canopy density and tree species spatiotemporal distribution map of the region to be detected.

10. A system for simultaneous detection of forest canopy density and tree species type, characterized by, The forest canopy density and tree species synchronous detection method according to any one of claims 1 to 9, comprising: A sample construction module is configured to acquire unmanned aerial vehicle multi-source remote sensing data of a test area and construct a sample data set; A model construction module is configured to construct a multi-task neural network model based on multi-task learning and train the model according to the sample data set, and obtain a forest canopy density and tree species synchronous detection model; A data acquisition module is configured to acquire satellite multi-spectral remote sensing image data of a region to be detected; A detection module is configured to input the satellite multi-spectral remote sensing image data of the region to be detected into the forest canopy density and tree species synchronous detection model, and obtain a forest canopy density and tree species synchronous detection result.

11. A readable storage medium, characterized by, The readable storage medium has a computer program stored thereon, and the computer program is executed to implement the steps of the forest canopy density and tree species synchronous detection method according to any one of claims 1 to 9.

Images