Method and system for synchronously detecting forest canopy density and tree species type, and computer program

Abstract

The invention discloses a forest canopy density and tree species type synchronous detection method, which comprises the following steps: acquiring unmanned aerial vehicle multi-source remote sensing data of a test area, and constructing a sample data set; according to the sample data set, a multi-task neural network model is constructed based on multi-task learning and is trained, and a forest canopy density and tree species type synchronous detection model is obtained; acquiring satellite multispectral remote sensing image data of the to-be-detected area; and inputting the satellite multispectral remote sensing image data of the to-be-detected area into the forest canopy density and tree species type synchronous detection model to obtain a forest canopy density and tree species type synchronous detection result. According to the technical scheme provided by the invention, synchronous detection of the forest canopy density and the tree species type is realized through the multi-task neural network model; performing fine adjustment and improvement on the network model; performing statistical analysis on a detection result; and constructing a forest canopy density and tree species type change space-time distribution diagram of the to-be-detected area.

Description

Technical Field

[0001] The invention relates to the field of forest image processing, and in particular to a method for synchronously detecting forest canopy density and tree species types. Background Art

[0002] Forest canopy density usually refers to the ratio of the vertical projection area of ​​the tree crown in the forest to the forest area. Canopy density is an important factor reflecting the forest structure and forest environment. It is widely used in soil erosion, water conservation, forest quality evaluation, forest landscape construction, etc. At the same time, it is an important factor in the division of small and medium-sized forest management, determining the intensity of tending and felling, and judging whether it is a forest.

[0003] The identification and classification of forest tree species are of great significance for counting forest resources, exploring the diversity of ecosystems, and even studying the evolution of organisms. Therefore, the accurate identification and statistics of tree species in forest areas are 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] By simultaneously testing the forest canopy density and tree species types, we can obtain more complete and abundant forest-related information and conditions in the area to be tested. By combining tree species types and canopy density, we can more accurately assess the structure and density of the forest stand, making it easier to manage the forest or protect its ecology.

[0005] However, existing technical methods can only calculate the forest density or identify the forest tree species separately, but fail to realize the simultaneous detection of the two. The detection results are single and the application is greatly limited. Summary of the invention

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

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

[0008] A method for synchronously detecting forest canopy density and tree species type comprises the following steps:

[0009] Obtain multi-source remote sensing data from drones in the test area and construct a sample data set;

[0010] According to the sample data set, a multi-task neural network model was constructed and trained based on multi-task learning to obtain a simultaneous detection model for forest density 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 simultaneous detection model of forest canopy density and tree species type to obtain the simultaneous detection results of forest canopy density and tree species type.

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

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

[0015] According to one aspect of the present invention, the method for synchronously detecting forest density and tree species type further includes: preprocessing the multi-source remote sensing data of unmanned aerial vehicles in the test area and the satellite multi-spectral 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 multi-spectral remote sensing image data of the area to be detected includes: noise removal, geometric correction, radiation calibration, atmospheric correction and image registration.

[0017] According to one aspect of the present invention, the multi-task neural network model is constructed based on multi-task learning and trained, including:

[0018] Based on the Swin Transformer model, we introduce hierarchical feature expression and a self-attention mechanism based on a shifted window to construct a deep shared feature extraction layer.

[0019] After the shared feature extraction layer, branch networks are constructed for the forest canopy density 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, integrates the weighted losses of the forest canopy density detection task and the tree species type detection task, introduces an adjustable weight coefficient, and dynamically adjusts the weight according to the task characteristics and complexity.

[0021] According to one aspect of the present invention, the multi-task neural network model is constructed based on multi-task learning and trained further comprising:

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

[0023] The 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.

[0024] According to one aspect of the present invention, the multi-task neural network model is constructed based on multi-task learning and trained further comprising:

[0025] Cross-validation, learning rate adjustment, and weight distribution methods are used to optimize the performance of the model.

[0026] According to one aspect of the present invention, the multi-task neural network model is constructed based on multi-task learning and trained further comprising:

[0027] Use multiple indicators, including at least precision, recall, F1 score, and overall accuracy, to comprehensively evaluate model performance;

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

[0029] According to one aspect of the present invention, the method for synchronously detecting forest density and tree species type further comprises the following steps:

[0030] The detection results of forest canopy density and tree species types were statistically analyzed.

[0031] According to one aspect of the present invention, the method for synchronously detecting forest density and tree species type further comprises the following steps:

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

[0033] A system for synchronous detection of forest canopy density and tree species type, based on the method for synchronous detection of forest canopy density and tree species type as described above, comprises:

[0034] The sample construction module is used to obtain multi-source remote sensing data from drones in the test area and construct a sample data set;

[0035] The model building module is used to build and train a multi-task neural network model based on multi-task learning according to the sample data set to obtain a simultaneous detection model for forest density and tree species type;

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

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

[0038] A computer program, when executed, implements the steps of the method for synchronously detecting forest density and tree species type as described above.

[0039] Advantages of the present invention:

[0040] The present invention provides a method for synchronous detection of forest canopy density and tree species types. By constructing a multi-task neural network model based on multi-task learning, applying Swin Transformer to perform feature extraction, comprehensively extracting different characteristics of multi-source remote sensing data, and performing deep learning, the synchronous detection of forest canopy density and tree species types is achieved.

[0041] This method can also conduct a comprehensive evaluation of the network model based on multiple indicators, and fine-tune and improve the network model according to the evaluation results; after obtaining the forest canopy density and tree species type detection results of the area to be detected through the detection model, the detection results are statistically analyzed to further obtain forest-related information and conditions in the area, which is convenient for subsequent ecological protection, etc.; according to the forest canopy density and tree species type detection results, the data is integrated to construct the spatiotemporal distribution map of the 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-making analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0043] Figure 1 This is a flow chart of a method for synchronously detecting forest density and tree species type according to Embodiment 1 of the present invention;

[0044] Figure 2 It is a structural diagram of the forest canopy density and tree species type synchronous detection model of the present invention;

[0045] Figure 3 This is a flow chart of a method for synchronously detecting forest density and tree species type described in Embodiment 2 of the present invention. DETAILED DESCRIPTION

[0046] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0047] Embodiment 1

[0048] like Figure 1 As shown, a method for synchronously detecting forest canopy density and tree species type comprises the following steps:

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

[0050] The multi-source remote sensing data of the UAV includes space image data based on CCD high-resolution remote sensing image data and distance image data based on LiDAR scanning data.

[0051] CCD high-resolution remote sensing images are remote sensing images with high spatial resolution acquired through charge coupled device (CCD) technology. CCD high-resolution remote sensing images play an important role in environmental monitoring, resource investigation, urban planning, military reconnaissance and other fields due to their high resolution, high temporal resolution, high spectral resolution and rich texture features.

[0052] LiDAR scan data is a type of three-dimensional data acquired by a laser scanner (LiDAR scanner). LiDAR mainly 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 synchronously detecting forest density and tree species type also includes: preprocessing the multi-source remote sensing data of unmanned aerial vehicles in the test area.

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

[0055] Radiometric calibration means that when calculating the spectral reflectance or spectral radiance of an object, or when comparing images acquired at different times and with different sensors, the brightness grayscale value of the image must be converted into absolute radiance.

[0056] Atmospheric correction refers to eliminating the radiation error caused by atmospheric absorption, especially scattering, and inverting the true surface reflectivity of the ground.

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

[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 real situation of the ground objects; image registration allows remote sensing data at different times and under different conditions to be compared and integrated, providing a basis for ground object change monitoring and data analysis.

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

[0060] When constructing a sample dataset, data augmentation is performed on the image data to construct a more extensive sample dataset.

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

[0062] In practical applications, this method tries to establish a large-scale sample database that is diverse and widely distributed in space to support iterative training and parameter optimization of deep learning models. The goal is to increase the training accuracy of the model to more than 85%. For models that do not meet this accuracy standard, detailed inspections will be carried out, and their interpretation signs and labels of change types will be supplemented, training parameters will be adjusted, and iterative training will be repeated to achieve higher accuracy.

[0063] S2: Based on the sample data set, a multi-task neural network model is constructed and trained based on multi-task learning to obtain a simultaneous detection model for forest density and tree species type.

[0064] The multi-task neural network model is constructed based on multi-task learning and trained, including:

[0065] Based on the Swin Transformer model, we introduce hierarchical feature expression and a self-attention mechanism based on a shifted window to construct a deep shared feature extraction layer.

[0066] After the shared feature extraction layer, branch networks are constructed for the forest canopy density 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, integrates the weighted losses of the forest canopy density detection task and the tree species type detection task, introduces an adjustable weight coefficient, and dynamically adjusts the weight according to the task characteristics and complexity.

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

[0069] The multi-task fully convolutional neural network model (MT-FCN) is a deep learning framework that can handle multiple related visual tasks simultaneously. It usually contains a shared feature extraction part and multiple branch networks designed for specific tasks. By sharing the feature extraction part, MT-FCN can reduce repeated calculations, improve efficiency, and improve the performance of each task through the correlation between tasks.

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

[0071] Swin Transformer is an advanced deep learning model designed for computer vision tasks. It combines the powerful capabilities of the Transformer model in sequence data processing with the efficient local feature extraction advantages of convolutional neural networks (CNNs) in image recognition. Swin Transformer solves the computational complexity problem in high-resolution image processing while maintaining high performance by introducing hierarchical feature representation and a self-attention mechanism based on a shifted window.

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

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

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

[0075] The multi-task neural network model is constructed based on multi-task learning and trained further comprising:

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

[0077] The 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.

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

[0079] Cross-validation, learning rate adjustment, weight distribution and other methods are used to optimize the performance of the model.

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

[0081] In practical applications, the satellite multispectral remote sensing image data of the area to be detected used in this method is taken from Sentinel-2. Sentinel-2 is a series of high-resolution multispectral imaging satellites launched by the European Space Agency (ESA). The Sentinel-2 satellite is equipped with a state-of-the-art multispectral imager (MSI), which can capture images in 13 different spectral bands, covering the range from visible light to short-wave infrared. Sentinel-2 data are widely used in the field of environmental monitoring, including vegetation cover, soil moisture, water quality monitoring, etc. Its multispectral data helps to accurately identify the type of land objects and provide a scientific basis for environmental protection.

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

[0083] Spatial image data can be used to detect forest tree species types, and distance image data can be used to detect forest canopy density.

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

[0085] The preprocessing of the satellite multispectral remote sensing image data of the area to be detected includes: noise removal, geometric correction, radiation 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 density and tree species type synchronous detection model to obtain the forest canopy density and tree species type synchronous detection results.

[0087] The output of the detection model of this method consists of two parts: one is the forest canopy density detection result; the other is the tree species type detection result. These two parts of the output together constitute a comprehensive analysis of the model on satellite multispectral remote sensing image data, which can provide rich information for subsequent application scenarios.

[0088] The beneficial effect of this embodiment is that the method constructs a multi-task neural network model based on multi-task learning, applies Swin Transformer for feature extraction, comprehensively extracts different characteristics of multi-source remote sensing data, performs deep learning, and realizes simultaneous detection of forest density and tree species types.

[0089] Embodiment 2

[0090] like Figure 3 As shown, a method for synchronously detecting forest canopy density and tree species type comprises the following steps:

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

[0092] The multi-source remote sensing data of the UAV includes space image data based on CCD high-resolution remote sensing image data and distance image data based on LiDAR scanning data.

[0093] CCD high-resolution remote sensing images are remote sensing images with high spatial resolution acquired through charge coupled device (CCD) technology. CCD high-resolution remote sensing images play an important role in environmental monitoring, resource investigation, urban planning, military reconnaissance and other fields due to their high resolution, high temporal resolution, high spectral resolution and rich texture features.

[0094] LiDAR scan data is a type of three-dimensional data acquired by a laser scanner (LiDAR scanner). LiDAR mainly 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 synchronously detecting forest density and tree species type also includes: preprocessing the multi-source remote sensing data of unmanned aerial vehicles in the test area.

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

[0097] Radiometric calibration means that when calculating the spectral reflectance or spectral radiance of an object, or when comparing images acquired at different times and with different sensors, the brightness grayscale value of the image must be converted into absolute radiance.

[0098] Atmospheric correction refers to eliminating the radiation error caused by atmospheric absorption, especially scattering, and inverting the true surface reflectivity of the ground.

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

[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 real situation of the ground objects; image registration allows remote sensing data at different times and under different conditions to be compared and integrated, providing a basis for ground object change monitoring and data analysis.

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

[0102] When constructing a sample dataset, data augmentation is performed on image data to construct a more extensive sample dataset.

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

[0104] In practical applications, this method tries to establish a large-scale sample database that is diverse and widely distributed in space to support iterative training and parameter optimization of deep learning models. The goal is to increase the training accuracy of the model to more than 85%. For models that do not meet this accuracy standard, detailed inspections will be carried out, and their interpretation signs and labels of change types will be supplemented, training parameters will be adjusted, and iterative training will be repeated to achieve higher accuracy.

[0105] S2: Based on the sample data set, a multi-task neural network model is constructed and trained based on multi-task learning to obtain a simultaneous detection model for forest density and tree species type.

[0106] The multi-task neural network model is constructed based on multi-task learning and trained, including:

[0107] Based on the Swin Transformer model, we introduce hierarchical feature expression and a self-attention mechanism based on a shifted window to construct a deep shared feature extraction layer.

[0108] After the shared feature extraction layer, branch networks are constructed for the forest canopy density 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, integrates the weighted losses of the forest canopy density detection task and the tree species type detection task, introduces an adjustable weight coefficient, and dynamically adjusts the weight according to the task characteristics and complexity.

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

[0111] The multi-task fully convolutional neural network model (MT-FCN) is a deep learning framework that can handle multiple related visual tasks simultaneously. It usually contains a shared feature extraction part and multiple branch networks designed for specific tasks. By sharing the feature extraction part, MT-FCN can reduce repeated calculations, improve efficiency, and improve the performance of each task through the correlation between tasks.

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

[0113] Swin Transformer is an advanced deep learning model designed for computer vision tasks. It combines the powerful capabilities of the Transformer model in sequence data processing with the efficient local feature extraction advantages of convolutional neural networks (CNNs) in image recognition. Swin Transformer solves the computational complexity problem in high-resolution image processing while maintaining high performance by introducing hierarchical feature representation and a self-attention mechanism based on a shifted window.

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

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

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

[0117] The multi-task neural network model is constructed based on multi-task learning and trained further comprising:

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

[0119] The 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.

[0120] The multi-task neural network model is constructed based on multi-task learning and trained further comprising:

[0121] Cross-validation, learning rate adjustment, weight distribution and other methods are used to optimize the performance of the model.

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

[0123] Use multiple indicators, including at least precision, recall, F1 score, and overall accuracy, to comprehensively evaluate model performance;

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

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

[0126] In practical applications, the satellite multispectral remote sensing image data of the area to be detected used in this method is taken from Sentinel-2. Sentinel-2 is a series of high-resolution multispectral imaging satellites launched by the European Space Agency (ESA). The Sentinel-2 satellite is equipped with a state-of-the-art multispectral imager (MSI), which can capture images in 13 different spectral bands, covering the range from visible light to short-wave infrared. Sentinel-2 data are widely used in the field of environmental monitoring, including vegetation cover, soil moisture, water quality monitoring, etc. Its multispectral data helps to accurately identify the type of land objects and provide a scientific basis for environmental protection.

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

[0128] Spatial image data can be used to detect forest tree species types, and distance image data can be used to detect forest canopy density.

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

[0130] The preprocessing of the satellite multispectral remote sensing image data of the area to be detected includes: noise removal, geometric correction, radiation 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 density and tree species type synchronous detection model to obtain the forest canopy density and tree species type synchronous detection results.

[0132] The output of the detection model of this method consists of two parts: one is the forest canopy density detection result; the other is the tree species type detection result. These two parts of the output together constitute the model's comprehensive analysis of satellite multispectral remote sensing image data, which can provide rich information for subsequent application scenarios.

[0133] S5: Conduct statistical analysis on the detection results of forest density and tree species types.

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

[0135] For example, after obtaining the forest density in different regions, the forest density in each region can be displayed in the form of a heat map, thereby conducting an overall analysis of a large range or region.

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

[0137] Preferably, the method further comprises:

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

[0139] Satellite multispectral remote sensing image data of the area to be detected at different time points can be obtained, and input into the forest canopy density and tree species type synchronous detection model to obtain the forest canopy density and tree species type synchronous detection results at different time points in the area. After that, these data are integrated to construct a spatiotemporal distribution map showing the distribution of changes in forest canopy density and tree species types in the area and their time course, which can provide relevant information for environmental monitoring and decision-making analysis.

[0140] The beneficial effects of this embodiment are: this method can also conduct a comprehensive evaluation of the network model based on multiple indicators, and fine-tune and improve the network model according to the evaluation results; after obtaining the forest canopy density and tree species type detection results of the area to be detected through the detection model, the detection results are statistically analyzed to further obtain forest-related information and conditions in the area, which is convenient for subsequent ecological protection, etc.; according to the forest canopy density and tree species type detection results, the data is integrated to construct a spatiotemporal distribution map of the 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-making analysis.

[0141] Embodiment 3

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

[0143] The sample construction module is used to obtain multi-source remote sensing data from drones in the test area and construct a sample data set;

[0144] The model building module is used to build and train a multi-task neural network model based on multi-task learning according to the sample data set to obtain a simultaneous detection model for forest density and tree species type;

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

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

[0147] Preferably, the system further comprises:

[0148] A preprocessing module is used to preprocess the multi-source remote sensing data of UAVs in the test area and the satellite multi-spectral remote sensing image data of the area to be detected;

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

[0150] Preferably, the system further comprises:

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

[0152] Embodiment 4

[0153] A computer program, when executed, implements the steps of the method for synchronously detecting forest density and tree species type as described in embodiment one or two.

[0154] Embodiment 5

[0155] A readable storage medium stores a computer program as described in Example 4, and when the computer program is executed, the steps of the method for synchronously detecting forest density and tree species type as described in Example 1 or 2 are implemented.

[0156] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by any technician familiar with the art within the technical scope disclosed in the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A method for synchronously detecting forest density and tree species type, characterized in that: The following steps are involved: Obtain multi-source remote sensing data from drones in the test area and construct a sample data set; According to the sample data set, a multi-task neural network model was constructed and trained based on multi-task learning to obtain a simultaneous detection model for forest density and tree species type. Acquire satellite multispectral remote sensing image data of the area to be detected; The satellite multispectral remote sensing image data of the area to be detected is input into the simultaneous detection model of forest canopy density and tree species type to obtain the simultaneous detection results of forest canopy density and tree species type.

2. The method for synchronously detecting forest density and tree species type according to claim 1, characterized in that: The multi-source remote sensing data of the UAV includes space image data based on CCD high-resolution remote sensing image data and distance image data based on LiDAR scanning data.

3. The method for synchronously detecting forest density and tree species type according to claim 2, characterized in that: When constructing the sample dataset, the tree species type is calibrated for the spatial image data, and the forest canopy density is calibrated for the distance image data.

4. The method for synchronously detecting forest density and tree species type according to claim 1, characterized in that: The method for synchronously detecting forest density and tree species types also includes: preprocessing the multi-source remote sensing data of unmanned aerial vehicles in the test area and the satellite multi-spectral remote sensing image data of the area to be detected.

5. The method for synchronously detecting forest density and tree species type according to claim 1, characterized in that: The preprocessing of the UAV multi-source remote sensing data of the test area and the satellite multi-spectral remote sensing image data of the area to be detected includes: noise removal, geometric correction, radiation calibration, atmospheric correction and image registration.

6. The method for synchronously detecting forest density and tree species type according to claim 1, characterized in that: The multi-task neural network model is constructed based on multi-task learning and trained, including: Based on the Swin Transformer model, we introduce hierarchical feature expression and a self-attention mechanism based on a shifted 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 type detection task respectively. The branch networks adopt weight sharing, self-attention and shared attention strategies. The branch network adopts a multi-task loss function, integrates the weighted losses of the forest canopy density detection task and the tree species type detection task, introduces an adjustable weight coefficient, and dynamically adjusts the weight according to the task characteristics and complexity.

7. The method for synchronously detecting forest density and tree species type according to claim 6, characterized in that: The multi-task neural network model is constructed based on multi-task learning and trained further comprising: When training the neural network model, the shared feature extraction layer is trained first, and then the task-specific layers are gradually added for fine-tuning; The 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.

8. The method for synchronously detecting forest density and tree species type according to claim 7, characterized in that: The multi-task neural network model is constructed based on multi-task learning and trained further comprising: Cross-validation, learning rate adjustment, and weight distribution methods are used to optimize the performance of the model.

9. The method for synchronously detecting forest density and tree species type according to claim 8, characterized in that: The multi-task neural network model is constructed based on multi-task learning and trained further comprising: Use multiple indicators, including at least precision, recall, F1 score, and overall accuracy, to comprehensively evaluate model performance; Based on the comprehensive evaluation results, the network model is fine-tuned.

10. The method for synchronously detecting forest density and tree species type according to claim 1, characterized in that: The method for synchronously detecting forest density and tree species type also includes the following steps: The detection results of forest canopy density and tree species types were statistically analyzed.

11. The method for synchronously detecting forest density and tree species type according to claim 1, characterized in that: The method for synchronously detecting forest density and tree species type also includes the following steps: According to the detection results of forest canopy density and tree species types at different time points in the area to be detected, a spatiotemporal distribution map of forest canopy density and tree species types in the area to be detected is constructed.

12. A synchronous detection system for forest density and tree species type, characterized in that: The method for synchronously detecting forest canopy density and tree species type according to any one of claims 1 to 11 comprises: The sample construction module is used to obtain multi-source remote sensing data from drones in the test area and construct a sample data set; The model building module is used to build and train a multi-task neural network model based on multi-task learning according to the sample data set to obtain a simultaneous detection model for forest density and tree species type; A data acquisition module, used to acquire satellite multispectral remote sensing image data of the area to be detected; The detection module is used to input the satellite multispectral remote sensing image data of the area to be detected into the forest canopy density and tree species type synchronous detection model to obtain the forest canopy density and tree species type synchronous detection results.

13. A computer program, characterized in that When the computer program is executed, the steps of the method for synchronously detecting forest density and tree species type as claimed in any one of claims 1 to 11 are implemented.

Images