Grassland aboveground biomass measurement method and device based on unmanned aerial vehicle remote sensing image

Abstract

This application provides a method and apparatus for measuring grassland biomass based on UAV remote sensing images. The method involves identifying relevant vegetation regions of grassland organisms from remote sensing images of a target grassland area, performing local texture enhancement on these vegetation regions to obtain texture enhancement features of grassland organisms in the remote sensing images, identifying grassland vegetation spectral features from the remote sensing images, and determining a multi-source feature fusion map of grassland organisms based on these spectral features and texture enhancement features. Calibration constraints for grassland biomass in the target grassland area are determined using historical grassland vegetation spectral features and grassland vegetation spectral features from historical remote sensing images of the target grassland area. Finally, a biomass measurement map of the grassland in the target grassland area is calculated based on the multi-source feature fusion map and calibration constraints. Using this method, grassland biomass can be collaboratively measured based on the texture structure and spectral information of grassland vegetation.

Description

Technical Field

[0001] This application relates to the field of biomass measurement technology, and more specifically, to a method and apparatus for measuring grassland biomass based on UAV remote sensing images. Background Technology

[0002] Biomass measurement refers to a method for quantitatively assessing the total amount of biological organisms in a specific time and region. It is usually expressed as dry weight or carbon storage per unit area or volume. It is widely used in fields such as ecology, forestry, agriculture and climate change research. It quantifies the biomass of plants, animals or microorganisms through direct sampling, remote sensing technology or model estimation.

[0003] Grassland, as an important component of terrestrial ecosystems, relies heavily on aboveground biomass as a core indicator for assessing its ecological health, productivity, and carbon sequestration capacity. This biomass is crucial for the rational utilization of grassland resources, ecological protection and restoration, and the sustainable development of animal husbandry. Traditional grassland aboveground biomass measurement largely depends on in-situ sampling and weighing. While this method offers high accuracy, it suffers from drawbacks such as being time-consuming, labor-intensive, having limited coverage, being highly destructive, and unable to quickly acquire regional-scale data. Existing satellite remote sensing methods, while capable of large-scale monitoring, are limited by spatial resolution, making it difficult to accurately capture the detailed vegetation features of small-scale grasslands, thus limiting measurement accuracy. Unmanned aerial vehicle (UAV) remote sensing technology, with its advantages of high resolution, flexibility, convenience, and low cost, is increasingly being applied to grassland monitoring. However, current UAV-based remote sensing methods often rely solely on spectral features, neglecting the auxiliary role of grassland vegetation texture details in biomass measurement. Furthermore, they fail to consider the spatiotemporal differences in spectral characteristics of remote sensing data from different periods, resulting in measurement accuracy and stability that cannot meet practical application requirements. Therefore, how to collaboratively measure grassland aboveground biomass based on the texture structure and spectral information of grassland vegetation has become a challenge for the industry. Summary of the Invention

[0004] This application provides a method and apparatus for measuring grassland biomass based on UAV remote sensing images, which can collaboratively measure grassland biomass based on the texture structure and spectral information of grassland vegetation.

[0005] In a first aspect, this application provides a method for measuring grassland biomass based on UAV remote sensing images, comprising the following steps: Remote sensing images of the target grassland area were collected using drones; Based on historical vegetation information of the target grassland area, the relevant vegetation areas of grassland organisms are identified from the remote sensing image, and local texture enhancement is performed on the relevant vegetation areas in the remote sensing image to obtain the texture enhancement features of grassland organisms in the remote sensing image. Grassland vegetation spectral features are identified from the remote sensing image, and the grassland vegetation spectral features and the texture-enhanced vegetation features are interactively fused to obtain a multi-source feature fusion map of grassland organisms. Historical remote sensing images of the target grassland area are acquired, and the historical grassland vegetation spectral features in the historical remote sensing images and the grassland vegetation spectral features are subjected to consistency constraints to obtain calibration constraints on the aboveground biomass of the grassland in the target grassland area. Based on the multi-source feature fusion map and the calibration constraints, the biomass of grassland in the target grassland area is calculated collaboratively to obtain a biomass calculation map of grassland in the target grassland area.

[0006] In some embodiments, identifying the relevant vegetation areas of grassland organisms from the remote sensing image based on historical vegetation information of the target grassland area specifically includes: Obtain historical vegetation information for the target grassland area; The remote sensing image is preprocessed to obtain a preprocessed remote sensing image; Using the historical vegetation information as a vegetation reference, the relevant vegetation areas of grassland organisms are identified from the preprocessed remote sensing images.

[0007] In some embodiments, performing local texture enhancement on relevant vegetation areas in the remote sensing image to obtain texture enhancement features of organisms on grassland in the remote sensing image specifically includes: Convert the relevant vegetation areas in the remote sensing image into grayscale images; The grayscale image is locally enhanced by sliding to obtain the enhanced image of the relevant vegetation area; The enhanced image is fused with the remote sensing image to obtain the texture enhancement features of organisms on the grassland in the remote sensing image.

[0008] In some embodiments, identifying grassland vegetation spectral features from the remote sensing image specifically includes: Extract the regional reflectance data of the relevant vegetation areas in the remote sensing image; The spectral characteristics of grassland vegetation are determined based on the reflectance data of the region.

[0009] In some embodiments, the interactive fusion of the grassland vegetation spectral features and the texture-enhanced vegetation features to obtain a multi-source feature fusion map of grassland aboveground organisms specifically includes: Spatial registration is performed between the grassland vegetation spectral features and the texture-enhanced vegetation features to obtain spatial registration features; The spatial registration features are normalized to obtain normalized spatial registration features; The normalized spatial registration features are fused to obtain a multi-source feature fusion map of grassland organisms.

[0010] In some embodiments, the consistency constraint between the historical grassland vegetation spectral features in the historical remote sensing image and the grassland vegetation spectral features to obtain the calibration constraint for grassland aboveground biomass in the target grassland area specifically includes: Extract the spectral features of historical grassland vegetation from the historical remote sensing images; Determine the deviation values ​​of the historical grassland vegetation spectral characteristics and the grassland vegetation spectral characteristics; The deviation data are fitted and analyzed to construct calibration constraints for grassland biomass in the target grassland area.

[0011] In some embodiments, the biomass of grassland in the target grassland area is jointly measured based on the multi-source feature fusion map and the calibration constraints to obtain a biomass measurement map of grassland in the target grassland area, specifically including: Based on the multi-source feature fusion map, determine the basic budget value data of grassland aboveground biomass in the target grassland area; The biomass calibration amount for each pixel is determined based on the calibration constraints. The basic budget value data is collaboratively corrected by various biomass calibration values ​​to obtain the biomass correction calculation value for each pixel. Based on the biomass correction calculation value of each pixel, a biomass calculation map of the grassland in the target grassland area is obtained.

[0012] Secondly, this application provides a grassland biomass measurement device based on UAV remote sensing images, comprising: The acquisition module is used to acquire remote sensing images of the target grassland area via drone; The processing module is used to identify the relevant vegetation areas of grassland organisms in the remote sensing image based on the historical vegetation information of the target grassland area, and to perform local texture enhancement on the relevant vegetation areas in the remote sensing image, thereby obtaining the texture enhancement features of grassland organisms in the remote sensing image. The processing module is also used to identify grassland vegetation spectral features from the remote sensing image, and to interactively fuse the grassland vegetation spectral features and the texture-enhanced vegetation features to obtain a multi-source feature fusion map of grassland organisms. The processing module is also used to acquire historical remote sensing images of the target grassland area, and to apply consistency constraints to the historical grassland vegetation spectral features in the historical remote sensing images and the grassland vegetation spectral features to obtain calibration constraints on the aboveground biomass of the grassland in the target grassland area. The execution module is used to collaboratively calculate the biomass of grassland in the target grassland area based on the multi-source feature fusion map and the calibration constraints, so as to obtain a biomass calculation map of grassland in the target grassland area.

[0013] Thirdly, this application provides a computer device, the computer device including a memory and a processor, the memory storing code, and the processor being configured to acquire the code and execute the above-described method for measuring grassland biomass based on UAV remote sensing images.

[0014] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for measuring grassland biomass based on UAV remote sensing images.

[0015] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects: The method and apparatus for grassland biomass measurement based on UAV remote sensing images provided in this application firstly acquires remote sensing images of a target grassland area using a UAV; based on historical vegetation information of the target grassland area, relevant vegetation regions of grassland organisms are identified from the remote sensing images, and local texture enhancement is performed on the relevant vegetation regions in the remote sensing images to obtain texture enhancement features of grassland organisms in the remote sensing images; grassland vegetation spectral features are identified from the remote sensing images, and the grassland vegetation spectral features and texture enhancement vegetation features are interactively fused to obtain a multi-source feature fusion map of grassland organisms; historical remote sensing images of the target grassland area are acquired, and consistency constraints are applied to the historical grassland vegetation spectral features and the grassland vegetation spectral features in the historical remote sensing images to obtain calibration constraints for grassland biomass in the target grassland area; based on the multi-source feature fusion map and the calibration constraints, the biomass of grassland in the target grassland area is collaboratively measured to obtain a biomass measurement map of grassland in the target grassland area.

[0016] Therefore, this application demonstrates that in the process of grassland biomass measurement, it acquires remote sensing images of the target grassland area using UAV remote sensing images. The vegetation area identification and texture enhancement steps leverage historical vegetation information to accurately select the target area and avoid interference from non-vegetated areas. Simultaneously, local texture enhancement strengthens the structural characteristics of grassland vegetation, solving the problem of blurred grassland texture information and its susceptibility to background noise in natural scenes. The spectral feature extraction and interactive fusion steps achieve complementary advantages between texture structure information and spectral information through multi-source feature fusion. Texture features reflect the morphological density differences of grassland vegetation, while spectral features characterize the physiological growth state of vegetation. The fusion of these two features overcomes the limitation that a single texture or spectral feature cannot comprehensively depict biomass-related information. The historical spectral consistency constraint step effectively offsets the interference of environmental factors such as light and weather on real-time spectral features through spatiotemporal spectral feature calibration, solving the industry problem of decreased measurement accuracy caused by spectral feature fluctuations in dynamic environments. Finally, the collaborative measurement step integrates multi-source fusion features and calibration constraints to form a closed-loop measurement mechanism, fully leveraging the synergistic gain effect of texture and spectral information. Compared to traditional single-feature measurement methods, this approach can achieve accurate measurement from local to global perspectives. Using the above scheme, grassland aboveground biomass can be calculated collaboratively based on the texture structure and spectral information of grassland vegetation. Attached Figure Description

[0017] Figure 1 This is an exemplary flowchart illustrating a method for measuring grassland biomass based on UAV remote sensing images, according to some embodiments of this application. Figure 2 This is an exemplary flowchart illustrating the determination of relevant vegetation areas according to some embodiments of this application; Figure 3 This is an exemplary flowchart illustrating the determination of a multi-source feature fusion map according to some embodiments of this application; Figure 4 This is a schematic diagram of a grassland biomass measurement device based on UAV remote sensing images, according to some embodiments of this application. Figure 5 This is a schematic diagram of the structure of a computer device for implementing a method for measuring grassland biomass based on UAV remote sensing images, according to some embodiments of this application. Detailed Implementation

[0018] To better understand the technical solution of this application, the technical solution of this application will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0019] refer to Figure 1The figure is an exemplary flowchart of a grassland biomass measurement method based on UAV remote sensing images, according to some embodiments of this application. The grassland biomass measurement method based on UAV remote sensing images mainly includes the following steps: In step 101, remote sensing images of the target grassland area are acquired using a drone.

[0020] In practice, a multi-rotor UAV with a multispectral camera is selected. First, based on the area and terrain features of the target grassland area, the flight altitude of 50-100 meters, the flight speed of no more than 8 meters per second, the forward overlap rate of ≥80%, and the lateral overlap rate of ≥60% are set. At the same time, the flight operation is carried out in a clear weather with no strong wind and avoiding backlight and large areas of shadow. Before the operation, the UAV positioning module needs to be calibrated and the parameters such as camera focal length and exposure time need to be adjusted. Then, the UAV automatically flies and collects data according to the preset flight path. During the flight, the location metadata such as the shooting time, latitude and longitude, and altitude of each remote sensing image are recorded simultaneously. Finally, remote sensing images without obvious distortion and noise are obtained to ensure that the images can clearly present the canopy morphology and distribution characteristics of the grassland vegetation.

[0021] It should be noted that the remote sensing images in this application represent digital imaging data generated by an optical sensor mounted on a drone taking pictures of a target grassland area from high altitude. They reflect the apparent characteristics of the grassland vegetation, such as canopy morphology, cover density, and color differences. They also indirectly reflect the physiological growth status of the vegetation through reflectance data of different bands. In addition, they can also reflect the topographic relief of the area and the distribution of non-grassland areas such as bare soil, water bodies, and artificial facilities. The remote sensing images include pixel array data that constitute the main body of the image, multi-band spectral information collected by the sensor, and metadata used for geolocation and data calibration, such as shooting time, latitude and longitude coordinates, flight altitude, camera focal length, and exposure parameters.

[0022] In step 102, based on the historical vegetation information of the target grassland area, the relevant vegetation areas of grassland organisms are identified from the remote sensing image, and the relevant vegetation areas in the remote sensing image are locally textured to obtain the texture enhancement features of grassland organisms in the remote sensing image.

[0023] In some embodiments, reference Figure 2 As shown, this figure is an exemplary flowchart for determining relevant vegetation areas in some embodiments of this application. In this embodiment, the identification of relevant vegetation areas of grassland organisms from the remote sensing image based on historical vegetation information of the target grassland area can be achieved by the following steps: In step 1021, historical vegetation information of the target grassland area is obtained; In step 1022, the remote sensing image is preprocessed to obtain a preprocessed remote sensing image; In step 1023, the historical vegetation information is used as a vegetation reference to identify the relevant vegetation areas of grassland organisms from the preprocessed remote sensing image.

[0024] It should be noted that the historical vegetation information in this application refers to the set of vegetation-related data in the target grassland area over the past 3-5 years. It reflects the long-term growth and evolution of grassland vegetation in the area, the community structure characteristics of different growing seasons, and the intrinsic relationship between vegetation spectral reflectance characteristics and aboveground biomass. It also reflects the distribution changes of non-grassland areas such as bare soil and water bodies in the area, as well as the influence of meteorological conditions, human activities and other factors on vegetation growth. The historical vegetation information includes a historical vegetation geographic distribution boundary vector layer generated by remote sensing interpretation, vegetation coverage classification maps for each year, multispectral reflectance threshold ranges of typical grass species, a list of dominant grass species and community proportion data obtained from field surveys, measured aboveground biomass data in different regions, vegetation canopy morphology parameters, and corresponding meteorological factor data and records of human interference factors for the corresponding time period.

[0025] In specific implementation, the remote sensing image is preprocessed to obtain the preprocessed remote sensing image. This can be achieved in the following way: First, based on the factory radiometric calibration coefficient of the sensor carried by the UAV, the digital quantization value of the image is converted into the atmospheric top reflectance. Then, atmospheric correction is performed using the dark pixel method to eliminate the influence of atmospheric scattering and absorption on image quality. Next, at least 16 ground control points are selected evenly distributed within the target grassland area. Geometric correction is performed using a cubic polynomial transformation model combined with the nearest neighbor interpolation method to accurately match the geographic coordinates of the image to the Gauss-Kruger projection coordinate system. The geometric accuracy of the corrected image is controlled within 1 pixel. Finally, a 3×3 window median filtering algorithm is used to suppress noise in the corrected image, filtering out impulse noise and salt-and-pepper noise caused by sensor jitter during UAV flight, while preserving the texture details of the vegetation canopy. This yields the preprocessed remote sensing image. Other methods can also be used in other embodiments, which are not limited here.

[0026] In addition, in specific implementation, using the historical vegetation information as a vegetation reference, the identification of relevant vegetation areas of grassland organisms from the preprocessed remote sensing image can be achieved in the following way: using historical vegetation information as a vegetation reference, a combination of threshold screening, spatial overlay analysis, and spectral similarity analysis is used to identify vegetation areas related to grassland biomass from the preprocessed remote sensing image. First, the reflectance values ​​of the red, green, blue, and near-infrared bands of the preprocessed image are extracted pixel by pixel, and these values ​​are compared one by one with the reflectance threshold ranges of each band of historical vegetation. Candidate vegetation pixels that simultaneously meet the threshold conditions of each band are selected. Then, the historical grassland vegetation geographical distribution boundary vector layer is imported, and the results are processed by... The vector raster overlay tool of the geographic information system clips the candidate vegetation pixel raster and the boundary vector layer, removing non-grassland pixels such as bare soil, water bodies, and buildings outside the boundary. Finally, it calculates the cosine similarity between the spectral curve of the candidate pixels and the typical spectral curve of the historical dominant grass species. Pixels with a similarity ≥ 0.8 are finally identified as vegetation areas related to grassland organisms. The correlation between the similarity value and the measured biomass is verified. After multiple statistical tests, it was found that when the similarity is ≥ 0.8, the deviation between the biomass corresponding to the identified vegetation area and the measured value can be controlled within an acceptable range of ±10%. At the same time, it can take into account the accuracy and false negative rate of vegetation area identification. Other methods can be used in other embodiments, which are not limited here.

[0027] It should be noted that the relevant vegetation area in this application refers to a specific spatial range of herbaceous vegetation that is directly related to grassland organisms. This area reflects not only the actual distribution pattern of herbaceous vegetation within the target grassland area, the concentrated growth area of ​​dominant grass species, and the density of the vegetation canopy, but also the spatial heterogeneity of the spectral reflectance characteristics and texture features of the vegetation.

[0028] In some embodiments, local texture enhancement of relevant vegetation areas in the remote sensing image to obtain texture enhancement features of organisms on grassland in the remote sensing image can be achieved through the following steps: Convert the relevant vegetation areas in the remote sensing image into grayscale images; The grayscale image is locally enhanced by sliding to obtain the enhanced image of the relevant vegetation area; The enhanced image is fused with the remote sensing image to obtain the texture enhancement features of organisms on the grassland in the remote sensing image.

[0029] In specific implementation, converting the relevant vegetation area in the remote sensing image into a grayscale image can be achieved in the following way: First, extract the surface reflectance data of the red light band and near-infrared band in the relevant vegetation area of ​​the remote sensing image after radiometric calibration and atmospheric correction. These two bands are highly sensitive to the vegetation growth status. Among them, the red light band is easily absorbed by vegetation chlorophyll and has low reflectance, while the near-infrared band is easily reflected by the vegetation leaf structure and has high reflectance. Perform pixel-by-pixel difference operation on the two bands. The operation formula is: difference pixel value = near-infrared band reflectance value - red light band reflectance value. After the operation, first remove abnormal pixels with negative difference values. These pixels are mostly residual non-vegetation noise points. Then, count the global minimum and global maximum values ​​of the remaining effective difference pixel values. Use a linear stretching algorithm to normalize the difference pixel values ​​to the 0-255 grayscale range of the 8-bit grayscale image to obtain the grayscale image. Other methods can also be used in other embodiments, which are not limited here.

[0030] In addition, in specific implementation, the local sliding enhancement of the grayscale image to obtain the enhanced image of the relevant vegetation area can be achieved in the following way: a 3×3 pixel local sliding window is selected and the entire grayscale image is traversed with a pixel-by-pixel step size. For the pixel set in each window, a grayscale co-occurrence matrix is ​​first constructed with four directions (0°, 45°, 90°, and 135°) and a step size of 1 pixel. From this matrix, four core texture parameters that can characterize the density of the vegetation canopy and the arrangement of branches are extracted: contrast, entropy, correlation, and homogeneity. Then, a 3×3 Laplacian edge enhancement operator is used, with a convolution kernel value of [0,1,0;1,-4,1;0,1,0]. Convolution operation is performed on each pixel in the window to calculate the pixel grayscale gradient change value, thereby enhancing the vegetation area. The image features fine texture edges formed by overlapping canopy leaves and intertwined branches. Simultaneously, the enhancement weight is adaptively adjusted based on the standard deviation of pixel grayscale values ​​within each window, specifically divided into three gradient intervals: when the standard deviation is ≥25, it is identified as a dense grass area, and the enhancement coefficient is increased to 1.3 to highlight texture details; when the standard deviation is 10 ≤ standard deviation <25, it is identified as a medium-density grass area, and a base enhancement coefficient of 1.0 is used; when the standard deviation is <10, it is identified as a sparse grass area, and the enhancement coefficient is reduced to 0.7 to avoid excessive enhancement of background noise. After enhancement, the image pixel values ​​are normalized a second time to ensure that the values ​​remain within the 0-255 range, resulting in an enhanced image with prominent texture details. Other methods can be used in other embodiments, which are not limited here.

[0031] In addition, in specific implementation, fusing the enhanced image with the remote sensing image to obtain the texture enhancement features of grassland organisms in the remote sensing image can be achieved in the following way: a weighted fusion algorithm is used to fuse the enhanced image with the relevant vegetation areas in the original remote sensing image, selecting the near-infrared band for fusion. The weight of the enhanced image is set to 0.6 to focus on preserving the texture enhancement effect, and the weight of the corresponding band in the original remote sensing image is set to 0.4 to preserve the original spectral information of the vegetation. The fusion is completed by performing the calculation formula "fusion pixel value = 0.6 × enhanced image pixel value + 0.4 × original remote sensing image corresponding band pixel value" pixel by pixel. Then, the fused pixel values ​​are mapped to the standard grayscale range of 0-255, finally obtaining a clear image. Texture enhancement feature maps that clearly distinguish the differences in texture between dense and sparse grass and show a significant correlation with grassland biomass are generated. This involves collecting historical measured biomass data of the target grassland area, along with corresponding texture enhancement images and original remote sensing images. Multiple weight ratios are then set between the enhanced and original remote sensing images, such as 0.2:0.8, 0.3:0.7, 0.4:0.6…0.8:0.2. Different fused feature maps are obtained pixel-by-pixel according to each weight. The correlation coefficient between each fused feature map and the measured biomass is then calculated. Finally, the ratio of 0.6:0.4, which has the highest correlation coefficient and the smallest measurement deviation, is selected as the fusion weight between the enhanced and original remote sensing images. Other methods can be used in other embodiments, which are not limited here.

[0032] It should be noted that the grayscale image in this application represents a digital image in which the multispectral information of the relevant vegetation area in the remote sensing image is converted into a single grayscale value, reflecting the basic light and dark differences and texture contours of the vegetation canopy, and can initially present the density distribution trend of the grass; the enhanced image represents the image obtained after the relevant vegetation area in the image has been enhanced, highlighting the subtle texture details formed by the arrangement of leaves and the interlacing of branches in the vegetation canopy, reflecting the texture boundary differences of grass with different densities, and enhancing the distinction between dense and sparse grass; the texture enhancement feature represents the set of texture parameters extracted after the relevant vegetation area in the remote sensing image has been enhanced, reflecting the intrinsic relationship between vegetation texture features and aboveground biomass.

[0033] In step 103, grassland vegetation spectral features are identified from the remote sensing image, and the grassland vegetation spectral features and the texture-enhanced vegetation features are interactively fused to obtain a multi-source feature fusion map of grassland organisms.

[0034] In some embodiments, identifying the spectral features of grassland vegetation from the remote sensing image can be achieved using the following steps: Extract the regional reflectance data of the relevant vegetation areas in the remote sensing image; The spectral characteristics of grassland vegetation are determined based on the reflectance data of the region.

[0035] In specific implementation, the regional reflectance data of the relevant vegetation area in the remote sensing image can be extracted in the following way: extract the surface reflectance data of the relevant vegetation area in the remote sensing image pixel by pixel in the inner red band, green band, blue band and near infrared band, and simultaneously remove abnormal pixels with reflectance values ​​less than 0 or greater than 1 to obtain the regional reflectance data of the relevant vegetation area. Other methods can also be used in other embodiments, which are not limited here.

[0036] In addition, in specific implementation, the determination of grassland vegetation spectral characteristics based on the regional reflectance data can be achieved in the following way: the vegetation spectral characteristic index calculation method is then used to calculate the normalized vegetation index, ratio vegetation index, and difference vegetation index pixel by pixel based on the regional reflectance dataset. The calculation formulas are: normalized vegetation index = (near-infrared band reflectance - red band reflectance) / (near-infrared band reflectance + red band reflectance), ratio vegetation index = near-infrared band reflectance / red band reflectance, and difference vegetation index = near-infrared band reflectance - red band reflectance. The above indices and regional reflectance data are used as grassland vegetation spectral characteristics. Other methods can also be used in other embodiments, which are not limited here.

[0037] It should be noted that the regional reflectance data in this application represents the set of surface reflectance values ​​for specific bands of red, green, blue, and near-infrared light in the vegetation area related to the remote sensing image. These values ​​reflect the reflectance characteristics of grassland vegetation under different wavelengths of electromagnetic waves and can be used to preliminarily distinguish vegetation from bare soil and non-vegetated background water bodies. The spectral characteristics of grassland vegetation represent the key growth states of grassland vegetation coverage, photosynthetic intensity, and biomass accumulation potential in the target grassland area and can be used for feature analysis of vegetation in the target grassland area.

[0038] In some embodiments, reference Figure 3 As shown, this figure is an exemplary flowchart for determining a multi-source feature fusion map in some embodiments of this application. In this embodiment, the grassland vegetation spectral features and the texture-enhanced vegetation features are interactively fused to obtain a multi-source feature fusion map of grassland organisms. This can be achieved by the following steps: In step 1031, the grassland vegetation spectral features and the texture-enhanced vegetation features are spatially registered to obtain spatially registered features; In step 1032, the spatial registration features are normalized to obtain normalized spatial registration features; In step 1033, the normalized spatial registration features are fused to obtain a multi-source feature fusion map of grassland organisms.

[0039] In specific implementation, the spectral features of the grassland vegetation and the texture-enhanced vegetation features are spatially registered to obtain the spatial registration features. This can be achieved in the following way: First, using the geographic coordinate system of the original remote sensing image as a reference, a quadratic polynomial geometric correction algorithm is used to spatially register the values ​​of each dimension of the grassland vegetation spectral features with the values ​​of each dimension of the texture-enhanced vegetation features. By selecting no less than 20 evenly distributed ground control points, spatial transformation parameters are calculated, and the spectral feature values ​​and texture feature values ​​corresponding to the same spatial location are accurately associated and matched to ensure that the two types of features correspond one-to-one in spatial dimensions, thus obtaining the spatial registration features. Other methods can also be used in other embodiments, which are not limited here. In addition, in specific implementation, the spatial registration features are normalized to obtain normalized spatial registration features. This can be achieved by the following method: each spectral feature value and each texture feature value after spatial registration is processed using a linear normalization algorithm. During processing, the global maximum and global minimum values ​​of each spectral feature dimension and each texture feature dimension are first calculated. The global minimum value of the corresponding dimension is subtracted from the single feature value. Then, the calculation result is divided by the difference between the global maximum and global minimum values ​​of the corresponding dimension. In this way, all spectral feature values ​​and texture feature values ​​are uniformly mapped to the interval between 0 and 1, eliminating the dimensional differences between different feature dimensions, and obtaining normalized spatial registration features. Other methods can also be used in other embodiments, which are not limited here.

[0040] Furthermore, in specific implementation, the fusion processing of the normalized spatial registration features to obtain the multi-source feature fusion map of grassland aboveground organisms can be achieved in the following way: Based on historical measured grassland aboveground biomass data of the target grassland area, Pearson correlation analysis is performed with each normalized spectral feature dimension and each normalized texture feature dimension. The correlation coefficient between each feature dimension and historical measured biomass is calculated. Then, weight gradients are assigned according to the absolute value of the correlation coefficient; the larger the absolute value of the correlation coefficient, the higher the weight coefficient is assigned. At the same time, multiple weight combinations are set for cross-validation, and the fusion features are selected and... The weight combination with the highest correlation coefficient of historically measured biomass and the smallest measurement deviation is selected as the optimal weight. Finally, the fusion process is completed by performing a weighted summation operation on a spatial location basis. That is, each normalized spectral feature value is multiplied by the corresponding weight coefficient, each normalized texture feature value is multiplied by the corresponding weight coefficient, and all product results are added to obtain the fusion feature value. Then, all fusion feature values ​​are normalized twice to the range of 0 to 255, and all fusion feature values ​​are integrated in spatial location order to obtain a multi-source feature fusion map of grassland organisms. Other methods can be used in other embodiments, which are not limited here.

[0041] It should be noted that the spatial registration features in this application represent the feature set obtained by associating and matching the values ​​of each dimension of grassland vegetation spectral features and the values ​​of each dimension of texture-enhanced vegetation features with the geographic coordinates of the original remote sensing image as a reference, reflecting the correspondence consistency of the two types of features in spatial dimensions; the normalized spatial registration features represent the feature set obtained by linearly normalizing the spectral and texture feature values ​​in the spatial registration features, reflecting the standardized feature information after eliminating the differences in the dimensions of different features, solving the problem that spectral and texture features are difficult to directly fuse due to different numerical ranges; the multi-source feature fusion map represents the feature map of the fusion of spectral and texture features in the remote sensing image, reflecting the comprehensive information of combining the spectral growth status of grassland vegetation and the details of canopy texture, providing a core basis for the measurement of grassland aboveground biomass.

[0042] In step 104, historical remote sensing images of the target grassland area are acquired, and the historical grassland vegetation spectral features in the historical remote sensing images and the grassland vegetation spectral features are subjected to consistency constraints to obtain calibration constraints on the aboveground biomass of the grassland in the target grassland area.

[0043] It should be noted that the historical remote sensing images in this application represent a set of remote sensing observation data of the target grassland area acquired by satellite or airborne remote sensing sensors at different time points in the past and subjected to standardized preprocessing such as radiometric calibration, atmospheric correction, and geometric correction. They reflect the spectral reflectance characteristics, canopy texture and morphology, spatial distribution pattern of grassland vegetation in different historical periods of the region, as well as the comprehensive influence of environmental factors such as topography and climate on vegetation growth.

[0044] In some embodiments, the calibration constraint for grassland aboveground biomass in the target grassland area by applying consistency constraints to the historical grassland vegetation spectral features in the historical remote sensing image and the grassland vegetation spectral features can be achieved by the following steps: Extract the spectral features of historical grassland vegetation from the historical remote sensing images; Determine the deviation values ​​of the historical grassland vegetation spectral characteristics and the grassland vegetation spectral characteristics; The deviation data are fitted and analyzed to construct calibration constraints for grassland biomass in the target grassland area.

[0045] In specific implementation, the historical grassland vegetation spectral features in the historical remote sensing image are extracted. That is, the historical grassland vegetation spectral features are extracted using the current grassland vegetation spectral feature extraction method, ensuring that each feature dimension and calculation standard is completely consistent with the currently acquired grassland vegetation spectral features. Other methods can also be used in other embodiments, which are not limited here.

[0046] In addition, in specific implementation, the deviation values ​​of the historical grassland vegetation spectral features and the grassland vegetation spectral features can be determined in the following way: using the geographic coordinates of the remote sensing image as a reference, spatial registration is performed on the historical and current grassland vegetation spectral features. By matching ground control points again, it is ensured that the two types of feature values ​​at the same spatial location correspond one-to-one. Then, the deviation values ​​of the two types of features in each vegetation index dimension are obtained by using a dimension-by-dimensional and pixel-by-pixel difference calculation method. At the same time, the deviation values ​​are screened for anomalies using the 3σ principle. First, the mean and standard deviation of the deviation values ​​in each dimension are calculated. Deviation values ​​that exceed the range of mean plus or minus 3 times the standard deviation are judged as anomalies and removed. Valid deviation values ​​are retained and integrated to form complete deviation value data. Other methods can also be used in other embodiments, which are not limited here.

[0047] Furthermore, in specific implementation, the calibration constraints for grassland aboveground biomass in the target grassland area can be constructed by fitting and analyzing the deviation data in the following manner: First, collect measured data of grassland aboveground biomass in the target grassland area for historical periods. Combine this with the historical grassland vegetation spectral characteristics to calculate the estimated biomass value for the same period. Calculate the difference between the measured value and the estimated value and use it as the dependent variable. Simultaneously, use the effective deviation data filtered by the 3σ principle as the independent variable. Use a linear regression algorithm to fit and analyze the independent and dependent variables, calculate the quantitative correlation coefficient between the deviation values ​​of each vegetation index dimension and the biomass measurement deviation, and then verify the reliability of the correlation model through a goodness-of-fit test. The historical effective deviation values ​​of each vegetation index dimension are statistically calculated to obtain the mean and standard deviation. Then, the 3σ principle is used to determine the deviation threshold range, that is, the lower limit of the threshold is the mean minus 3 times the standard deviation, and the upper limit of the threshold is the mean plus 3 times the standard deviation. Deviation values ​​exceeding this range are judged as outliers. Finally, based on the obtained quantitative correlation coefficient and deviation threshold range, a specific correction formula is formulated as "Current biomass measurement correction amount = deviation value of each dimension × sum of corresponding correlation coefficients", as well as constraint rules that require re-verification of spectral feature data when the deviation threshold is exceeded. This completes the construction of grassland biomass calibration constraints in the target grassland area. Other methods can be used in other embodiments, which are not limited here.

[0048] It should be noted that the deviation data in this application represents the degree of deviation between the historical grassland vegetation spectral characteristics and the current grassland vegetation spectral characteristics, reflecting the degree of dimensional numerical deviation between the two types of spectral characteristics due to differences in imaging time, sensor parameters, environmental conditions, etc.; the calibration constraint represents the constraint conditions for calibrating the grassland aboveground biomass in the target grassland area, reflecting the quantitative correspondence between the spectral characteristic deviation and the grassland aboveground biomass measurement deviation, and can be used to correct the current grassland aboveground biomass measurement results.

[0049] In step 105, the biomass on the grassland in the target grassland area is calculated collaboratively based on the multi-source feature fusion map and the calibration constraints to obtain a biomass calculation map of the grassland in the target grassland area.

[0050] In some embodiments, the biomass of grassland in the target grassland area is jointly measured based on the multi-source feature fusion map and the calibration constraints to obtain a biomass measurement map of grassland in the target grassland area, which can be achieved by the following steps: Based on the multi-source feature fusion map, determine the basic budget value data of grassland aboveground biomass in the target grassland area; The biomass calibration amount for each pixel is determined based on the calibration constraints. The basic budget value data is collaboratively corrected by various biomass calibration values ​​to obtain the biomass correction calculation value for each pixel. Based on the biomass correction calculation value of each pixel, a biomass calculation map of the grassland in the target grassland area is obtained.

[0051] In specific implementation, the basic budget value of grassland biomass in the target grassland area can be determined based on the multi-source feature fusion map in the following way: First, extract the pixel-by-pixel fusion feature value of the target grassland area in the multi-source feature fusion map. This fusion value has been normalized to the range of 0 to 255 and contains spectral growth information and texture detail information. Substitute it into a pre-constructed linear regression calculation model. This model uses the historical measured grassland biomass data of the target grassland area in the past five years as the dependent variable and the corresponding historical multi-source feature fusion value as the independent variable. The training set and validation set are divided in a 7:3 ratio to carry out model training. After passing the goodness-of-fit test and the coefficient of determination is not less than 0.7, the basic budget value of grassland biomass in the target grassland area is obtained pixel-by-pixel through model calculation.

[0052] In addition, in specific implementation, the biomass calibration amount of each pixel can be determined according to the calibration constraints in the following way: for each pixel in the target grassland area, extract the feature values ​​of the current grassland vegetation spectral features and the historical grassland vegetation spectral features in each dimension, calculate the deviation value of the two types of feature values ​​in each dimension, and then substitute the deviation value of each dimension into the quantization correlation coefficient in the calibration constraints. The single-dimensional correction amount is weighted and summed according to the correlation weight of each dimension to obtain the biomass calibration amount of each pixel. At the same time, referring to the deviation threshold range determined based on the 3σ principle in the calibration constraints, pixels with deviation values ​​exceeding the threshold in any dimension are marked as abnormal pixels.

[0053] In addition, in specific implementation, the biomass correction calculation value of each pixel can be obtained by coordinating the correction of the basic budget value data through various biomass calibration values. This can be achieved in the following way: the preliminary biomass correction calculation value of each pixel is obtained by adding the basic budget value to the corresponding biomass calibration value. For marked abnormal pixels, the average correction calculation value of normal pixels in the 3×3 neighborhood is used for interpolation. If the proportion of abnormal pixels in the 3×3 neighborhood exceeds 50%, the interpolation is expanded to the 5×5 neighborhood to ensure that all pixels have effective biomass correction calculation values.

[0054] In addition, in specific implementation, the biomass measurement map of the grassland in the target grassland area can be obtained based on the biomass correction measurement value of each pixel in the target grassland area in the following way: using the geographic coordinate system, pixel resolution and spatial range of the original remote sensing image as a unified benchmark, the biomass correction measurement value of the grassland corresponding to each pixel in the target grassland area is accurately matched and assigned to the corresponding spatial location; then, the integrity verification of the measurement values ​​of all pixels is carried out, and for missing pixels without valid measurement values, the mean of the correction measurement values ​​of normal pixels in the 3×3 neighborhood is used for interpolation to supplement them; then, key metadata information such as imaging time, measurement model parameters, calibration constraint coefficients and biomass data units are added to the data; finally, the integrated pixel measurement values ​​are encapsulated into a standardized raster data format to obtain the biomass measurement map of the grassland in the target grassland area.

[0055] It should be noted that the basic budget value data in this application represents the set of preliminary calculated values ​​of grassland biomass in the target grassland area, reflecting the basic biomass calculation level that combines vegetation spectral growth characteristics and canopy texture details; the biomass calibration value reflects the correction range of biomass calculation deviation caused by factors such as spatiotemporal differences and different sensor parameters; the biomass correction calculation value reflects the grassland aboveground biomass level after deviation correction; the biomass calculation map represents the biomass calculation results map of the target grassland area, reflecting the spatial distribution pattern and difference characteristics of grassland aboveground biomass in the target grassland area.

[0056] In another aspect, in some embodiments, this application provides a device for measuring grassland biomass based on UAV remote sensing images, with reference to... Figure 4 The figure is a schematic diagram of a grassland biomass measurement device based on UAV remote sensing images, according to some embodiments of this application. The grassland biomass measurement device 400 based on UAV remote sensing images includes: an acquisition module 401, a processing module 402, and an execution module 403, which are described below: Acquisition module 401, in this application, is mainly used to acquire remote sensing images of the target grassland area via a drone; Processing module 402, in this application, is used to identify the relevant vegetation areas of grassland organisms from the remote sensing image based on the historical vegetation information of the target grassland area, and to perform local texture enhancement on the relevant vegetation areas in the remote sensing image, thereby obtaining the texture enhancement features of grassland organisms in the remote sensing image. It should be noted that the processing module 402 in this application is also used to identify grassland vegetation spectral features from the remote sensing image, and to interactively fuse the grassland vegetation spectral features and the texture-enhanced vegetation features to obtain a multi-source feature fusion map of grassland organisms. Additionally, it should be noted that the processing module 402 in this application is also used to acquire historical remote sensing images of the target grassland area, and to apply consistency constraints to the historical grassland vegetation spectral features in the historical remote sensing images and the grassland vegetation spectral features to obtain calibration constraints on the aboveground biomass of the grassland in the target grassland area. The execution module 403 in this application is mainly used to perform collaborative calculation of biomass on grassland in the target grassland area based on the multi-source feature fusion map and the calibration constraints, so as to obtain a biomass calculation map of grassland in the target grassland area.

[0057] In addition, this application also provides a computer device, which includes a memory and a processor. The memory stores code, and the processor is configured to acquire the code and execute the above-described method for measuring grassland biomass based on UAV remote sensing images.

[0058] In some embodiments, reference Figure 5 The figure is a schematic diagram of the structure of a computer device for implementing a method for measuring grassland biomass based on UAV remote sensing images, according to some embodiments of this application. The grassland biomass measurement method based on UAV remote sensing images in the above embodiments can be implemented through... Figure 5 The computer device shown is used to implement this, and the computer device 500 includes at least one processor 501, a communication bus 502, a memory 503, and at least one communication interface 504.

[0059] Processor 501 can be a general-purpose central processing unit (CPU) or an application-specific integrated circuit (ASIC).

[0060] The communication bus 502 can be used to transmit information between the aforementioned components.

[0061] Memory 503 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CDROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital versatile optical discs, Blu-ray discs, etc.), magnetic disks or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. Memory 503 may exist independently and be connected to processor 501 via communication bus 502. Memory 503 may also be integrated with processor 501.

[0062] The memory 503 stores program code for executing the scheme of this application, and its execution is controlled by the processor 501. The processor 501 executes the program code stored in the memory 503. The program code may include one or more software modules. The method used in the above embodiments can be implemented by the processor 501 and one or more software modules in the program code in the memory 503.

[0063] Communication interface 504 uses any transceiver-like device to communicate with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area networks (WLAN), etc.

[0064] In a specific implementation, as one example, a computer device may include multiple processors, each of which may be a single-core (single CPU) processor or a multi-core (multi CPU) processor. Here, a processor may refer to one or more devices, circuits, and / or processing cores used to process data (e.g., computer program instructions).

[0065] The aforementioned computer device can be a general-purpose computer device or a special-purpose computer device. In specific implementations, the computer device can be a desktop computer, a portable computer, a network server, a handheld digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device, or an embedded device. This application does not limit the type of computer device.

[0066] In addition, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for measuring grassland biomass based on UAV remote sensing images.

[0067] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0068] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for measuring grassland aboveground biomass based on unmanned aerial vehicle remote sensing images, characterized in that, Includes the following steps: Remote sensing images of the target grassland area were collected using drones; Based on historical vegetation information of the target grassland area, the relevant vegetation areas of grassland organisms are identified from the remote sensing image, and local texture enhancement is performed on the relevant vegetation areas in the remote sensing image to obtain the texture enhancement features of grassland organisms in the remote sensing image. Grassland vegetation spectral features are identified from the remote sensing image, and the grassland vegetation spectral features and the texture-enhanced vegetation features are interactively fused to obtain a multi-source feature fusion map of grassland organisms. Historical remote sensing images of the target grassland area are acquired, and the historical grassland vegetation spectral features in the historical remote sensing images and the grassland vegetation spectral features are subjected to consistency constraints to obtain calibration constraints on the aboveground biomass of the grassland in the target grassland area. Based on the multi-source feature fusion map and the calibration constraints, the biomass on the grassland in the target grassland area is calculated collaboratively to obtain a biomass calculation map of the grassland in the target grassland area.

2. The method of claim 1, wherein, The relevant vegetation areas for grassland organisms identified from the remote sensing images based on historical vegetation information of the target grassland area specifically include: Obtain historical vegetation information for the target grassland area; The remote sensing image is preprocessed to obtain a preprocessed remote sensing image; Using the historical vegetation information as a vegetation reference, the relevant vegetation areas of grassland organisms are identified from the preprocessed remote sensing images.

3. The method of claim 1, wherein, Local texture enhancement of relevant vegetation areas in the remote sensing image, thereby obtaining texture enhancement features of organisms on grassland in the remote sensing image, specifically includes: Convert the relevant vegetation areas in the remote sensing image into grayscale images; The grayscale image is locally enhanced by sliding to obtain the enhanced image of the relevant vegetation area; The enhanced image is fused with the remote sensing image to obtain the texture enhancement features of organisms on the grassland in the remote sensing image.

4. The method as described in claim 1, characterized in that, Identifying the spectral features of grassland vegetation from the remote sensing image specifically includes: Extract the regional reflectance data of the relevant vegetation areas in the remote sensing image; The spectral characteristics of grassland vegetation are determined based on the reflectance data of the region.

5. The method as described in claim 1, characterized in that, The interaction and fusion of the grassland vegetation spectral features and the texture-enhanced vegetation features to obtain a multi-source feature fusion map of grassland organisms specifically includes: Spatial registration is performed between the grassland vegetation spectral features and the texture-enhanced vegetation features to obtain spatial registration features; The spatial registration features are normalized to obtain normalized spatial registration features; The normalized spatial registration features are fused to obtain a multi-source feature fusion map of grassland organisms.

6. The method as described in claim 1, characterized in that, The calibration constraints for grassland aboveground biomass in the target grassland area are obtained by applying consistency constraints between the historical grassland vegetation spectral features in the historical remote sensing images and the grassland vegetation spectral features. These constraints specifically include: Extract the spectral features of historical grassland vegetation from the historical remote sensing images; Determine the deviation values ​​of the historical grassland vegetation spectral characteristics and the grassland vegetation spectral characteristics; The deviation data are fitted and analyzed to construct calibration constraints for grassland biomass in the target grassland area.

7. The method as described in claim 1, characterized in that, Based on the multi-source feature fusion map and the calibration constraints, the biomass of grassland in the target grassland area is collaboratively measured to obtain a biomass measurement map of grassland in the target grassland area, specifically including: Based on the multi-source feature fusion map, determine the basic budget value data of grassland aboveground biomass in the target grassland area; The biomass calibration amount for each pixel is determined based on the calibration constraints. The basic budget value data is collaboratively corrected by various biomass calibration values ​​to obtain the biomass correction calculation value for each pixel. Based on the biomass correction calculation value of each pixel, a biomass calculation map of the grassland in the target grassland area is obtained.

8. A device for measuring grassland biomass based on UAV remote sensing images, characterized in that, include: The acquisition module is used to acquire remote sensing images of the target grassland area via drone; The processing module is used to identify the relevant vegetation areas of grassland organisms in the remote sensing image based on the historical vegetation information of the target grassland area, and to perform local texture enhancement on the relevant vegetation areas in the remote sensing image, thereby obtaining the texture enhancement features of grassland organisms in the remote sensing image. The processing module is also used to identify grassland vegetation spectral features from the remote sensing image, and to interactively fuse the grassland vegetation spectral features and the texture-enhanced vegetation features to obtain a multi-source feature fusion map of grassland organisms. The processing module is also used to acquire historical remote sensing images of the target grassland area, and to apply consistency constraints to the historical grassland vegetation spectral features in the historical remote sensing images and the grassland vegetation spectral features to obtain calibration constraints on the aboveground biomass of the grassland in the target grassland area. The execution module is used to collaboratively calculate the biomass of grassland in the target grassland area based on the multi-source feature fusion map and the calibration constraints, so as to obtain a biomass calculation map of grassland in the target grassland area.

9. A computer device, characterized in that, The computer device includes a memory and a processor, the memory storing code, and the processor being configured to acquire the code and execute the grassland biomass measurement method based on UAV remote sensing images as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the grassland biomass measurement method based on UAV remote sensing images as described in any one of claims 1 to 7.

Images