Sensitized artemisia multi-element remote sensing identification method based on FSDAF spatio-temporal fusion

Abstract

This invention discloses a multi-source remote sensing identification method for allergenic wormwood based on FSDAF spatiotemporal fusion. It integrates high-resolution remote sensing satellite imagery (optical and radar images) combined with UAV data and multi-source data from ground meteorology and vegetation monitoring. Utilizing advanced technologies such as FSDAF spatiotemporal fusion, it constructs an identification model for allergenic wormwood based on the spectral, texture, density, and geometric characteristics of the plant's growth period. This model identifies severely invaded areas of allergenic wormwood, achieving high-precision automatic identification and spatial distribution mapping of allergenic wormwood over large areas. Based on the identification model and growth status analysis model, it accurately analyzes the distribution area and severity of damage of allergenic wormwood within different administrative divisions, providing a basis for implementing targeted wormwood control measures.

Description

Technical Field

[0001] This invention relates to the field of multi-source remote sensing identification technology, specifically to a multi-source remote sensing identification method for sensitized artemisia based on FSDAF spatiotemporal fusion. Background Technology

[0002] Artemisia species are naturally distributed in the core of the North Temperate Zone, where the dry climate facilitates pollen dispersal. As pollen spreads, allergy problems become increasingly prominent. Artemisia allergies in my country exhibit a distinct pattern of high incidence in the north and strong seasonality, with allergy-affected areas centered in North China, Northeast China, and Northwest China, and locally impacting parts of East China, Central China, and Southwest China. Allergenic Artemisia species (such as Artemisia capillaris, Artemisia argyi, and Artemisia annua) are the primary source of airborne allergenic pollen in northern my country during the summer and autumn seasons (July-September), and their pollen dispersal poses a serious threat to public health.

[0003] Accurate identification of allergenic artemisia is crucial for allergy control and forms the basis for remote sensing monitoring of allergenic artemisia plant growth and control of allergenic pollen. Traditional ground survey methods are time-consuming, labor-intensive, and have limited coverage, making it difficult to achieve large-scale, dynamic, and accurate monitoring. Currently, the rapid development of computer vision and deep learning technologies, which enable automatic and rapid classification through the analysis of plant images, has become a core focus of research in various industries such as agriculture, forestry, and animal husbandry. Summary of the Invention

[0004] This application provides a multivariate remote sensing identification method for sensitized artemisia based on FSDAF spatiotemporal fusion, in order to solve the technical problems of existing ground survey methods, such as being time-consuming and labor-intensive, having limited coverage, and being unable to achieve large-scale, dynamic, and accurate monitoring.

[0005] According to the first aspect, one embodiment provides a multivariate remote sensing identification method for sensitized artemisia based on FSDAF spatiotemporal fusion, the method comprising: By detecting Pettitt mutation points in the surface meteorological monitoring data of the target area, the critical time point of the vigorous growth period of sensitized Artemisia in the target area was determined. High-resolution remote sensing images, UAV aerial images, and radar images of the target area during the vigorous growth period of sensitized wormwood after the critical time point are collected. The collected multi-dimensional image data are fused using the FSDAF spatiotemporal fusion model to generate a large-area, long-term sequence image dataset of the growth and development characteristics of sensitized wormwood. Based on the obtained image dataset of sensitized Artemisia growth and development characteristics, feature extraction was performed from four dimensions: spectrum, texture, density, and geometry, resulting in a sensitized Artemisia feature dataset that includes normalized vegetation index, enhanced vegetation index, pixel shape index, texture feature index, and spectral feature index. Based on the aforementioned sensitized artemisia feature dataset, the trained CART classification tree algorithm is used for classification and identification processing, outputting the identification results of sensitized artemisia types and distributions.

[0006] Furthermore, by performing Pettitt mutation point detection on surface meteorological monitoring data of the target area, the critical time point of the vigorous growth period of sensitized artemisia in the target area was determined, specifically including: Based on the surface meteorological monitoring data series of the target area, calculate the Mann-Whitney nonparametric statistic: ; In the formula, t = 2, 3, 4, ..., n; is a statistical sequence; sgn is the sign function; X t X is the t-th data point in the statistical sequence; i is the i-th data point in the statistical sequence; n is the number of data points in the statistical sequence. Calculate the mutation point and significance level p-value based on the Mann-Whitney nonparametric statistic: ; ; In the formula, K t,N If p≤0.05, the mutation point is determined to be valid, and the corresponding mutation point is the critical time point of the vigorous growth period of the sensitized Artemisia in the target area.

[0007] Furthermore, the normalized vegetation index is calculated as follows: ; In the formula, NDVI is the normalized vegetation index; It is in the near-infrared band; It is in the red light band.

[0008] Furthermore, the enhanced vegetation index is calculated as follows: ; In the formula, denoted as , where is the surface reflectance in the near-infrared, red, and blue bands; G is the gain factor; C1 and C2 are atmospheric impedance coefficients; and L is the soil adjustment parameter.

[0009] Furthermore, the pixel shape index is calculated as follows: The pixel shape index uses basic shape features as classification features for allergenic wormwoods. The basic shape features of allergenic wormwoods include the perimeter, area, compactness, and shape coefficient of similar pixel groups formed by allergenic wormwood communities. Perimeter: The perimeter of a planar object is calculated using its boundary pixel information. Eight-directional chain codes are used to represent the boundary, and the step sizes of all boundary chain codes are summed. The calculation formula is as follows: ; In the formula, L i P is the distance between adjacent pixels, and P is the perimeter of the surface object; ; Where n is the result of taking the remainder of chain code direction i with 2; Area: The area S of a planar object is the sum of the number of pixels within the object area multiplied by the ground area corresponding to each pixel. Compactness: This is a morphological indicator used to quantify the compactness of a two-dimensional shape. It measures how close the shape is to a perfect circle, and is calculated as follows: ; In the formula, S is the area of ​​the planar object, and P is the perimeter of the planar object; Shape factor: An indicator used to quantify the compactness or near-circularity of a shape; the calculation formula is as follows: ; In the formula, F is the shape coefficient, S is the area of ​​the planar object, and P is the perimeter of the planar object.

[0010] Furthermore, the texture feature index is calculated as follows: Texture feature indices include contrast, correlation, energy, homogeneity, and second-order entropy; Contrast ratio: Reflects the brightness difference between a given pixel and its neighboring pixels. The formula is as follows: ; In the formula f CON For contrast, x and y are the pixels with grayscale values ​​of x and y, respectively, and N is the total number of pixels. Let y be the probability of a pixel with gray level y appearing at a distance d from a pixel with gray level x and in the direction θ. Correlation: Represents the degree of similarity between elements in the spatial gray-level co-occurrence matrix in the row or column direction. Therefore, the correlation value reflects the gray-level correlation of the layout in the image. The formula is as follows: ; In the formula f COR For correlation, x and y are the pixels with gray levels x and y, respectively, and N is the total number of pixels. Let x be the mean of the grayscale values. Let be the probability of a pixel with gray level y appearing at a distance d from a pixel with gray level x and in the direction θ. Let x be the variance of the gray level. Let be the variance of the grayscale value y; Energy: This is a measure of the amount of information contained in an image, representing the degree of non-uniformity or complexity of texture in the image. The formula is as follows: ; In the formula f ENE Let N be the energy, x and y be the pixels with grayscale values ​​of x and y, respectively, and N be the total number of pixels. Let y be the probability of a pixel with gray level y appearing at a distance d from a pixel with gray level x and in the direction θ. Homogeneity: This is a measure describing the uniformity of pixels in an image, indicating whether the distribution of pixels in the image is uniform. The formula is as follows: ; In the formula f HOM To represent homogeneity, x and y are the pixels with gray levels x and y, respectively, and N is the total number of pixels. Let y be the probability of a pixel with gray level y appearing at a distance d from a pixel with gray level x and in the direction θ. Second-order entropy: Characterizes the spatial features of grayscale information, reflecting the comprehensive characteristics of the grayscale value at a certain pixel location and the grayscale distribution of surrounding pixels. The formula is as follows: ; In the formula H ENT Let N be the second-order entropy, x and y be the pixels with gray levels x and y, respectively, and N be the total number of pixels. Let be the probability of a pixel with gray level y appearing at a distance d from a pixel with gray level x and in the direction θ. for The constant logarithm.

[0011] Furthermore, the spectral characteristic index is calculated as follows: Spectral characteristics include spectral brightness and mean values ​​for each band; The band mean is calculated as follows: ; In the formula, The average grayscale value is the number of rows in the image, where N is the number of rows in the image. This represents the row and column position of a unit pixel in the image. The grayscale value of a pixel in the image; The spectral brightness is calculated using a weighted average value, and the calculation formula is as follows: ; In the formula, Pi is the spectral brightness in the row direction of the image, which is the gray-weighted average value in the row direction of the image. Pi is the weight of the pixels in the image in the row direction.

[0012] Furthermore, the method also includes: Based on the ground survey data of allergenic artemisia, the identification results are interactively interpreted and optimized, and the distribution and area of ​​allergenic artemisia types are statistically analyzed based on the optimization results.

[0013] According to the second aspect, one embodiment provides a multivariate remote sensing identification system for sensitized artemisia based on FSDAF spatiotemporal fusion, the system comprising: The Pettitt mutation detection module is used to detect Pettitt mutation points in the ground meteorological monitoring data of the target area and determine the critical time point of the vigorous growth period of sensitized wormwood in the target area. The FSDAF spatiotemporal fusion module is used to collect high-resolution remote sensing images, UAV aerial images, and radar images of the target area during the vigorous growth period of sensitized wormwood after the critical time point. The FSDAF spatiotemporal fusion model is used to fuse the collected multi-dimensional image data to generate a large-area, long-term sequence image dataset of the growth and development characteristics of sensitized wormwood. The feature extraction module is used to extract features from four dimensions—spectrum, texture, density, and geometry—based on the obtained image dataset of sensitized wormwood growth and development. This results in a sensitized wormwood feature dataset that includes normalized vegetation index, enhanced vegetation index, pixel shape index, texture feature index, and spectral feature index. The identification module is used to perform classification and identification processing based on the sensitized artemisia feature dataset using a trained CART classification tree algorithm, and output the identification results of the sensitized artemisia type and distribution.

[0014] According to a third aspect, one embodiment provides an electronic device, the device comprising: a processor and a memory; The memory is used to store one or more program instructions; The processor is configured to run one or more program instructions to perform the steps of a multivariate remote sensing identification method for sensitized artemisia based on FSDAF spatiotemporal fusion as described in any of the preceding claims.

[0015] According to a fourth aspect, one embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of a multivariate remote sensing identification method for sensitized artemisia based on FSDAF spatiotemporal fusion as described in any of the preceding claims.

[0016] This application provides a multi-source remote sensing identification method for allergenic wormwood based on FSDAF spatiotemporal fusion, which has the following beneficial effects: This invention integrates multi-source high-resolution imagery (optical imagery, radar imagery) from high-resolution remote sensing satellites and UAVs, as well as multi-source data from ground meteorology and vegetation monitoring. Utilizing advanced technologies such as FSDAF spatiotemporal fusion, it constructs an identification model for allergenic wormwood based on the spectral, texture, density, and geometric characteristics of the wormwood's growth period. This model identifies severely invaded areas of allergenic wormwood, achieving high-precision automatic identification and spatial distribution mapping of allergenic wormwood over large areas. Based on the allergenic wormwood identification model and growth status analysis model, it accurately statistically analyzes the distribution area and degree of damage of allergenic wormwood within different administrative divisions, providing a basis for taking targeted wormwood control measures. Attached Figure Description

[0017] Figure 1 A flowchart of a multivariate remote sensing identification method for sensitized artemisia based on FSDAF spatiotemporal fusion is provided as an embodiment of the present invention; Figure 2 This is a schematic diagram of the logical structure of a multi-dimensional remote sensing identification system for sensitized artemisia based on FSDAF spatiotemporal fusion, provided as an embodiment of the present invention. Detailed Implementation

[0018] The present invention will now be described in further detail with reference to specific embodiments and accompanying drawings. Similar elements in different embodiments are referred to by associated similar element reference numerals. In the following embodiments, many details are described to facilitate a better understanding of this application. However, those skilled in the art will readily recognize that some features may be omitted in different situations, or may be replaced by other elements, materials, or methods. In some cases, certain operations related to this application are not shown or described in the specification. This is to avoid obscuring the core parts of this application with excessive description. For those skilled in the art, detailed description of these related operations is not necessary; they can fully understand the related operations based on the description in the specification and general technical knowledge in the art.

[0019] Furthermore, the features, operations, or characteristics described in the specification can be combined in any suitable manner to form various embodiments. At the same time, the steps or actions in the method description can be rearranged or adjusted in a manner obvious to those skilled in the art. Therefore, the various orders in the specification and drawings are only for the clear description of a particular embodiment and do not imply a necessary order, unless otherwise stated that a particular order must be followed.

[0020] The first embodiment of this invention provides a multivariate remote sensing identification method for sensitized artemisia based on FSDAF spatiotemporal fusion, which will be discussed below. Figure 1Please provide a detailed explanation.

[0021] like Figure 1 As shown, in step S100, the critical time point of the vigorous growth period of sensitized wormwood in the target area is determined by detecting Pettitt mutation points in the ground meteorological monitoring data of the target area.

[0022] Sensitive wormwoods, such as *Artemisia annua*, *Artemisia serrata*, and *Artemisia annua*, are highly sensitive to water conditions, especially during the summer when rainfall and heat coincide, resulting in faster growth than other plants. This is the critical time point for distinguishing sensitive wormwoods from other plants. Therefore, determining the critical point of the vigorous growth period of sensitive wormwoods is fundamental to reducing the redundancy of remote sensing data on their growth period. This critical time point also serves as the starting point for distinguishing sensitive wormwoods from other plants. The Pettitt test is a robust and classic method for detecting catastrophes in time series, specifically designed to detect the location of individual catastrophes in a sequence. It identifies catastrophes by comparing the consistency of the distribution of two sequences before and after the catastrophe. Therefore, the Pettitt test can be used to determine the starting point of the vigorous growth period of sensitive wormwoods in a region. The Pettitt test is a method based on the Mann-Whitney nonparametric test, commonly used in hydrological and meteorological element sequences to analyze catastrophes and obtain catastrophes, quantifying the statistical significance of the catastrophes. The Mann-Whitney nonparametric statistic is:

[0023] In the formula, t = 2, 3, 4, ..., n; is a statistical sequence; sgn is the sign function; X t X is the t-th data point in the statistical sequence; i is the i-th data point in the statistical sequence; n is the number of data points in the statistical sequence.

[0024]

[0025] In the formula, K t,N A mutation point exists when p ≤ 0.05.

[0026] like Figure 1 As shown, in step S200, high-resolution remote sensing images, UAV aerial images, and radar images of the target area during the vigorous growth period of sensitized wormwood after the critical time point are collected, and the collected multi-dimensional image data are fused using the FSDAF spatiotemporal fusion model to generate a large-area, long-term sequence image dataset of the growth and development characteristics of sensitized wormwood.

[0027] FSDAF stands for Flexible Spatiotemporal Data Fusion. FSDAF fusion combines high temporal resolution, low spatial resolution spatial raster data with high spatial resolution, low temporal resolution spatial raster data to create a composite image with both high spatiotemporal resolution. Its core advantage lies in effectively predicting changes in land surface conditions during phenological transitions and abrupt changes in land cover types with minimal input data. This invention, based on Petitt mutation analysis, identifies the critical growth period of sensitized artemisia in the region. The FSDAF spatiotemporal fusion model is then used to fuse subsequent high-resolution remote sensing images, UAV aerial images, and radar images from this period, generating a large-area, long-term sequence image dataset of the growth and development characteristics of sensitized artemisia. The FSDAF implementation path is as follows: The FSDAF spatiotemporal fusion model takes as input a high-resolution image Ra (referring to sub-meter level remote sensing imagery, including sub-meter level satellite imagery and UAV imagery) at time ta and two low-resolution images Rb and Rc (including meter-level imagery of long-term series, such as the Sentinel series and the Landsat series) at times ta and tb. An unsupervised classification method is used to classify the high-resolution image at time ta, obtaining the land cover category weight Pc per pixel. Based on the spectral linear decomposition theory, m pixels whose land surface type remains unchanged from time ta to tb are selected, and the linear fusion equation is solved using formula (4) to obtain the temporal variation ΔR(c) for each band.

[0028]

[0029] In the formula, m represents the total number of unsupervised classifications. For pixels in the image at time tc The proportion of land cover categories .

[0030] Calculate the prediction results of the time variation for each band. The calculation formula is shown in (5):

[0031] True pixel value and temporal predicted value The difference is represented by the residual E, which can be calculated using formula (6):

[0032] Predicted spatial variation value at time tb It can be calculated using thin plate spline functions, and the calculation formula (7) is as follows:

[0033] in when When the value of is the minimum, formula (7) is determined to be the optimal parameter.

[0034] The FSDAF algorithm calculates... Then the weight function is obtained. The calculation formula (8) is shown below: (8) in When the p-th cell within a moving window is of the same class, ;otherwise The value of HI ranges from 0 to 1; a larger value indicates a smaller difference in variation. By... Normalization Then, by calculating the residual value Using the ΔR(c) value, the change in the high-resolution image from time tb to time ta for category c is obtained. The calculation formula (9) is as follows:

[0035] Finally, similar pixels with close pixel values ​​are selected using the spatial distance weighting formula (10):

[0036] right After weighted summation, the result is added to the surface reflectance value of the high-resolution image at time ta, which is obtained by solving formula (11). , where c is the center pixel position of the moving window and w is the window size.

[0037]

[0038] in .

[0039] like Figure 1 As shown, in step S300, based on the obtained image dataset of the growth and development characteristics of sensitized wormwood, feature extraction processing is performed from four dimensions: spectrum, texture, density, and geometry, to obtain a feature dataset of sensitized wormwood that includes normalized vegetation index, enhanced vegetation index, pixel shape index, texture feature index, and spectral feature index.

[0040] The core of this multi-dimensional feature extraction model for the growth stages of Artemisia species lies in integrating remote sensing and field data to accurately characterize their growth dynamics through feature engineering. The model extracts features collaboratively from four dimensions: In the spectral dimension, it captures unique reflectance spectral characteristics of key phenological stages such as greening, flowering, and yellowing by analyzing temporal remote sensing imagery using NDVI and enhanced vegetation index; in the texture dimension, it quantifies the spatial heterogeneity of canopy structure using methods such as gray-level co-occurrence matrices; and in the density and geometric dimensions, it analyzes morphological parameters such as community density and geometry of allergenic Artemisia species by combining object-oriented segmentation and UAV imagery. Finally, through multi-source feature fusion and machine learning algorithms, a comprehensive model is constructed that can accurately identify Artemisia species, quantify growth status, and support allergen risk assessment, providing a basis for precise monitoring and control.

[0041] (a) Normalized Difference Vegetation Index The Normalized Difference Vegetation Index (NDVI) is a crucial parameter reflecting vegetation health, nutrient information, growth vigor, and greenness. It is used to detect vegetation growth status, vegetation cover, and eliminate image radiometric errors. The NDVI directly reflects vegetation conditions influenced by both natural and anthropogenic factors, eliminating most of the effects of topography, solar angle, cloud shadows, and atmospheric conditions. It is one of the most widely used vegetation indices. Its calculation formula is as follows:

[0042] In the formula, NDVI is the normalized vegetation index; It is in the near-infrared band; It is in the red light band.

[0043] (ii) Enhanced vegetation index The Enhanced Vegetation Index (EVI) is an optimized vegetation index designed to improve upon the shortcomings of the NDVI, such as its saturation in high-biomass areas and its sensitivity to soil background and atmospheric influences. Its core principle is to incorporate blue bands for atmospheric correction and use adjustment coefficients to reduce background noise. In models for extracting the growth characteristics of sensitized Artemisia species, EVI can be incorporated as a core spectral feature. The clearer, less noisy time-series curves constructed from EVI more accurately reflect the growth dynamics of Artemisia species during key phenological stages such as greening, flowering (which may be accompanied by spectral changes), and yellowing, thus improving the accuracy of species identification and phenological monitoring.

[0044]

[0045] In the formula, denoted as , where is the surface reflectance in the near-infrared, red, and blue bands; G is the gain factor, a constant of 2.5; C1 and C2 are atmospheric impedance coefficients, with C1 being a constant of 6 and C2 a constant of 7.5; L is the soil adjustment parameter, a constant of 1.0.

[0046] (iii) Pixel Shape Index Allergenic Artemisia species exhibit distinctive shape characteristics that differentiate them from other plant communities during their growth period. Object-based shape features include both contour-based and region-based features. Object-based features utilize the outer edges of the shape, while region-based features utilize the overall information of the target area. This invention's pixel shape index uses basic shape features as classification characteristics for allergenic Artemisia species. These basic shape features include the perimeter, area, compactness, and shape coefficient of similar pixel groups formed by allergenic Artemisia species communities.

[0047] Perimeter: The perimeter of a planar object can be calculated using its boundary pixel information. If an eight-direction chain code is used to represent the boundary, the perimeter is calculated by summing the step sizes of all boundary chain codes. The step size for each step depends on the chain code direction value i: a shorter step size is used when direction i is even (representing a horizontal or vertical direction), and a longer step size is used when direction i is odd (representing a diagonal direction). When calculating the total length, it is also necessary to multiply by the actual physical size represented by each pixel (i.e., spatial resolution) to obtain the true perimeter value.

[0048]

[0049] In the formula, L i P is the distance between adjacent pixels, and P is the perimeter of the surface object.

[0050]

[0051] Where n is the result of taking the remainder of chain code direction i with 2;

[0052] Area: The area S of a planar object is the sum of the number of pixels within the object area multiplied by the ground area corresponding to each pixel.

[0053] Compactness is a morphological index used to quantify the compactness of a two-dimensional shape. It measures how close a shape is to a perfect circle. The more rounded the shape, the higher its compactness value; the more elongated and irregular the shape, the lower its compactness value.

[0054]

[0055] In the formula, S is the area of ​​the planar object, and P is the perimeter of the planar object; Shape factor: A metric used to quantify the compactness or near-circularity of a shape, calculated as follows:

[0056] In the formula, F is the shape coefficient, S is the area of ​​the planar object, and P is the perimeter of the planar object.

[0057] (iv) Texture features The texture feature index is calculated using the gray-level co-occurrence matrix (GLCM). Based on the GLCM's information regarding the direction, interval, and rate of change of image gray levels, statistical attributes that quantitatively describe texture features are extracted. It represents the probability of a pixel with gray level j appearing at a distance d from a pixel with gray level i and a direction θ. The texture feature index is composed of contrast, correlation, energy, homogeneity, and second-order entropy (ENT). The specific calculation formula is as follows: Contrast ratio: Reflects the brightness contrast between a given pixel and its neighboring pixels. If elements off-diagonal have large values, indicating rapid changes in image brightness, the CON value will be large. Contrast ratio also reflects the sharpness of an image and the depth of its texture. Deeper textures result in greater contrast and a clearer visual effect; conversely, lower contrast indicates shallower textures and a blurrier effect.

[0058]

[0059] f CON For contrast, x and y are the pixels with grayscale values ​​of x and y, respectively, and N is the total number of pixels. Let y be the probability of a pixel with gray level y appearing at a distance d from a pixel with gray level x and in the direction θ.

[0060] Correlation: Represents the degree of similarity between elements in the spatial gray-level co-occurrence matrix along the row or column direction. Therefore, the magnitude of the correlation value reflects the gray-level correlation in the image layout. When the matrix element values ​​are uniformly equal, the correlation value is large; conversely, if the matrix pixel values ​​differ greatly, the correlation value is small.

[0061]

[0062] In the formula f COR For correlation, x and y are the pixels with gray levels x and y, respectively, and N is the total number of pixels. Let x be the mean of the grayscale values. Let be the probability of a pixel with gray level y appearing at a distance d from a pixel with gray level x and in the direction θ. Let x be the variance of the gray level. Let y be the variance of the gray level.

[0063] Energy: This is a measure of the amount of information contained in an image, and a measure of randomness, representing the degree of non-uniformity or complexity of texture in the image. When all elements in the co-occurrence matrix have maximum randomness, all values ​​in the spatial co-occurrence matrix are almost equal, and the elements in the co-occurrence matrix are dispersed, the entropy is relatively high.

[0064]

[0065] In the formula f ENE Let N be the energy, x and y be the pixels with grayscale values ​​of x and y, respectively, and N be the total number of pixels. Let y be the probability of a pixel with gray level y appearing at a distance d from a pixel with gray level x and in the direction θ.

[0066] Homogeneity is a measure that describes the uniformity of pixels in an image, indicating whether the distribution of pixels in the image is uniform. If the pixels are the same or nearly similar in gray level, the homogeneity is high; if the pixel gray level distribution is significantly uneven, the homogeneity is low.

[0067]

[0068] In the formula f HOM To represent homogeneity, x and y are the pixels with gray levels x and y, respectively, and N is the total number of pixels. Let y be the probability of a pixel with gray level y appearing at a distance d from a pixel with gray level x and in the direction θ.

[0069] Second-order entropy: Characterizes the spatial features of grayscale information, reflecting the comprehensive characteristics of the grayscale value at a certain pixel location and the grayscale distribution of surrounding pixels. Second-order entropy reflects the degree of disorder in an image. The larger the second-order entropy value, the more disordered and complex the image.

[0070]

[0071] In the formula H ENT Let N be the second-order entropy, x and y be the pixels with gray levels x and y, respectively, and N be the total number of pixels. Let be the probability of a pixel with gray level y appearing at a distance d from a pixel with gray level x and in the direction θ. for The constant logarithm.

[0072] (v) Spectral characteristics Spectral features of remote sensing images are crucial for extracting and identifying plant community information, and are also a key factor in selecting training samples of sensitized artemisia. Spectral features include spectral brightness and the mean value of each band. The specific calculation formula is as follows:

[0073] In the formula, The average grayscale value is the number of rows in the image, where N is the number of rows in the image. This represents the row and column position of a unit pixel in the image. This represents the grayscale value of a pixel in the image.

[0074] During the vigorous growth period of the sensitized Artemisia communities, the gray values ​​of adjacent pixels differ significantly. Therefore, the spectral brightness characteristic is calculated using a weighted average to determine the mean brightness value. The calculation formula is as follows:

[0075] In the formula, Pi is the spectral brightness in the row direction of the image, which is the gray-weighted average value in the row direction of the image. Pi is the weight of the pixels in the image in the row direction.

[0076] like Figure 1 As shown, in step S400, based on the sensitized artemisia feature dataset, the trained CART classification tree algorithm is used for classification and identification processing, and the identification results of the sensitized artemisia type and distribution are output.

[0077] Based on the training samples and the allergenic Artemisia feature dataset (Normalized Difference Vegetation Index, Enhanced Vegetation Index, Pixel Shape Index, Texture Features, and Spectral Features), the CART (Classification and Regression Tree) classification tree algorithm was used to obtain the allergenic Artemisia types and distribution data. The CART classification and regression tree can perform both classification and regression during the classification process; its output is the category of the classified sample, and the output of the regression tree is a real number.

[0078] The CART classification tree algorithm uses the Gini coefficient to select features. The Gini coefficient represents the impurity of the model. The smaller the Gini coefficient, the lower the impurity and the better the features.

[0079] The formula for calculating the Gini value is:

[0080]

[0081] In the formula, For classification The probability of occurrence, where n is the number of categories. This reflects the probability that two random samples in dataset D have inconsistent labels. The smaller the value, the higher the purity of the two samples.

[0082] Based on the ground survey data of allergenic artemisia, the identification and interpretation results were interactively optimized, and the distribution and area of ​​allergenic artemisia types were statistically analyzed based on the optimization results.

[0083] Corresponding to the aforementioned method for multivariate remote sensing identification of sensitized artemisia based on FSDAF spatiotemporal fusion, this invention also discloses a multivariate remote sensing identification system for sensitized artemisia based on FSDAF spatiotemporal fusion, such as... Figure 2 As shown, it specifically includes: The Pettitt mutation detection module is used to detect Pettitt mutation points in the ground meteorological monitoring data of the target area and determine the critical time point of the vigorous growth period of sensitized wormwood in the target area. The FSDAF spatiotemporal fusion module is used to collect high-resolution remote sensing images, UAV aerial images, and radar images of the target area during the vigorous growth period of sensitized wormwood after the critical time point. The FSDAF spatiotemporal fusion model is used to fuse the collected multi-dimensional image data to generate a large-area, long-term sequence image dataset of the growth and development characteristics of sensitized wormwood. The feature extraction module is used to extract features from four dimensions—spectrum, texture, density, and geometry—based on the obtained image dataset of sensitized wormwood growth and development. This results in a sensitized wormwood feature dataset that includes normalized vegetation index, enhanced vegetation index, pixel shape index, texture feature index, and spectral feature index. The identification module is used to perform classification and identification processing based on the sensitized artemisia feature dataset using a trained CART classification tree algorithm, and output the identification results of the sensitized artemisia type and distribution.

[0084] It should be noted that for a detailed description of the multivariate remote sensing identification system for sensitized artemisia based on FSDAF spatiotemporal fusion provided in the embodiments of the present invention, please refer to the relevant description of the multivariate remote sensing identification method for sensitized artemisia based on FSDAF spatiotemporal fusion provided in the embodiments of this application, which will not be repeated here.

[0085] In addition, embodiments of the present invention also provide an electronic device, the device comprising: a processor and a memory; the memory being used to store one or more program instructions; the processor being used to execute one or more program instructions to perform the steps of a multivariate remote sensing identification method for sensitized artemisia based on FSDAF spatiotemporal fusion as described in any of the preceding embodiments.

[0086] It should be noted that for a detailed description of an electronic device provided in the embodiments of the present invention, please refer to the relevant description of a multivariate remote sensing identification method for sensitized artemisia based on FSDAF spatiotemporal fusion provided in the embodiments of this application, which will not be repeated here.

[0087] In addition, embodiments of the present invention also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the multivariate remote sensing identification method for sensitized artemisia based on FSDAF spatiotemporal fusion as described in any of the preceding claims.

[0088] It should be noted that for a detailed description of the computer-readable storage medium provided in the embodiments of the present invention, please refer to the relevant description of the multivariate remote sensing identification method for sensitized artemisia based on FSDAF spatiotemporal fusion provided in the embodiments of this application, which will not be repeated here.

[0089] Those skilled in the art will understand that all or part of the functions of the various methods in the above embodiments can be implemented by hardware or by computer programs. When all or part of the functions in the above embodiments are implemented by computer programs, the program can be stored in a computer-readable storage medium, which may include: read-only memory, random access memory, disk, optical disk, hard disk, etc., and the program is executed by a computer to achieve the above functions. For example, the program can be stored in the memory of a device, and when the program in the memory is executed by the processor, all or part of the above functions can be achieved. In addition, when all or part of the functions in the above embodiments are implemented by computer programs, the program can also be stored in a server, another computer, disk, optical disk, flash drive, or external hard drive, etc., and can be downloaded or copied to the memory of a local device, or the system of the local device can be updated. When the program in the memory is executed by the processor, all or part of the functions in the above embodiments can be achieved.

[0090] The above examples illustrate the present invention only to aid in understanding it and are not intended to limit the scope of the invention. Those skilled in the art can make various simple deductions, modifications, or substitutions based on the principles of this invention.

Claims

1. A multivariate remote sensing identification method for sensitized artemisia based on FSDAF spatiotemporal fusion, characterized in that, The method includes: By detecting Pettitt mutation points in the surface meteorological monitoring data of the target area, the critical time point of the vigorous growth period of sensitized Artemisia in the target area was determined. High-resolution remote sensing images, UAV aerial images, and radar images of the target area during the vigorous growth period of sensitized wormwood after the critical time point are collected. The collected multi-dimensional image data are fused using the FSDAF spatiotemporal fusion model to generate a large-area, long-term sequence image dataset of the growth and development characteristics of sensitized wormwood. Based on the obtained image dataset of sensitized Artemisia growth and development characteristics, feature extraction was performed from four dimensions: spectrum, texture, density, and geometry, resulting in a sensitized Artemisia feature dataset that includes normalized vegetation index, enhanced vegetation index, pixel shape index, texture feature index, and spectral feature index. Based on the aforementioned sensitized artemisia feature dataset, the trained CART classification tree algorithm is used for classification and identification processing, outputting the identification results of sensitized artemisia types and distributions.

2. The multivariate remote sensing identification method for sensitized artemisia based on FSDAF spatiotemporal fusion as described in claim 1, characterized in that, By performing Pettitt mutation point detection on surface meteorological monitoring data of the target area, the critical time points of the vigorous growth period of sensitized artemisia in the target area were determined, specifically including: Based on the surface meteorological monitoring data series of the target area, calculate the Mann-Whitney nonparametric statistic: ； In the formula, t = 2, 3, 4, ..., n; is a statistical sequence; sgn is the sign function; X t X is the t-th data point in the statistical sequence; i is the i-th data point in the statistical sequence; n is the number of data points in the statistical sequence. Calculate the mutation point and significance level p-value based on the Mann-Whitney nonparametric statistic: ； ； In the formula, K t,N If p≤0.05, the mutation point is determined to be valid, and the corresponding mutation point is the critical time point of the vigorous growth period of the sensitized Artemisia in the target area.

3. The multivariate remote sensing identification method for sensitized artemisia based on FSDAF spatiotemporal fusion as described in claim 1, characterized in that, The normalized vegetation index is calculated as follows: ； In the formula, NDVI is the normalized vegetation index; It is in the near-infrared band; It is in the red light band.

4. The multivariate remote sensing identification method for sensitized artemisia based on FSDAF spatiotemporal fusion as described in claim 1, characterized in that, The enhanced vegetation index is calculated as follows: ； In the formula, denoted as , where is the surface reflectance in the near-infrared, red, and blue bands; G is the gain factor; C1 and C2 are atmospheric impedance coefficients; and L is the soil adjustment parameter.

5. The multivariate remote sensing identification method for sensitized artemisia based on FSDAF spatiotemporal fusion as described in claim 1, characterized in that, The pixel shape index is calculated as follows: The pixel shape index uses basic shape features as classification features for allergenic wormwoods. The basic shape features of allergenic wormwoods include the perimeter, area, compactness, and shape coefficient of similar pixel groups formed by allergenic wormwood communities. Perimeter: The perimeter of a planar object is calculated using its boundary pixel information. Eight-directional chain codes are used to represent the boundary, and the step sizes of all boundary chain codes are summed. The calculation formula is as follows: ； In the formula, L i P is the distance between adjacent pixels, and P is the perimeter of the surface object; ； Where n is the result of taking the remainder of chain code direction i with 2; Area: The area S of a planar object is the sum of the number of pixels within the object area multiplied by the ground area corresponding to each pixel. Compactness: This is a morphological indicator used to quantify the compactness of a two-dimensional shape. It measures how close the shape is to a perfect circle, and is calculated as follows: ； In the formula, S is the area of ​​the planar object, and P is the perimeter of the planar object; Shape factor: An indicator used to quantify the compactness or near-circularity of a shape; the calculation formula is as follows: ； In the formula, F is the shape coefficient, S is the area of ​​the planar object, and P is the perimeter of the planar object.

6. The multivariate remote sensing identification method for sensitized artemisia based on FSDAF spatiotemporal fusion as described in claim 1, characterized in that, Texture feature index is calculated as follows: Texture feature indices include contrast, correlation, energy, homogeneity, and second-order entropy; Contrast ratio: Reflects the brightness difference between a given pixel and its neighboring pixels. The formula is as follows: ； In the formula f CON For contrast, x and y are the pixels with grayscale values ​​of x and y, respectively, and N is the total number of pixels. Let y be the probability of a pixel with gray level y appearing at a distance d from a pixel with gray level x and in the direction θ. Correlation: Represents the degree of similarity between elements in the spatial gray-level co-occurrence matrix in the row or column direction. Therefore, the correlation value reflects the gray-level correlation of the layout in the image. The formula is as follows: ； In the formula f COR For correlation, x and y are the pixels with gray levels x and y, respectively, and N is the total number of pixels. Let x be the mean of the grayscale values. Let be the probability of a pixel with gray level y appearing at a distance d from a pixel with gray level x and in the direction θ. Let x be the variance of the gray level. Let be the variance of the grayscale value y; Energy: This is a measure of the amount of information contained in an image, representing the degree of non-uniformity or complexity of texture in the image. The formula is as follows: ； In the formula f ENE Let N be the energy, x and y be the pixels with grayscale values ​​of x and y, respectively, and N be the total number of pixels. Let y be the probability of a pixel with gray level y appearing at a distance d from a pixel with gray level x and in the direction θ. Homogeneity: This is a measure describing the uniformity of pixels in an image, indicating whether the distribution of pixels in the image is uniform. The formula is as follows: ； In the formula f HOM To represent homogeneity, x and y are the pixels with gray levels x and y, respectively, and N is the total number of pixels. Let y be the probability of a pixel with gray level y appearing at a distance d from a pixel with gray level x and in the direction θ. Second-order entropy: Characterizes the spatial features of grayscale information, reflecting the comprehensive characteristics of the grayscale value at a certain pixel location and the grayscale distribution of surrounding pixels. The formula is as follows: ； In the formula H ENT Let N be the second-order entropy, x and y be the pixels with gray levels x and y, respectively, and N be the total number of pixels. Let be the probability of a pixel with gray level y appearing at a distance d from a pixel with gray level x and in the direction θ. for The constant logarithm.

7. The multivariate remote sensing identification method for sensitized artemisia based on FSDAF spatiotemporal fusion as described in claim 1, characterized in that, The spectral characteristic index is calculated as follows: Spectral characteristics include spectral brightness and mean values ​​for each band; The band mean is calculated as follows: ； In the formula, The average grayscale value is the number of rows in the image, where N is the number of rows in the image. This represents the row and column position of a unit pixel in the image. The grayscale value of a pixel in the image; The spectral brightness is calculated using a weighted average value, and the calculation formula is as follows: ； In the formula, Pi is the spectral brightness in the row direction of the image, which is the gray-weighted average value in the row direction of the image. Pi is the weight of the pixels in the image in the row direction.

8. The multivariate remote sensing identification method for sensitized artemisia based on FSDAF spatiotemporal fusion as described in claim 1, characterized in that, The method further includes: Based on the ground survey data of allergenic artemisia, the identification results are interactively interpreted and optimized, and the distribution and area of ​​allergenic artemisia types are statistically analyzed based on the optimization results.

9. A multi-element remote sensing identification system for sensitized artemisia based on FSDAF spatiotemporal fusion, characterized in that, The system includes: The Pettitt mutation detection module is used to detect Pettitt mutation points in the ground meteorological monitoring data of the target area and determine the critical time point of the vigorous growth period of sensitized wormwood in the target area. The FSDAF spatiotemporal fusion module is used to collect high-resolution remote sensing images, UAV aerial images, and radar images of the target area during the vigorous growth period of sensitized wormwood after the critical time point. The FSDAF spatiotemporal fusion model is used to fuse the collected multi-dimensional image data to generate a large-area, long-term sequence image dataset of the growth and development characteristics of sensitized wormwood. The feature extraction module is used to extract features from four dimensions—spectrum, texture, density, and geometry—based on the obtained image dataset of sensitized wormwood growth and development. This results in a sensitized wormwood feature dataset that includes normalized vegetation index, enhanced vegetation index, pixel shape index, texture feature index, and spectral feature index. The identification module is used to perform classification and identification processing based on the sensitized artemisia feature dataset using a trained CART classification tree algorithm, and output the identification results of the sensitized artemisia type and distribution.

10. An electronic device, characterized in that, The device includes: a processor and a memory; The memory is used to store one or more program instructions; The processor is configured to run one or more program instructions to perform the steps of a multivariate remote sensing identification method for sensitized artemisia based on FSDAF spatiotemporal fusion as described in any one of claims 1 to 8.

Images