A Method and System for Diagnostic of Secondary Degradation of Alpine Grasslands Based on Multi-Source Data

Abstract

This invention relates to the field of UAV image processing technology, specifically to a method and system for diagnosing secondary degradation of alpine grasslands based on multi-source data. Existing technologies suffer from the following drawbacks: low image processing accuracy, large data processing errors, and a lack of spatiotemporal comparability. To address these shortcomings, this invention first acquires images using the hypotenuse target method and calculates the modulation transfer function characteristic parameters to construct a multi-scale Gaussian space. An adaptive threshold algorithm is used to determine and fix thresholds, achieving binarization of the multi-scale images. Connected components meeting certain conditions are extracted from the binary images at each scale, and their directional robust perimeters are calculated, forming a perimeter sequence. Based on the perimeter sequence, a logarithmic scale-perimeter curve is established, and the second derivative is calculated to determine the collapse scale and standard scale, calculating the geometric features of the connected components. Combining degradation judgment thresholds and proportional constraints, crack-type early degradation patches satisfying the condition with the maximum negative second-order curvature are identified, achieving automatic identification of secondary grassland degradation.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) image processing technology, and more specifically, to a method and system for diagnosing secondary degradation of alpine grasslands based on multi-source data. Background Technology

[0002] Alpine meadows are a typical ecosystem type on the Qinghai-Tibet Plateau, represented by dwarf Kobresia communities. These meadows have thick root mats and strong soil-fixing and water-retention functions. Secondary degradation of alpine meadows refers to the re-entry of initially restored alpine grasslands into a degraded state due to natural or human disturbance. This is manifested by the appearance of degraded patches, a sharp drop in vegetation cover, invasion of toxic weeds, damage / cracking / breakage of the sod, expansion of bare patches, the appearance of freeze-thaw fissures, connection of bare patches, and accelerated loss of soil carbon and nitrogen, ultimately degrading into secondary bare land ("black soil beaches"). Unlike primary degradation, secondary degradation is often accompanied by complete destruction of the sod, leading to the loss of soil water and heat regulation functions and significantly increasing the difficulty of restoration.

[0003] In natural environments, aeolian dust and fine-grained sediments carried by runoff intermittently fill fissures, causing them to exhibit alternating bands of high reflectance (exposed sections) and low reflectance (filled sections) in the surface spectrum. The optical blur function (PSF) and pixel sampling size of UAV imaging systems limit their ability to capture contrast variations exceeding a threshold. When the actual width of a fissure is smaller than the pixel size, only segments of the partially filled fissure are visible in the image, resulting in the originally continuous linear boundary being fragmented into discontinuous short lines or alternating bright and dark dashed lines. These dashed boundaries are systematically amplified during image stitching or hierarchical thresholding, leading to an overestimation of the fissure perimeter and distortion of its shape index. This results in errors in subsequent data processing.

[0004] As image resolution or smoothing scale changes, the gaps in the dashed fissures become connected into continuous lines at a certain level of ambiguity, resulting in a sudden drop in perimeter. Due to differences in UAV platforms, flight altitudes, lens parameters, and post-processing algorithms, the location of the fissure connection inflection point varies across images, leading to significant inconsistencies in the appearance of the same land cover across different data sources. Existing degradation identification methods typically rely on fixed thresholds or manual interpretation, failing to adaptively determine the location of this inflection point or identify which scale of boundary in the image best represents the true surface structure. Consequently, the identification of early-degraded patches in alpine meadows exhibits inconsistencies between different batches within the same region, severely impacting the spatiotemporal comparability and quantitative accuracy of degradation monitoring. Summary of the Invention

[0005] To address the problems of existing technologies, this invention provides a method and system for diagnosing secondary degradation of alpine grasslands based on multi-source data. The purpose is to identify secondary degradation of alpine grasslands by using alpine grassland images as input images.

[0006] A method for diagnosing secondary degradation of alpine grasslands based on multi-source data includes the following steps: Step 1: Obtain the original image of the grassland; Step 2: Calculate the maximum negative second-order curvature, perimeter of the connected region, area of ​​the connected region, and roundness of the connected region from the obtained original grassland image; The extraction steps for the maximum negative second-order curvature, the perimeter of the connected region, the area of ​​the connected region, and the roundness of the connected region are as follows: Step 2.1 Calculate the characteristic parameters of the modulation transfer function, and calculate the point spread function parameters based on the characteristic parameters to serve as the starting scale of the Gaussian scale space; Step 2.2 Construct a Gaussian scale space based on the initial scale, and perform multi-scale Gaussian smoothing on the original grassland image to obtain a multi-scale smoothed image; Step 2.3 Use an adaptive thresholding algorithm to determine the target threshold on the multi-scale smoothed image at the initial scale, fix the target threshold and use it for binarization of the multi-scale smoothed image to obtain the binary image corresponding to each scale. Step 2.4 In the binary images at each scale, extract the connected components that satisfy the area condition and obtain the outer boundary of each connected component; use the orientation-robust perimeter estimation method to calculate the perimeter of each outer boundary at the corresponding scale, forming a perimeter sequence arranged by scale; Step 2.5 Construct a logarithmic scale-perimeter curve based on the perimeter sequence and calculate the corresponding second derivative to obtain the maximum negative second curvature; determine the scale corresponding to the maximum negative second curvature as the collapse scale, and determine the first scale after the collapse scale as the standard scale; determine the binary image corresponding to the standard scale, thereby determining the perimeter, area, and roundness of the connected domain; Step 3: Set degradation judgment thresholds and proportion limits to diagnose secondary degradation of alpine grasslands.

[0007] Preferably, step 1 uses the oblique target method to obtain the original image of the grassland, also known as the oblique target image.

[0008] Preferably, step 2.1 specifically includes the following steps: Step a: Resample the oblique target image along the edge normal direction to obtain the edge spread function; calculate the first difference of the edge spread function to obtain the line spread function; perform a discrete Fourier transform on the line spread function and normalize it at zero frequency to obtain the modulation transfer function. The modulation transfer function satisfies: ; in, For modulation transfer function, The line spread function is obtained by discrete Fourier transform at frequency. The value at that location, The value of the line spread function at frequency 0 after discrete Fourier transform; Step b, determine on the modulation transfer function that... equal to half the spatial frequency ,Will As a characteristic parameter of the modulation transfer function; Step c, making the point spread function Gaussian in form, combined with the characteristic parameters of the modulation transfer function, yields the corresponding modulation transfer function satisfying: ; in, An indicator function with the natural constant e as its base. Given the parameters of the point spread function to be solved, we obtain... ; Step d, will As the starting scale of Gaussian scale space .

[0009] Preferably, step 2.2 specifically includes the following steps: Step A, starting with the scale Based on this, set the scale ratio. Construct an absolute scale sequence with a scale level K > 1 and a scale layer number K > 1. Where, 0≤ ≤K-1, It is a positive integer; Step B, Calculate the increment scale Where, when k=0, =0; Step C, construct the incremental Gaussian kernel. Convolve the incremental Gaussian kernel with the oblique target image to obtain a multi-scale smooth image: in, A collection of smooth images at multiple scales. These are pixel coordinates.

[0010] Preferably, step 2.3 specifically includes the following steps: Step ①: Determine the starting scale Corresponding multi-scale smoothed images ; Step ②: Use Otsu's method as the adaptive threshold algorithm; Step 3, setting The number of gray levels is Calculate its normalized histogram , 0≤ ≤ -1; Set a candidate threshold t and calculate the corresponding inter-class variance: in, for Pixel-like ratio, for Pixel-like ratio; Indicates candidate threshold The corresponding inter-class variance, 1≤ ≤ -2, It is a positive integer; Step 4: Select the candidate threshold that maximizes the inter-class variance. As the target threshold ; Step 5: Fix the target threshold for the multi-scale smoothing image set. Binarization for each scale of According to the formula: Generate binary image A set of binary images is obtained. .

[0011] Preferably, step 2.4 specifically includes the following steps: Step S1: For the binary image at each scale The set of connected components is obtained by using eight-neighbor connectivity labels. Calculate the pixel area of ​​each connected component. Retain connected components that satisfy the condition that the pixel area is greater than or equal to a preset pixel area threshold; Step S2: Determine the outer boundary of the retained connected components. : in, These are pixel coordinates; Step S3, set the direction set and pixel side length d Calculate the number of intersections between the outer boundary and grid lines in each direction. According to the formula: in The perimeter is calculated using the weights of direction d. ; Step S4: Arrange the perimeters of the same connected component at different scales in ascending order of scale to form a perimeter sequence. : 。

[0012] Preferably, step 2.5 specifically includes the following steps: Step (1), for each absolute scale Convert to logarithmic scale, the formula is: ; Step (2), using a logarithmic scale x-axis, corresponding to perimeter Using the vertical axis, construct the first... Logarithmic scale-perimeter curve of connected components ; Step (3), for logarithmic scale-perimeter curve middle Calculate the forward first-order slope at interior points With the backward first slope The discrete value of the second derivative is calculated based on the first-order slope, using the following formula: ;in, For logarithmic scale-perimeter curves Discrete values ​​of the second derivative at point; Step (4), obtain the first A connected component in The set of discrete values ​​of the second derivative within the range With the corresponding absolute scale set , Find to make - Maximum Scale Index The corresponding absolute scale is determined to be the collapse scale. ; Step (5), select the collapse scale index The next adjacent index The absolute scale corresponding to this index is determined as the standard scale. ; Step (6): Select the binary image corresponding to the standard scale, and the first binary image in the binary image. Connected components ; Step (7): Extract the perimeter corresponding to the standard scale from the perimeter sequence. ; Step (8): Calculate the area of ​​the connected components by counting pixels. ; Step (9), according to the formula Calculate roundness .

[0013] Preferably, step 3 specifically includes the following steps: Step 3.1, for the first Given a connected component, take its position... The set of discrete values ​​of the second derivative within the range The largest second-order derivative discrete value in the set of second-order derivative discrete values ​​is selected to obtain the first... The maximum negative second curvature of a connected region ; Step 3.2: Within the same oblique target image, select connected components corresponding to non-degenerate continuous boundaries to form a sample set, and calculate the sample set. Median of each connected component robust standard deviation According to the formula Set a degradation threshold ;in, It is a constant greater than or equal to 3; Step 3.3, according to the formula Set proportional limits for the collapse scale, among which For the first The collapse scale of a connected domain It is a constant greater than 1; Step 3.4: When the maximum negative second-order curvature of the connected component is greater than or equal to the degradation judgment threshold and the collapse scale meets the proportional limit, it is marked as a fissure-type early degradation patch, including: For the first For each connected component, obtain its maximum negative second-order curvature, degradation threshold, collapse scale, pixel side length, and scaling constant, and construct a judgment function. : Step 3.6, based on the decision function The value of is the first Each connected component is assigned a label; Step 3.7: If the value of the determination function is 1, then it is marked as a crack-type early degradation plaque, and the first... The area, circularity, and perimeter of a connected component.

[0014] This invention also provides a system for implementing the above-mentioned method for diagnosing secondary degradation of alpine grasslands based on multi-source data, comprising: Input module: Input the original image of the grassland; The calculation module calculates the corresponding parameters from the original image; The diagnostic module sets degradation judgment thresholds and proportion limits to diagnose secondary degradation of alpine grasslands.

[0015] The present invention also provides a computer-readable storage medium having a computer program stored thereon for implementing the above-described method for diagnosing secondary degradation of alpine grasslands.

[0016] The technical solution of the present invention achieves the following beneficial technical effects: 1. By introducing the inclined target method to determine the modulation transfer function and point spread function parameters of the UAV image system, a Gaussian scale space matching the image resolution was established, and quantitative control of image blur characteristics was realized. 2. Combining multi-scale Gaussian smoothing, adaptive threshold segmentation, and direction-robust perimeter estimation algorithms, a degradation feature identification method based on logarithmic scale-perimeter curve second derivative analysis is proposed, thereby reducing data processing errors. 3. It can adaptively determine the standard scale of fractured degraded patches, effectively overcoming the inconsistencies in identification caused by image resolution, lighting conditions, and surface differences in traditional manual interpretation or fixed threshold methods. This enables rapid and accurate automatic identification of early degradation of alpine meadows.

[0017] Obviously, based on the above description of the present invention, and according to common technical knowledge and conventional methods in the field, various other modifications, substitutions or alterations can be made without departing from the basic technical concept of the present invention.

[0018] The following detailed embodiments further illustrate the above-described content of the present invention. However, this should not be construed as limiting the scope of the present invention to the following examples. All technologies implemented based on the above-described content of the present invention fall within the scope of the present invention. Attached Figure Description

[0019] Figure 1 This is a flowchart of a method for identifying degraded patches in alpine meadows based on UAV imagery, according to the present invention. Detailed Implementation

[0020] It should be noted that the algorithms for data acquisition, transmission, storage and processing steps not specifically described in the embodiments, as well as the hardware structures and circuit connections not specifically described, can all be implemented using content already disclosed in the prior art.

[0021] Example 1: A Diagnostic Method for Secondary Degradation of Alpine Grasslands Based on Multi-Source Data Input module: Input the original image of the grassland; The calculation module calculates the corresponding parameters from the original image; The diagnostic module sets degradation judgment thresholds and proportion limits to diagnose secondary degradation of alpine grasslands.

[0022] The process framework of this example is as follows: Figure 1 As shown.

[0023] The oblique target method is used to acquire oblique target images, and the modulation transfer function characteristic parameters are calculated. Based on the characteristic parameters, the point spread function parameters are calculated as the starting scale of the Gaussian scale space, including: The edge spread function is obtained by resampling the oblique target image along the edge normal direction. The first-order difference of the edge spread function is obtained to obtain the line spread function. The line spread function is then subjected to discrete Fourier transform and normalized at zero frequency to obtain the modulation transfer function. The modulation transfer function satisfies: in, For modulation transfer function, The line spread function is obtained by discrete Fourier transform at frequency. f The value at that location, The value of the line spread function at frequency 0 after discrete Fourier transform; Determine on the modulation transfer function that equal to half the spatial frequency ,Will As a characteristic parameter of the modulation transfer function; If the point spread function is Gaussian in form, and considering the characteristic parameters of the modulation transfer function, then the corresponding modulation transfer function satisfies: in, An indicator function with the natural constant e as its base. Given the parameters of the point spread function to be solved, we obtain... Will As the starting scale of Gaussian scale space .

[0024] The slanted target is a standard tool for calibrating optical imaging systems. Its structure consists of a high-contrast black and white slanted edge, and its geometric parameters (edge ​​angle, strip width) are precisely calibrated to meet industry standards for calibrating the resolution and blur characteristics of optical systems.

[0025] The oblique target image is an image obtained by taking pictures of the oblique target in a flat area at the edge of the alpine meadow operation area by a UAV with the same flight altitude, focal length and exposure parameters as the meadow image. It must meet the requirements of no edge obstruction and no motion blur, and the grayscale must show obvious black and white double peaks. It is used to provide the reference data for system blur characteristic calibration.

[0026] The edge normal direction is perpendicular to the edge direction in the oblique target image and points in the direction of a sharp change in grayscale. Edge pixel sequences are extracted using an edge detection algorithm, and the edge tangent direction is obtained by fitting adjacent pixels. The perpendicular direction of the tangent is the edge normal direction, which is the direction in which the transition from black to white at the edge is most significant.

[0027] Uniform resampling along the edge normal direction with a step size of 0.1 pixels yields a one-dimensional grayscale sequence, which is the edge spread function. The imaging system blurs the ideal sharp edge of the slanted target into a grayscale gradient. The edge spread function reflects the system's blur response to the ideal edge; the smoother the gradient, the more severe the blurring.

[0028] The line spread function is obtained by taking the first-order difference of the edge spread function. Mathematically, the edge spread function is the integral of the line spread function, and the line spread function can be obtained by taking its derivative. The line spread function describes the blur distribution of a one-dimensional thin line target in an imaging system.

[0029] Performing a Discrete Fourier Transform on the line spread function converts the blurred, thin-line signal in the spatial domain into a frequency domain signal. While spatial signals struggle to distinguish the transmission capabilities of different details, frequency domain signals can separate different spatial frequency components, reflecting the system's effectiveness in transmitting various details.

[0030] The absolute value of the line spread function (LFS) at frequency f after its discrete Fourier transform represents the signal amplitude at that frequency. The LFS value at zero frequency represents the system's ability to transmit uniform signals. The modulation transfer function (MTF) ranges from 0 to 1. The closer it is to 1, the better the system responds to spatial frequencies. f The stronger the ability to convey details; The closer the value is to 0, the more difficult it is for the system to transmit details at that frequency. Normalization can eliminate the overall gain difference of the system, so that the modulation transfer function only reflects the proportion of detail transmitted.

[0031] Find the modulation transfer function curve that makes The spatial frequency with a value of 0.5 is denoted as the characteristic parameter. A value of 0.5 indicates that the system can transmit 50% of the contrast of details at that frequency. This frequency is called MTF50, which is an industry-recognized indicator of system resolution. The larger the characteristic parameter, the stronger the system's ability to transmit details. Characteristic parameter quantization represents the maximum frequency at which the system can effectively transmit high-frequency details such as early cracks.

[0032] After the superposition of the charge diffusion and other effects of the optical diffraction sensor in the imaging system, the intensity distribution of the point spread function is approximately Gaussian, and the Gaussian function is still Gaussian after Fourier transform. This makes it easier to establish the mapping relationship between the frequency domain and the spatial domain and reduce the computational complexity.

[0033] The standard deviation of the Gaussian point spread function represents the degree of diffusion of the blurred light spot in the system. The larger the standard deviation of the Gaussian point spread function, the more severe the system blurring. The solution yields... s The inherent fuzziness of the quantization system. The point spread function parameters are solved. s This serves as the initial scale for the Gaussian scale space. The initial scale for the Gaussian scale space must match the inherent fuzziness of the system; if the initial scale is smaller than... s The inability to cancel system ambiguity will lead to signal mixing; if the initial scale is larger than s This will excessively blur the details of the cracks.

[0034] In an embodiment of the present invention, a Gaussian scale space is constructed using an initial scale, and multi-scale Gaussian smoothing is performed on the UAV imagery to obtain a multi-scale smoothed image, including: Starting scale Based on this, set the scale ratio. >1 and the number of scale layers K >1, construct an absolute scale sequence ;in, 0≤ ≤K-1 , It is a positive integer; Calculate the incremental scale Among them, when 0; Construct an incremental Gaussian kernel. Convolving the incremental Gaussian kernel with the UAV image yields a multi-scale smoothed image. ： in, A collection of smooth images at multiple scales. These are pixel coordinates.

[0035] Scale ratio A value greater than 1 is used to control the rate of increase in smoothness between adjacent scales. The larger the value, the more significant the smoothing difference between adjacent scales; The closer to 1, the more subtle the smoothing differences between adjacent scales. (Selection) A value greater than 1 is used to allow the scale sequence to gradually cover the range from the inherent fuzziness of the system to a smooth range sufficient to connect the dashed lines of the cracks into a continuous line.

[0036] Scale layer A value greater than 1 indicates the total number of multi-scale smoothed images. The settings need to cover the complete changes in the crack structure, including both fine scales that can show the crack dashed lines and coarse scales that can connect the crack dashed lines.

[0037] index From 0 to A positive integer of -1, used to mark the position of each scale in the sequence, starting from 0 to correspond to the initial scale, and so on. As the scale increases, the smoothness gradually improves.

[0038] The Gaussian kernel corresponding to the incremental scale (incremental Gaussian kernel) only contains the amount of smoothing that is greater in the current scale than in the previous scale. Convolution with this kernel avoids redundant smoothing operations, significantly improving computational efficiency while ensuring continuous smoothing effects. When the value is 0, the incremental scale is 0 (because the smoothing corresponding to the initial scale is the first smoothing, and there is no preceding smoothing amount to be added).

[0039] The incremental Gaussian kernel is a weight kernel constructed based on incremental scales and conforming to a Gaussian function distribution. Its weights exhibit a characteristic of being high in the center and low at the edges, meaning that the weights of the central pixels are the largest and gradually decrease towards the surrounding pixels. This is consistent with the natural characteristic of optical systems where the blurred center is clear and the edges gradually change, which can simulate the smoothing effect of real images and avoid false structures (such as distortion at the edges of cracks) caused by unreasonable weight distribution.

[0040] Convolution is a process of performing pixel-by-pixel weighted calculations on an incremental Gaussian kernel and UAV imagery: Centered on each pixel in the image, the weights at each location within the kernel are multiplied by the corresponding pixel's grayscale value, and then all products are summed to obtain the new grayscale value of that central pixel. After traversing all pixels of the image, a smoothed image of the corresponding scale is obtained. Thus, by fusing pixel and surrounding information through weighted processing, fine noise and sub-pixel dashed lines (early cracks) in the image are smoothed, while preserving the integrity of large-scale structures such as meadow backgrounds.

[0041] Each convolution with an incremental Gaussian kernel produces a smoothed image corresponding to the absolute scale; [and so on]. The smoothed images corresponding to each scale are aggregated to form a multi-scale smoothed image set.

[0042] In an embodiment of the present invention, an adaptive thresholding algorithm is used to determine a target threshold on a multi-scale smoothed image at an initial scale. This target threshold is then fixed for binarization of the multi-scale smoothed image to obtain binary images corresponding to each scale, including: Determine the starting scale Corresponding multi-scale smoothed images ; Otsu's method is used as the adaptive thresholding algorithm: set up The number of gray levels is Calculate its normalized histogram , 0≤ ≤ -1 ; Set candidate threshold t And calculate the corresponding inter-class variance: in, for Pixel-like ratio, for Pixel-like ratio, Indicates candidate threshold The corresponding inter-class variance, 1≤ ≤ -2 , It is a positive integer; Select the candidate threshold that maximizes the inter-class variance. As the target threshold ; A fixed target threshold is applied to a multi-scale smooth image ensemble. Binarization for each scale of According to the formula: Generate binary image A set of binary images is obtained. .

[0043] The smoothed image at the initial scale cancels out the inherent blur of the device without over-smoothing, thus preserving the grayscale difference between subpixel cracks (dark areas) and the meadow background (bright areas) to the greatest extent. That is, if a coarser-scale image is selected, the crack details have been smoothed, and the grayscale difference is smaller; if a finer-scale image is selected, the device blur is not canceled, and the noise interference is greater. Therefore, using the initial-scale image as the base threshold can better distinguish between cracks and the background.

[0044] gray levels This refers to the number of grayscale levels in the initial scale image (such as the common 0-255 grayscale levels). =256), covering all possible grayscale ranges of the image.

[0045] Normalized histogram : Statistical analysis of each gray level (range 0 to) -1) The proportion of the number of corresponding pixels to the total number of pixels in the image. Its function is to reflect the gray-level distribution characteristics of the image. That is, if there are obvious dark areas with cracks and bright areas with meadows in the image, the histogram will show two peaks, which correspond to the main gray levels of the dark areas and the bright areas, respectively.

[0046] Candidate threshold Value range 1 to A positive integer -2 is used to divide all gray levels into two categories: dark class (gray value ≤ 2). (corresponding to suspected crack areas) and bright categories (grayscale value > (Corresponding to a suspected meadow background). Select 1 to... -2 is to avoid Take 0 or When the value is -1, a certain class has no pixels, making it impossible to calculate inter-class differences.

[0047] Inter-class variance is a metric that measures the difference in gray levels between the dark and light classes. The greater the difference, the higher the inter-class variance value. When the dark area of ​​the fissure and the bright area of ​​the meadow are precisely separated, the difference in grayscale between the two classes is the greatest, and the inter-class variance reaches its peak; if... If the value deviates from the optimal value, pixels from the other class will be mixed in, the difference will become smaller, and the inter-class variance will decrease.

[0048] The target threshold is applied to multi-scale smoothed images at all scales (from the initial scale to the coarsest scale), and binary images are generated according to a uniform rule: for each pixel in the image at each scale, if its gray value is less than or equal to the target threshold, it is marked as a value representing a crack region; if it is greater than the target threshold, it is marked as a value representing a meadow background. The binary images at all scales are then aggregated to form a binary image set.

[0049] In embodiments of the present invention, connected components satisfying area conditions are extracted from binary images at various scales, and the outer boundaries of each connected component are obtained; a direction-robust perimeter estimation method is used to calculate the perimeter of each outer boundary at the corresponding scale, forming a perimeter sequence arranged by scale, including: Binary images at each scale The set of connected components is obtained by using eight-neighbor connectivity labels. Calculate the pixel area of ​​each connected component. Retain connected components that satisfy the condition that the pixel area is greater than or equal to a preset pixel area threshold; Determine the outer boundary of the preserved connected components. : in, These are pixel coordinates; Set direction set and pixel side length Calculate the number of intersections between the outer boundary and grid lines in each direction. According to the formula: in The perimeter is calculated using the weights of direction d. ; Arrange the perimeters of the same connected component at different scales in ascending order of scale to form a perimeter sequence. ： For binary images at each scale, an eight-neighbor connectivity labeling algorithm is used to identify connected components. An eight-neighbor domain refers to the eight adjacent pixels of each pixel: top, bottom, left, right, top-left, top-right, bottom-left, and bottom-right. If adjacent pixels all represent fracture regions, they are considered to be in the same connected component. Compared to four-neighbor domains (only top, bottom, left, and right), eight-neighbor domains can better capture the oblique connectivity structures that may exist in alpine meadow fractures, avoiding misclassifying obliquely connected fractures as multiple independent regions and ensuring that the shape of connected components is consistent with the actual fracture distribution. Through labeling, all continuous fracture regions in the image are divided into non-overlapping connected components, forming a set of connected components.

[0050] Calculate the pixel area of ​​each connected component, i.e., count the total number of pixels representing the fracture region within that connected component. Set a preset pixel area threshold, retaining connected components with an area greater than or equal to this threshold and discarding regions with smaller areas. The preset threshold is set based on the pixel area of ​​the smallest identifiable fracture in alpine meadows (based on the conversion of image ground sampling distance; for example, if the area of ​​the smallest identifiable fracture on the ground corresponds to 4 pixels in the image, then the threshold is set to 4). This filters out image noise (such as isolated dark pixels, pseudo-fracture regions caused by small impurities), preventing these invalid regions from interfering with subsequent perimeter calculations and scale change analysis, ensuring that only connected components corresponding to real fractures are processed.

[0051] The outer boundary of a connected component refers to the set of pixels at the edge of the connected component. It is defined as follows: If a pixel within the connected component has at least one pixel representing the meadow background among its four neighboring pixels (top, bottom, left, and right, i.e., the four-neighborhood), then that pixel belongs to the outer boundary. The four-neighborhood is used to determine the boundary because the outer boundary needs to reflect the boundary line between the connected component and the background. The four-neighborhood can completely cover the main adjacent directions of the pixel, avoiding the omission of boundary pixels or the inclusion of internal pixels.

[0052] The direction set refers to multiple sampling directions (such as 0 degrees, 45 degrees, 90 degrees, 135 degrees, etc.) used to calculate the perimeter. Choosing multiple directions avoids the directional sensitivity of traditional single-direction perimeter estimation. For example, calculating the perimeter of a diagonal boundary using horizontal / vertical directions will underestimate the perimeter; multi-direction sampling can offset this bias. The pixel side length, i.e., the ground sampling distance of the image, represents the actual length corresponding to each pixel on the ground. It is a parameter that converts the number of pixels into the actual length of the ground, ensuring that the perimeter result has actual geographical significance, rather than just the number of pixels.

[0053] For the outer boundary of each connected domain, the number of intersections with each grid line in the direction set is counted. A grid line refers to a parallel line uniformly distributed along a certain direction (e.g., 0 degrees is the horizontal direction), and each intersection of the outer boundary with a grid line is counted as one intersection. The number of intersections in each direction is multiplied by the cell side length, and then a weighted average is taken for all directions (usually with equal weights) to obtain the perimeter of the outer boundary at the current scale. This method calculates the perimeter by counting intersections in multiple directions, which effectively reduces the influence of boundary direction on perimeter estimation and ensures that perimeter results can be obtained for fracture boundaries with different orientations (horizontal, vertical, and oblique).

[0054] For each preserved connected component, its perimeter calculated from binary images at all scales is collected and arranged in ascending order of scale to form a perimeter sequence for that connected component. The scale from small to large corresponds to the smoothing degree from weak to strong. The perimeter sequence reflects the trend of perimeter change of the connected component boundary (fracture) as the smoothing degree increases. In the early stage, the fracture appears as a dashed line at the fine scale, and the perimeter is overestimated. As the smoothing increases, the dashed lines are connected into continuous lines, and the perimeter drops sharply (i.e., the perimeter collapses), while the perimeter of the continuous boundary only decreases gently.

[0055] In an embodiment of the present invention, a logarithmic scale-perimeter curve is constructed based on the perimeter sequence, and the corresponding second derivative is calculated, including: Each absolute scale Convert to logarithmic scale, the formula is: ; Logarithmic scale x-axis, corresponding to perimeter Using the vertical axis, construct the first... Logarithmic scale-perimeter curve of connected components ; For logarithmic scale-perimeter curves middle Calculate the forward first-order slope at interior points With the backward first slope The discrete value of the second derivative is calculated based on the first-order slope, using the following formula: ;in, For logarithmic scale-perimeter curves The discrete value of the second derivative at point .

[0056] Converting each absolute scale to a logarithmic scale makes the originally unevenly spaced scale points more evenly distributed. That is, the intervals between fine-scale intervals are widened and the intervals between coarse-scale intervals are narrowed, ensuring that the perimeter changes across the entire scale range can be clearly presented, especially highlighting the key changes in the crack perimeter at the fine scale.

[0057] Using the transformed logarithmic scale as the abscissa (reflecting the uniform change in smoothness) and the perimeter of the connected domain calculated at the corresponding scale as the ordinate (reflecting the change in boundary length), a separate logarithmic scale-perimeter curve is constructed for each preserved connected domain.

[0058] Each curve reflects the perimeter variation trend of the same connected domain under different smoothing levels: For early-stage gap-type connected domains, at fine scales (smaller logarithmic scales), the perimeter is overestimated due to the gaps appearing as sub-pixel dashed lines, resulting in a higher curve position; as the smoothing level increases (larger logarithmic scale), the dashed lines gradually connect into continuous lines, the perimeter drops sharply, and the curve shows a steep downward segment; after entering a coarse scale (larger logarithmic scale), the perimeter tends to stabilize, and the curve becomes gentle. For non-gap-type connected domains (such as continuous background boundaries), the perimeter only decreases slowly with increasing smoothing level, and the curve is gentle throughout without any steep segments.

[0059] For each logarithmic scale-perimeter curve, focus only on the interior points (i.e., the intermediate scale points excluding the two endpoints of the curve), and calculate the forward first slope and backward first slope for each interior point.

[0060] The forward first-order slope is calculated as follows: Take the difference in perimeter between the current interior point and its next scale point, divide it by the logarithmic scale difference between these two points, and the result is the forward slope of the current point, reflecting the rate of perimeter change from the current point to the next point. The backward first-order slope is calculated as follows: Take the difference in perimeter between the current interior point and its previous scale point, divide it by the logarithmic scale difference between these two points, and the result is the backward slope of the current point, reflecting the rate of perimeter change from the previous point to the current point.

[0061] The reason for choosing to calculate the internal point is that the endpoint has only one adjacent point, making it impossible to calculate the forward and backward slopes simultaneously. Calculating two slopes avoids the possible errors that may exist in the slope of a single direction (such as an abnormal slope caused by noise affecting an adjacent point). By combining the forward and backward slopes, the rate of change of the perimeter around the current point can be reflected.

[0062] The calculation of the discrete value of the second derivative is based on the forward first slope and the backward first slope: take the difference between the forward slope and the backward slope, divide it by half of the logarithmic scale interval corresponding to these two slopes (that is, half of the logarithmic scale distance between the two scale points before and after the current point), and the result is the discrete value of the second derivative of the current interior point.

[0063] The second derivative reflects the rate of change of the perimeter, i.e., the curvature of the curve: if the second derivative is negative and its absolute value is large, it indicates that the rate of change of the perimeter suddenly changes from gentle to steep (or the steepness intensifies), corresponding to a sharp downward inflection point on the curve (i.e., the abrupt drop in perimeter in the early stages of the fracture); if the second derivative is close to zero, it indicates that the rate of change of the perimeter is stable, corresponding to a gentle section of the curve (i.e., the overall trend of the non-fracture connected domain). By calculating the discrete values ​​of the second derivative, the existence of a sharp inflection point on the curve can be transformed into a quantifiable numerical indicator.

[0064] In an embodiment of the present invention, the scale corresponding to the maximum negative second-order curvature is determined as the collapse scale, and the first scale after the collapse scale is determined as the standard scale; determining the perimeter, area, and roundness of the connected region based on the binary image corresponding to the standard scale includes: Get the A connected component in The set of discrete values ​​of the second derivative within the range With the corresponding absolute scale set , Find - Maximum Scale Index The corresponding absolute scale is determined to be the collapse scale. ; Select collapse scale index The next adjacent index The absolute scale corresponding to this index is determined as the standard scale. ; Select the binary image corresponding to the standard scale, and the first binary image in the binary image. Connected components ; Extracting the perimeter corresponding to the standard scale from the perimeter sequence ; Calculate the area of ​​connected components by pixel count. ; According to the formula Calculate roundness .

[0065] Within the set of discrete values ​​of the second derivative, we find the scale index that maximizes the maximum negative second-order curvature. The maximum negative second-order curvature is the negative value of the discrete second-order derivative; the larger the value, the sharper the inflection point of the logarithmic scale-perimeter curve (i.e., the more drastic the change in perimeter from an overestimated state to a stable state). The process corresponding to this inflection point is the smooth connection of sub-pixel dashed cracks into a continuous line; this scale is the collapse scale. The absolute scale corresponding to this index is determined as the collapse scale, thus identifying the critical smoothing scale at which key changes occur in the crack structure.

[0066] Select the next adjacent index of the scale index corresponding to the collapse scale, and determine the absolute scale corresponding to that index as the standard scale. The reason for selecting the next index is that the collapse scale is the critical scale at which the fracture has just completed the transition from a dashed line to a continuous line, at which point there may still be slight fluctuations in the boundary; while the smoothness of its next adjacent scale is sufficient to make the boundary completely stable, and the perimeter no longer drops sharply (it is in a gentle section), which can most accurately reflect the true geometric contour of the fracture.

[0067] A standard-scale binary image is selected, whose connected component shape is determined solely by the smoothing effect of the standard scale (without threshold drift interference). The first [element / component] is located from this binary image. A connected domain whose boundary has stabilized as a continuous line, and whose shape is highly consistent with the actual crack.

[0068] Calculating the area of ​​connected components using pixel counting: Statistical analysis of the first pixel in a standard-scale binary image. The total number of pixels representing the fracture region within a connected component is the pixel area of ​​the connected component. Combined with the ground sampling distance of the image (implied in the previous imaging parameters), it can be converted into the actual ground area. The pixel counting method is directly based on the pixel distribution of the binary image, which can reflect the actual coverage of the connected component and avoid area errors caused by boundary blurring.

[0069] Roundness is calculated using a preset formula. Roundness is an indicator that quantifies the degree to which the shape of a connected region deviates from a circle: the closer the roundness value is to 1, the closer the shape of the connected region is to a circle; the smaller the value, the more irregular the shape. Early-stage fractured degraded patches in alpine meadows are often irregularly polygonal in boundary due to water erosion and fine-grain infilling, and their roundness values ​​are usually small; while connected regions in non-degraded areas (such as circular bare land) have roundness values ​​closer to 1.

[0070] In embodiments of the present invention, setting a degradation determination threshold and a ratio limit includes: For the first Given a connected component, take its position... The set of discrete values ​​of the second derivative within the range The largest second-order derivative discrete value in the set of second-order derivative discrete values ​​is selected to obtain the first... The maximum negative second curvature of a connected region ; Within the same UAV aerial photographs, connective components corresponding to non-degenerate continuous boundaries are selected to form a sample set. The median of each connective component in the sample set is then calculated. the median of robust standard deviation According to the formula Set a degradation threshold ;in, It is a constant greater than or equal to 3; According to the formula Set proportional limits for the collapse scale, among which For the first The collapse scale of a connected domain It is a constant greater than 1.

[0071] In embodiments of the present invention, when the maximum negative second-order curvature of the connected domain is greater than or equal to the degradation determination threshold and the collapse scale meets the proportional limit, it is marked as a fissure-type early degradation patch, including: For the first For each connected component, obtain its maximum negative second-order curvature, degradation threshold, collapse scale, pixel edge length, and scaling constant, according to the formula: Construct a decision function ; According to the decision function The value of is the first Each connected component is assigned a label: If the determination function takes the value of 1, it is labeled as a crack-type early degradation plaque, and the first value is output. The area, roundness, and perimeter of each connected component.

[0072] Within the same UAV aerial photographs as the connected component to be determined, a sample set is formed by selecting connected components corresponding to non-degenerate continuous boundaries. Non-degenerate continuous boundaries refer to naturally formed continuous contours in a meadow background (such as the edge of a complete meadow or the boundary of non-fractured bare land). These boundaries do not exhibit dashed lines due to fine-grained filling, and their perimeter changes gently with scale. Their maximum negative second-order curvature can represent the normal curvature level of non-degenerate structures. Selecting samples from the same aerial photographs ensures that the imaging conditions (illumination, resolution, blur characteristics) of the samples are consistent with those of the connected component to be determined, avoiding threshold deviations caused by differences in imaging from different aerial photographs.

[0073] First, the median of the maximum negative second-order curvature of each connected component in the sample set is calculated. The median effectively resists extreme values ​​in the sample (such as curvature anomalies in a non-degenerate connected component due to local noise) and reflects the curvature concentration level of non-degenerate structures better than the mean. Next, the robust standard deviation is calculated: first, the absolute deviation of each sample's curvature from the median is calculated; then, the median of these absolute deviations (i.e., absolute median deviation) is taken; finally, this value is multiplied by a fixed coefficient to obtain the robust standard deviation. The robust standard deviation also resists extreme values, ensuring that statistical parameters are not affected by individual outliers.

[0074] Based on the median and robust standard deviation, a degradation threshold is set according to a preset formula, which includes a constant greater than or equal to 3. The selection of this constant is based on the principle of statistical significance: when the maximum negative second-order curvature of the connected component to be determined exceeds the median plus 3 times the robust standard deviation, the probability that it belongs to a non-degenerate structure is extremely low, ensuring that the threshold can effectively distinguish between the sharp curvature of split gaps and the gentle curvature of non-split gaps.

[0075] The proportional limit for the collapse scale is set according to a preset formula, which includes the [missing information]. The collapse scale of each connected component, the pixel side length (i.e., the image ground sampling distance, representing the actual ground length of a single pixel), and a scaling constant greater than 1.

[0076] Setting proportional limits to screen for reasonable collapse scales: The collapse scale of early cracks (the critical scale for the transition from dashed lines to continuous lines) should be on the same order of magnitude as the cell edge length (because the crack width is close to the cell edge length). If the collapse scale is too small (much smaller than the cell edge length), it may be a false collapse caused by noise; if it is too large (much larger than the cell edge length), it may be a gradual change in a coarse continuous boundary rather than a crack. A proportional constant greater than 1 is used to limit the collapse scale to be greater than a certain multiple of the cell edge length, ensuring that it corresponds to the connectivity scale of a real crack, rather than an invalid noise scale, further reducing the risk of misjudgment.

[0077] For the first For each connected component, its maximum negative second-order curvature, degradation threshold, collapse scale, pixel edge length, and scaling constant are collected. A decision function is constructed according to a preset formula. The logic of this function is that two conditions must be met simultaneously: the decision function takes a value of 1 only when the maximum negative second-order curvature is greater than or equal to the degradation threshold (possessing the sharp perimeter collapse characteristic of the crack) and the collapse scale meets the scaling limit (possessing a reasonable crack connectivity scale); otherwise, the value is 0.

[0078] The dual-condition judgment is chosen because a single condition has limitations: if only curvature is considered, structures that are not cracked but have large local curvature may be misjudged as degenerate; if only scale is considered, structures with reasonable scale but gentle curvature may be misjudged as degenerate. Combining the two conditions can complement each other and significantly improve the accuracy of the judgment.

[0079] Based on the value of the decision function, the first... Each connected component is labeled: a value of 1 indicates an early-stage degraded patch of fissure type; a value of 0 indicates a non-degraded structure. The system also outputs the area, roundness, and perimeter of this connected component at a standard scale. These morphological parameters provide specific quantitative data for subsequent degradation monitoring (e.g., fissure area expansion rate analysis) and ecological assessment (e.g., studies on the correlation between degradation degree and morphology).

[0080] For example, in a preferred embodiment, the output area can be used to assist in the grading of the degree of degradation, for example, according to the following grading criteria: The NDVI value can be mapped to the degraded area, thereby inferring the grassland status through the area output in this embodiment and realizing grassland classification.

[0081] The technical solution of the present invention will be further illustrated by the following experiments. In the following experimental examples, the model structure or method steps not specifically described are the same as those described in Example 1.

[0082] Experimental Example 1: Case Study on Diagnosis and Post-Degradation Monitoring of Alpine Grasslands 1. Diagnostic methods and systems The method and system described in Example 1 are used.

[0083] 2. Implementation method In the experimental examples of this invention, a degradation diagnosis stage and a post-monitoring stage are included; The degradation diagnosis phase (pre-repair) includes: Step 1, Image Acquisition and Parameter Calibration: In the flat area of ​​the project area, oblique target images are captured using the oblique target method. Edge spread functions are obtained by resampling along the edge normal direction. The modulation transfer function characteristic parameters are calculated using first-order difference and discrete Fourier transform. The point spread function parameters are then solved as the initial Gaussian scale spatial scale. Simultaneously, orthophotos of the project area are captured by UAV, ensuring complete coverage of the treated area.

[0084] Step 2, Multi-scale image processing: Set the scale ratio based on the initial scale. >1. Number of scale layers >1, construct a Gaussian scale space, perform multi-scale Gaussian smoothing on the UAV imagery, and obtain a multi-scale smoothed image set.

[0085] Step 3, binarization and connected component extraction: On the initial scale-smoothed image, the Otsu method is used to determine the target threshold, and the threshold is fixed to perform binarization processing on all scale-smoothed images; the eight-neighbor connected label is used to extract connected components that meet the preset area conditions, and the outer boundaries of each connected component are obtained.

[0086] Step 4, Feature Calculation and Scale Determination: The perimeter of each connected domain at different scales is calculated using the directional robust perimeter estimation method to form a perimeter sequence; a logarithmic scale-perimeter curve is constructed and the second derivative is calculated. The scale corresponding to the maximum negative second curvature is determined as the collapse scale, and the first subsequent scale is the standard scale. The perimeter, area and roundness of the connected domains at the standard scale are extracted.

[0087] Step 5, Degraded Patch Identification: Select non-degraded continuous boundary connected regions within the project area to form a sample set, calculate the median and robust standard deviation, set a degradation identification threshold (constant ≥ 3) and a collapse scale ratio limit (constant > 1); when the maximum negative second-order curvature of the connected region is ≥ the degradation identification threshold and the collapse scale meets the ratio limit, it is marked as a crack-type early degradation patch.

[0088] Step 6, Degradation Classification: In conjunction with the GB19377 standard, the identified degraded patches are integrated with vegetation cover and species richness data to complete the degradation level classification (mild, moderate, severe) and determine the implementation boundaries of restoration technologies such as no-till reseeding and artificial grassland establishment.

[0089] The post-repair monitoring phase (3 years after restoration) includes: Repeat steps 1 to 5 above to acquire drone images of the project area again and identify degraded patches, and count the patch area, coverage, roundness and other indicators.

[0090] By combining monitoring indicators such as productivity and stability, and comparing the changes in degraded patches before and after restoration, a comprehensive effect assessment is conducted in accordance with the provisions of DB54 / T0117 and NY / T3461 to determine whether the ecological restoration goals have been achieved. Before restoration: Identified early-stage degraded patches with fissures were designated as precise implementation areas for soil improvement (application of ≥150kg total nutrients >40% + 2250kg bio-organic fertilizer per hectare) and no-till reseeding (total seeding rate of 39kg / hm², including 30kg / hm² of *Leymus chinensis*, etc.) to avoid indiscriminate implementation of restoration measures. After restoration: Monitoring data served as the basis for project acceptance. If the area of ​​degraded patches decreased by ≥60%, vegetation coverage increased to over 80%, and the requirements for grassland-livestock balance were met, the restoration effect was deemed satisfactory. Areas that did not meet the standards were required to optimize soil improvement and grazing management (grazing rest for more than 45 days or grazing ban for more than 1 year).

[0091] As can be seen from the above embodiments and experimental examples, this invention constructs a secondary degradation diagnosis method for alpine grasslands based on multi-source data. This method can monitor and diagnose the degradation of alpine grasslands, improving the accuracy and speed of grassland degradation diagnosis. This invention significantly improves the automation, robustness, and spatiotemporal comparability of early-stage degraded patches in alpine meadows.

Claims

1. A method for diagnosing secondary degradation of alpine grasslands based on multi-source data, characterized in that, Includes the following steps: Step 1: Obtain the original image of the grassland; Step 2: Calculate the maximum negative second-order curvature, perimeter of the connected region, area of ​​the connected region, and roundness of the connected region from the obtained original grassland image; The extraction steps for the maximum negative second-order curvature, the perimeter of the connected region, the area of ​​the connected region, and the roundness of the connected region are as follows: Step 2.1 Calculate the characteristic parameters of the modulation transfer function, and calculate the point spread function parameters based on the characteristic parameters to serve as the starting scale of the Gaussian scale space; Step 2.2 Construct a Gaussian scale space based on the initial scale, and perform multi-scale Gaussian smoothing on the original grassland image to obtain a multi-scale smoothed image; Step 2.3 Use an adaptive thresholding algorithm to determine the target threshold on the multi-scale smoothed image at the initial scale, fix the target threshold and use it for binarization of the multi-scale smoothed image to obtain the binary image corresponding to each scale. Step 2.4 In the binary images at each scale, extract the connected components that satisfy the area condition and obtain the outer boundary of each connected component; use the orientation-robust perimeter estimation method to calculate the perimeter of each outer boundary at the corresponding scale, forming a perimeter sequence arranged by scale; Step 2.5 Construct a logarithmic scale-perimeter curve based on the perimeter sequence and calculate the corresponding second derivative to obtain the maximum negative second curvature; determine the scale corresponding to the maximum negative second curvature as the collapse scale, and determine the first scale after the collapse scale as the standard scale; determine the binary image corresponding to the standard scale, thereby determining the perimeter, area, and roundness of the connected domain; Step 3: Set degradation judgment thresholds and proportion limits to diagnose secondary degradation of alpine grasslands.

2. The method for diagnosing secondary degradation of alpine grasslands based on multi-source data according to claim 1, characterized in that, Step 1 uses the oblique target method to obtain the original image of the grassland, also known as the oblique target image.

3. The method for diagnosing secondary degradation of alpine grasslands based on multi-source data according to claim 2, characterized in that, Step 2.1 specifically includes the following steps: Step a: Resample the oblique target image along the edge normal direction to obtain the edge spread function; calculate the first difference of the edge spread function to obtain the line spread function; perform a discrete Fourier transform on the line spread function and normalize it at zero frequency to obtain the modulation transfer function. The modulation transfer function satisfies: ; in, For modulation transfer function, The line spread function is obtained by discrete Fourier transform at frequency. The value at that location, The value of the line spread function at frequency 0 after discrete Fourier transform; Step b, determine on the modulation transfer function that... equal to half the spatial frequency ,Will As a characteristic parameter of the modulation transfer function; Step c, making the point spread function Gaussian in form, combined with the characteristic parameters of the modulation transfer function, yields the corresponding modulation transfer function satisfying: ; in, An indicator function with the natural constant e as its base. Given the parameters of the point spread function to be solved, we obtain... ; Step d, will As the starting scale of Gaussian scale space .

4. The method for diagnosing secondary degradation of alpine grasslands based on multi-source data according to claim 3, characterized in that, Step 2.2 specifically includes the following steps: Step A, starting with the scale Based on this, set the scale ratio. Construct an absolute scale sequence with a scale level K > 1 and a scale layer number K > 1. Where, 0≤ ≤K-1, It is a positive integer; Step B, Calculate the increment scale Where, when k=0, =0; Step C, construct the incremental Gaussian kernel. Convolve the incremental Gaussian kernel with the oblique target image to obtain a multi-scale smooth image: in, A collection of smooth images at multiple scales. These are pixel coordinates.

5. The method for diagnosing secondary degradation of alpine grasslands based on multi-source data according to claim 4, characterized in that, Step 2.3 specifically includes the following steps: Step ①: Determine the starting scale Corresponding multi-scale smoothed images ; Step ②: Use Otsu's method as the adaptive threshold algorithm; Step 3, setting The number of gray levels is Calculate its normalized histogram , 0≤ ≤ -1; Set a candidate threshold t and calculate the corresponding inter-class variance: in, for Pixel-like ratio, for Pixel-like ratio; Indicates candidate threshold The corresponding inter-class variance, 1≤ ≤ -2, It is a positive integer; Step 4: Select the candidate threshold that maximizes the inter-class variance. As the target threshold ; Step 5: Fix the target threshold for the multi-scale smoothing image set. Binarization for each scale of According to the formula: Generate binary image A set of binary images is obtained. .

6. The method for secondary degradation diagnosis of alpine grassland based on multi-source data according to claim 5, characterized in that, Step 2.4 specifically includes the following steps: Step S1: For the binary image at each scale The set of connected components is obtained by using eight-neighbor connectivity labels. Calculate the pixel area of ​​each connected component. Retain connected components that satisfy the condition that the pixel area is greater than or equal to a preset pixel area threshold; Step S2: Determine the outer boundary of the retained connected components. : in, These are pixel coordinates; Step S3, set the direction set and pixel side length δ Calculate the number of intersections between the outer boundary and grid lines in each direction. According to the formula: in The perimeter is calculated using the weights of direction d. ; Step S4: Arrange the perimeters of the same connected component at different scales in ascending order of scale to form a perimeter sequence. : 。 7. The method for diagnosing secondary degradation of alpine grasslands based on multi-source data according to claim 6, characterized in that, Step 2.5 specifically includes the following steps: Step (1), for each absolute scale Convert to logarithmic scale, the formula is: ; Step (2), using a logarithmic scale x-axis, corresponding to perimeter Using the vertical axis, construct the first... Logarithmic scale-perimeter curve of connected components ; Step (3), for logarithmic scale-perimeter curve middle Calculate the forward first-order slope at interior points With the backward first slope The discrete value of the second derivative is calculated based on the first-order slope, using the following formula: ;in, For logarithmic scale-perimeter curves Discrete values ​​of the second derivative at point; Step (4), obtain the first A connected component in The set of discrete values ​​of the second derivative within the range With the corresponding absolute scale set , Find to make - Maximum Scale Index The corresponding absolute scale is determined to be the collapse scale. ; Step (5), select the collapse scale index The next adjacent index The absolute scale corresponding to this index is determined as the standard scale. ; Step (6): Select the binary image corresponding to the standard scale, and the first binary image in the binary image. Connected components ; Step (7): Extract the perimeter corresponding to the standard scale from the perimeter sequence. ; Step (8): Calculate the area of ​​the connected components by counting pixels. ; Step (9), according to the formula Calculate roundness .

8. The method for diagnosing secondary degradation of alpine grasslands based on multi-source data according to claim 1 or 7, characterized in that, Step 3 specifically includes the following steps: Step 3.1, for the first Given a connected component, take its position... The set of discrete values ​​of the second derivative within the range The largest second-order derivative discrete value in the set of second-order derivative discrete values ​​is selected to obtain the first... The maximum negative second curvature of a connected region ; Step 3.2: Within the same oblique target image, select connected components corresponding to non-degenerate continuous boundaries to form a sample set, and calculate the sample set. Median of each connected component robust standard deviation According to the formula Set a degradation threshold ;in, It is a constant greater than or equal to 3; Step 3.3, according to the formula Set proportional limits for the collapse scale, among which For the first The collapse scale of a connected domain It is a constant greater than 1; Step 3.4: When the maximum negative second-order curvature of the connected component is greater than or equal to the degradation judgment threshold and the collapse scale meets the proportional limit, it is marked as a fissure-type early degradation patch, including: For the first For each connected component, obtain its maximum negative second-order curvature, degradation threshold, collapse scale, pixel side length, and scaling constant, and construct a judgment function. : Step 3.6, based on the decision function The value of is the first Each connected component is assigned a label; Step 3.7: If the value of the determination function is 1, then it is marked as a crack-type early degradation plaque, and the first... The area, circularity, and perimeter of a connected component.

9. A diagnostic system for secondary degradation of alpine grasslands based on multi-source data, applied to the diagnostic method for secondary degradation of alpine grasslands based on multi-source data as described in any one of claims 1-8, comprising: Input module: Input the original image of the grassland; The calculation module calculates the corresponding parameters from the original image; The diagnostic module sets degradation judgment thresholds and proportion limits to diagnose secondary degradation of alpine grasslands.

10. A computer-readable storage medium, characterized in that: It stores a computer program for performing secondary degradation diagnosis of alpine grassland as described in any one of claims 1-8.

Images