A method for identifying grassland mushroom circles based on multispectral threshold screening and geometric constraints

Abstract

The present application relates to the technical field of remote sensing image processing and ecological monitoring, and provides a grassland mushroom ring remote sensing identification method and system based on multispectral threshold screening and geometric constraint, computer equipment, computer readable storage medium and computer program product, the identification method comprising: S1, multispectral remote sensing image acquisition and preprocessing; S2, vegetation index construction and ring band feature enhancement; S3, adaptive threshold screening to generate candidate binary graph; S4, morphological arc breaking repair and connected domain candidate region extraction; S5, geometric constraint fitting and parameter extraction; S6, false target suppression and consistency verification; S7, identification result output. The identification method provided by the present application automatically identifies the mushroom ring target in the remote sensing image and extracts the structured parameters such as center coordinates, radius and ring width, effectively enhances the ring band features of the mushroom ring and suppresses the false detection caused by complex background interference, improves the identification accuracy and reliability, and the obtained results can be used for fungal resource investigation and grassland ecological monitoring.

Description

Technical Field

[0001] This invention relates to the fields of remote sensing image processing and ecological monitoring technology, and in particular to a remote sensing identification method for grassland mushroom circles based on multispectral threshold screening and geometric constraints. Background Technology

[0002] Mushroom rings refer to the ecological phenomenon of large, concentric or heterocentric mushroom fruiting bodies growing in rings on grasslands or woodlands. This phenomenon first gained attention through observations of dense, ring-shaped distributions of vegetation in pastures, and is known internationally as fairy rings. Mushroom rings typically range in diameter from several meters to hundreds of meters, depending primarily on their age, and expand outwards by 7.6–48 cm annually. Besides circular shapes, arc-shaped distributions are also common. Grassland ecosystems are typical open ecosystems, their structure and function regulated by a combination of factors including climate, water, soil nutrients, grazing disturbance, and soil microbial processes. In addition to the vegetation community itself, soil microorganisms, especially fungi, play a crucial role in grassland organic matter decomposition, nutrient cycling, soil structure maintenance, and rhizosphere interactions. However, fungal activity and its resulting soil and vegetation responses often exhibit strong spatial heterogeneity: within the same grassland area, significant differences in soil nutrients, water conditions, and microbial community structure across different microhabitats lead to patchy vegetation growth. Traditional survey methods based on manual reconnaissance or small-scale quadrat sampling suffer from limitations in large-scale grassland areas, including limited coverage, high time and labor costs, poor repeatability, and difficulty in generating continuous spatial distribution maps. Therefore, there is an urgent need for a technological means to rapidly acquire the spatial patterns of key ecological phenomena over a large area. Against this backdrop, grassland mushroom circles, as a surface phenomenon with typical geometric morphology and clear ecological process orientation, are considered capable of characterizing the coupling relationship between underground fungal processes and aboveground vegetation responses. They possess indicative significance for grassland ecosystems and have application value in fungal resource surveys. Therefore, the identification, location, and parametric mapping of mushroom circles are practically necessary.

[0003] From a mechanistic perspective, the formation of grassland mushroom rings is usually closely related to the long-term growth and radial expansion of underground fungal mycelia and their continuous modification of the soil microenvironment. Fungi exist in the soil as a network of mycelia. Under suitable conditions, the mycelia gradually expand outwards, forming a spatial process of continuously pushing outwards along an "active front," while simultaneously producing fruiting bodies asexually. Under unfavorable conditions, the mycelia can also sexually combine to produce basidiomycetes and basidiospores, forming mushrooms. The fruiting bodies of mushroom fungi are often distributed in concentric rings, with dense, lush vegetation on both sides of these rings, forming a "green grass ring" structure distinct from other areas. Because fungal fruiting bodies require specific light, humidity, and temperature conditions to germinate in natural and semi-natural grassland ecosystems, and have a relatively short maturation period, they are often hidden among the vegetation of the "green grass ring." Therefore, the commonly referred to "mushroom ring" actually refers to the "green grass ring" structure. As mycelium expands and metabolic activities continue, it exerts a comprehensive effect on decomposing organic matter, promoting or altering the transformation of nutrients such as nitrogen and phosphorus, influencing soil aggregate structure and porosity, and changing soil water retention capacity. This results in differences in soil nutrient availability and water conditions within the mycelium-influenced zone relative to the surrounding background, thereby inducing observable growth differences in vegetation inside and outside this zone. Especially in many grassland environments, mushroom rings often exhibit typical "green grass rings." Essentially, the mycelium-influenced zone or the vicinity of its activity front creates a rhizosphere environment more favorable for herbaceous plant growth. For example, at certain stages, there may be increased nutrient release and availability, enhanced soil water retention capacity, increased nitrogen uptake by plants, or altered rhizosphere microbial interactions. This leads to an increase in leaf area index, chlorophyll content, and aboveground biomass in the ring, resulting in a greener, more vigorous, and higher-coverage ring-shaped structure. Meanwhile, the internal ecological effects of the mushroom ring may be opposite to those of the annulus at different stages: for example, long-term resource depletion, root competition, accumulation of certain metabolites, or changes in the microhabitat may lead to relative vegetation decline, sparseness, or even slight exposure; at other times, it may appear similar to the outside while the annulus is more prominent, thus exhibiting differences between the inner and outer rings. These processes of radial expansion of mycelium, changes in the soil environment, and differences in vegetation physiological responses result in the mushroom ring forming an approximately circular or ring-like structure in space. These characteristics have both ecological interpretive significance and provide a structural basis for subsequent remote sensing identification.

[0004] Research indicates that over 60 species of fungi form mushroom circles. Among them, *Armillaria mellea* is common in the alpine meadows and plateaus of my country, while *Tricholoma mongolicum* and *Tricholoma fulvidraco* are prevalent in the temperate grasslands of northern China. Grassland mushrooms are widely popular due to their delicious flavor and high nutritional and medicinal value. However, in recent years, climate change and intensified human activities have adversely affected the grassland ecosystem upon which mushroom circles depend. Furthermore, the long-term predatory harvesting of wild grassland mushrooms has led to a sharp decline in their numbers, even pushing them to the brink of extinction. The function of grassland mushroom fungi as "decomposers" in the ecosystem has been severely weakened, consequently impacting the structural and functional stability of grassland ecosystems. Therefore, research on the developmental succession patterns and physiological and ecological mechanisms of mushroom circles, and exploring their indicative significance for grassland ecosystem changes under the background of climate change, is crucial for protecting the biodiversity and function of grassland ecosystems and maintaining ecological balance, possessing significant scientific and economic value. At the remote sensing observation level, the formation of "green rings" causes systematic changes in the spectral response of vegetation: when the vegetation in the ring is more vigorous and the chlorophyll content and leaf area increase, the vegetation absorbs more red light and reflects more near-infrared light, resulting in relatively high or significantly abnormal values ​​in vegetation indices such as NDVI, GNDVI, and NDRE. When the vegetation in the central area declines or becomes sparse, red light absorption weakens and near-infrared reflectance decreases, and the indices may show low values ​​in the center, forming a clear contrast with the ring. Therefore, constructing vegetation indices using multispectral remote sensing images and extracting the spatial structure of "ring anomalies and internal-external contrast" is a reasonable technical approach for identifying mushroom rings. Furthermore, mushroom ring identification not only serves ecological indicator analysis but is also directly related to fungal resource surveys and utilization: due to the obvious seasonality and sporadic occurrence of mushroom fruiting bodies on the ground, it is difficult to reliably discover resources over a large area by simply relying on fruiting body surveys; while abnormal vegetation structures such as "green grass rings" have relative sustainability within a certain period of time and can serve as spatial clues for fungal activity and potential resources. This can be used to quickly screen suspected fungal-affected areas, plan field sampling routes, set up sampling points, and conduct subsequent soil rhizosphere mycelial verification, thereby improving the ability to discover, locate, verify, and manage fungal resources.

[0005] At the remote sensing characterization level, the physiological differences in vegetation caused by mushroom rings can be reflected in changes in reflectance and vegetation indices across different spectral bands. For example, when vegetation growth is more vigorous in the ring, red light absorption and near-infrared reflectance are enhanced, often leading to relatively high values ​​for vegetation indices such as NDVI, GNDVI, and NDRE at the ring, thus forming an observable remote sensing feature of a "green grass ring." In cases of vegetation decline or central degradation, the indices may exhibit patterns of difference between the ring and the center, as well as between the inner and outer regions. Therefore, constructing vegetation indices and extracting ring-shaped anomaly structures using multispectral remote sensing images has a clear mechanistic basis and promising application prospects. However, factors such as lighting conditions, seasonal phenological changes, topographic relief and shadows, bare patches and pika disturbance, grazing footprints and road ruts, and waterlogged or saline boundaries in grassland scenes can all cause local anomalies or edge structures in vegetation indices, easily forming "pseudo-ring" interference and thus creating false mushroom rings. Meanwhile, genuine mushroom rings often exhibit broken arcs, local weakening, or overlap with other vegetation patches, making it difficult to consistently output the structured geometric parameters required for mushroom rings by relying solely on simple threshold segmentation or general patch recognition methods. This hinders field verification and resource surveys. Therefore, to meet the practical needs of grassland ecological indicators and fungal resource realization, there is an urgent need to propose a grassland mushroom ring identification technology that can stably enhance the zonal features of mushroom rings under multispectral remote sensing conditions, reduce false detections caused by complex interferences, and output the center location and scale parameters, thus supporting grassland ecosystem process research, resource surveys, and management applications.

[0006] Currently, there is no research on the monitoring and identification of grassland mushroom circles. The following are the key technical problems that are difficult to solve in the practical application of remote sensing identification of grassland mushroom circles: First, grassland mushroom circles often exhibit morphological characteristics such as "ring-shaped weak contrast, broken arc, and incompleteness" in remote sensing images. They are also significantly affected by seasonal phenology, lighting conditions, sensor differences, and terrain shadows, resulting in insufficient stability of detection methods based on fixed thresholds or single features, which easily leads to missed detections or positioning errors. Secondly, the grassland environment is subject to complex disturbances. Grazing trampling marks, pika disturbances and bare spot boundaries, road ruts, water accumulation boundaries, salt spots, and brightness gradients caused by shadows and slopes often form approximately ring-shaped or arc-shaped structures, making false detection of "mushroom circles" quite serious. Traditional circle detection or simple binary segmentation is difficult to effectively suppress false targets. Third, existing grassland remote sensing identification technologies mainly focus on land classification or abnormal patch extraction, and the output format is biased towards "patches". It is difficult to reliably provide the structured parameters required for mushroom circle surveys and ecological analysis, such as circle center coordinates, radius, ring width, and internal and external contrast intensity. This is not conducive to field navigation verification, fungal resource location and grassland ecological indicator analysis.

[0007] At the remote sensing characterization level, the physiological differences in vegetation caused by mushroom rings can be reflected in changes in reflectance and vegetation indices across different spectral bands. For example, when vegetation growth is more vigorous in the ring, red light absorption and near-infrared reflectance are enhanced, often leading to relatively high values ​​for vegetation indices such as NDVI, GNDVI, and NDRE at the ring, thus forming an observable remote sensing feature of a "green grass ring." In cases of vegetation decline or central degradation, the indices may exhibit patterns of difference between the ring and the center, as well as between the inner and outer regions. Therefore, constructing vegetation indices and extracting ring-shaped anomaly structures using multispectral remote sensing images has a clear mechanistic basis and promising application prospects. However, factors such as lighting conditions, seasonal phenological changes, terrain undulations and shadows, bare patches and pika disturbances, grazing tramplings and road ruts, and waterlogged or saline boundaries in grassland scenes can all cause local anomalies or edge structures in vegetation indices, easily forming "pseudo-ring" interference, thus leading to the problem of false mushroom circles. At the same time, real mushroom circles often have broken arcs, local weakening, or overlap with other vegetation patches, making it difficult to obtain irregular "patchwork" results by relying solely on simple threshold segmentation or general patch recognition methods. This makes it difficult to stably output the structured geometric parameters required for mushroom circles, which is not conducive to field verification and resource surveys. Summary of the Invention

[0008] To overcome or alleviate one or more of the above-mentioned technical problems, the present invention aims to provide a remote sensing identification method for grassland mushroom rings based on multispectral threshold screening and geometric constraints. Addressing the practical needs of grassland ecological indication and fungal resource realization, this method can effectively enhance the zonal features of mushroom rings under multispectral remote sensing conditions, reduce false detections caused by complex background interference, and achieve automatic identification with geometric parameterized output of mushroom rings. This supports the large-scale, rapid, and repeatable application of fungal resource surveys and grassland ecological monitoring. The present invention provides a grassland mushroom ring identification technology solution that can stably enhance the zonal features of mushroom rings under multispectral remote sensing conditions, reduce false detections caused by complex interference, and output center location and scale parameters, thereby supporting grassland ecosystem process research, resource surveys, and management applications.

[0009] This invention provides the following technical solution: In a first aspect, the present invention provides a remote sensing identification method for grassland mushroom circles based on multispectral threshold screening and geometric constraints, which includes the following steps: S1. Multispectral Remote Sensing Image Acquisition and Preprocessing Multispectral remote sensing images of the study area are acquired from UAV multispectral cameras or satellite sensors with near-infrared bands. Preprocessing involves radiometric calibration, reflectance conversion, band registration, geometric and orthorectification correction, image stitching and cropping, resampling, and noise suppression to obtain multiband reflectance images. ; S2, Vegetation Index Construction and Enhancing Zonal Characteristics Based on the multi-band reflectivity image Calculate at least one vegetation index among NDVI, GNDVI, EVI, and NDRE to reflect differences in vegetation physiological state and obtain an index map; The index map is subjected to ring enhancement processing, which employs any one or any combination of high-pass filtering, band-pass filtering, multi-scale difference, and local background removal to make local ring anomalies more prominent in space, resulting in an enhanced index map. ; S3. Adaptive threshold filtering generates candidate binary images. Based on the enhancement index diagram obtained in step S2 An adaptive threshold is used to adapt to changes in the index baseline under different regions, seasonal phenology, and illumination conditions. The adaptive threshold is obtained through background statistics. Pixels that meet the threshold conditions are marked as candidate pixels, and candidate binary maps are generated. ; S4. Morphological arc-break repair and connected region candidate extraction In the candidate binary map Morphological processing is performed on the region, including at least opening operations to remove small speckle noise, closing operations to connect broken arc segments, and hole filling. Connected regions are then filtered, and regions that do not conform to the target scale and shape are eliminated by calculating at least one of the following indices: area, perimeter, roundness, and aspect ratio of the circumscribed rectangle, resulting in a candidate region set. ; S5. Geometric Constraint Fitting and Parameter Extraction For the candidate region set Geometric fitting and parameter extraction are performed one by one. For each candidate region, a set of boundary points or a set of skeleton points are extracted, and the center of the circle is obtained by least squares circle fitting or RANSAC circle fitting. With radius Alternatively, ellipse fitting can be used to adapt to non-circular shapes, and geometric constraints can be applied to the fitting results to obtain preliminary screening of mushroom ring candidates. The geometric constraints include at least radius range constraints, fitting residual constraints, and circumferential coverage constraints. S6. False Target Suppression and Consistency Verification The preliminary screening of mushroom circle candidates undergoes pseudo-target suppression and consistency verification to eliminate false loops. The consistency verification includes at least comparative consistency verification using an "inner-loop-outer" sampling structure and multi-index consistency verification. Based on the verification results, a confidence score is calculated for each candidate target. This confidence score is determined at least based on fitting error, loop coverage, loop width rationality, comparative consistency, multi-index consistency, and a mask penalty term. Low-confidence pseudo-targets are then eliminated to obtain a high-quality mushroom circle set. ; S7. Output of Identification Results Output the mushroom ring set The system identifies the center point coordinates, radius, diameter, ring width, circumferential coverage, fitting residuals, and confidence scores for each mushroom ring. The identification results are then vectorized into a file containing the circle or ellipse boundary and center point. Simultaneously, a rasterized probability map or intensity map is generated for spatial statistics and ecological analysis.

[0010] Preferably, in step S2, the NDVI calculation formula is: NDVI=(NIR−Red) / (NIR+Red) (1) In the above formula, NDVI represents the normalized vegetation index, NIR represents the near-infrared reflectance, and Red represents the red light reflectance. The formula for calculating GNDVI is: GNDVI =(NIR - Green) / (NIR + Green) (2) In the above formula, GNDVI represents the green normalized vegetation index, NIR represents the near-infrared reflectance, and Green represents the green reflectance. The formula for calculating EVI is as follows: EVI=2.5×(NIR−Red) / (NIR+6×Red−7.5×Blue+1) (3) In the above formula, EVI represents the enhanced vegetation index, NIR represents the near-infrared reflectance, Red represents the red light reflectance, and Blue represents the blue light reflectance. The formula for calculating NDRE is as follows: NDRE=(NIR−RedEdge) / (NIR+RedEdge) (4) In the above formula, NDRE represents the normalized difference red edge index, NIR represents the near-infrared band reflectance, and RedEdge represents the red edge band reflectance.

[0011] Preferably, in step S3, the background statistics method can be selected from one of the following three methods: First, manually, select grassy areas in the image that clearly do not contain mushroom rings as the background ROI, and calculate the mean. with standard deviation ; Second, using automatic grid or sliding window methods, the image is divided into blocks for statistical analysis, and after removing extreme values ​​at the upper and lower quantiles, the background statistics are summarized. Third, robust statistical methods are used to reduce the impact of outliers; Construct a lower threshold based on background statistical results. ,in This is an adjustable coefficient, ranging from 0.5 to 2.5, with an upper threshold constructed based on requirements. This forms a bandpass filter.

[0012] Preferably, the multiple geometric constraints in step S5 are: Firstly, there is a radius range constraint. Where rmin is 1m and rmax is 30m; Secondly, fitting residual constraints: calculate the average distance error or median distance error from the boundary points to the fitting circle, and eliminate unstable candidates based on this. Third, circumferential coverage constraint: Divide the circumference into several sectors according to angles, and statistically analyze the proportion of sectors with effective edge point support. And it is required that c≥c0, where the value of c0 ranges from 0.4 to 0.8.

[0013] Preferably, the pseudo-target suppression and consistency verification in step S6 includes at least the following second and fourth items: First, ring width constraint Calculate the width of the ring using distance transformation or distance between inner and outer boundaries. and demand It falls within the preset range; Second, construct an "inner-loop-outer" sampling structure. Set up a ring sampling band near the fitting radius, and set up background sampling bands inside and outside the ring, and compare the exponential mean or quantile relationship: to satisfy the inner and outer contrast relationship of the green grass ring pattern or the central degradation pattern; Third, radial gradient consistency test Extract radial profiles along multiple directions of the circumference to determine whether exponential peaks or valleys appear stably near the fitting radius and whether the gradient signs are consistent. Fourth, multi-index consistency verification Repeat the internal and external comparisons or radial profile judgments on the NDVI, NDRE, GNDVI or EVI indices, and retain the results or increase the confidence level only when multiple indices give consistent support for the annular zone position. Fifth, mask removal Construct non-vegetation or water body masks using NDVI lower limit and NDWI, or import road, building, or bare rock vector boundaries, and discard candidate rings when the overlap ratio between the candidate ring and the mask exceeds a threshold. Sixth, multi-phase stability When multiple images are available, the candidate ring center position and scale change are verified to conform to a stable or slow evolution law; and the confidence score is obtained by weighting the fitting error, coverage, ring width, internal and external contrast consistency score, gradient consistency score, multi-index consistency score and mask overlap penalty term according to preset weights.

[0014] Preferably, in step S7, for each mushroom ring target, its structured parameters are output, including center point coordinates, radius and diameter, ring width, ring coverage, fitting residual and confidence score, and the type label is output as green grass ring type or central degradation type; at the same time, key meta-information such as image acquisition time, data source type and ground resolution are recorded synchronously.

[0015] Secondly, this invention provides a grassland mushroom circle remote sensing identification system based on multispectral threshold screening and geometric constraints, comprising: The multispectral remote sensing image acquisition and preprocessing module is used to acquire multispectral remote sensing images of the study area. These images are obtained from a UAV multispectral camera or a satellite sensor with a near-infrared band. The preprocessing involves radiometric calibration, reflectance conversion, band registration, geometric and orthorectification correction, stitching and cropping, resampling, and noise suppression of the multispectral remote sensing images to obtain multiband reflectance images. ; The vegetation index construction and zonal feature enhancement module is used to construct vegetation indices based on the multi-band reflectance image. At least one vegetation index among NDVI, GNDVI, EVI, and NDRE is calculated to reflect differences in vegetation physiological states, resulting in an index map. The index map is then subjected to ring enhancement processing, employing any one or any combination of high-pass filtering, band-pass filtering, multi-scale difference, and local background removal to make local ring-shaped anomalies more spatially prominent, resulting in an enhanced index map. ; An adaptive threshold filtering module for generating candidate binary maps is used to generate candidate binary maps based on the obtained enhanced exponential map. An adaptive threshold is used to adapt to changes in the index baseline under different regions, seasonal phenology, and illumination conditions. The adaptive threshold is obtained through background statistics. Pixels that meet the threshold conditions are marked as candidate pixels, and candidate binary maps are generated. ; The morphological arc-break repair and connected component candidate region extraction module is used in the candidate binary map. Morphological processing is performed on the region, including at least opening operations to remove small speckle noise, closing operations to connect broken arc segments, and hole filling. Connected regions are then filtered, and regions that do not conform to the target scale and shape are eliminated by calculating at least one of the following indices: area, perimeter, roundness, and aspect ratio of the circumscribed rectangle, resulting in a candidate region set. ; The geometric constraint fitting and parameter extraction module is used to process the candidate region set. Geometric fitting and parameter extraction are performed one by one. For each candidate region, a set of boundary points or a set of skeleton points are extracted, and the center of the circle is obtained by least squares circle fitting or RANSAC circle fitting. With radius Alternatively, ellipse fitting can be used to adapt to non-circular shapes, and geometric constraints can be applied to the fitting results to obtain preliminary screening of mushroom ring candidates. The geometric constraints include at least radius range constraints, fitting residual constraints, and circumferential coverage constraints. The pseudo-target suppression and consistency verification module is used to perform pseudo-target suppression and consistency verification on the initially screened mushroom circle candidates to eliminate pseudo-loops. The consistency verification includes at least comparative consistency verification of constructing an "inner-loop-outer" sampling structure and multi-index consistency verification. Based on the verification results, a confidence score is calculated for the candidate targets. The confidence score is determined based at least on fitting error, loop coverage, loop width rationality, comparative consistency, multi-index consistency, and mask penalty term. Based on this, low-confidence pseudo-targets are eliminated to obtain a high-quality mushroom circle set. ; The identification result output module is used to output the mushroom ring set. The system identifies the center point coordinates, radius, diameter, ring width, circumferential coverage, fitting residuals, and confidence scores for each mushroom ring. The identification results are then vectorized into a file containing the circle or ellipse boundary and center point. Simultaneously, a rasterized probability map or intensity map is generated for spatial statistics and ecological analysis.

[0016] Thirdly, the present invention provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the steps of the above-described method for remote sensing identification of grassland mushroom circles based on multispectral threshold screening and geometric constraints.

[0017] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that, when the computer program is executed by a processor, it implements the steps of the above-mentioned method for remote sensing identification of grassland mushroom circles based on multispectral threshold screening and geometric constraints.

[0018] Fifthly, the present invention provides a computer program product, including a computer program, characterized in that, when the computer program is executed by a processor, it implements the steps of the above-mentioned method for remote sensing identification of grassland mushroom circles based on multispectral threshold screening and geometric constraints.

[0019] Compared with the prior art, the present invention has the following beneficial effects: 1. In the multispectral data acquisition step of this invention, since the physiological differences in vegetation caused by mushroom rings can be reflected in the changes in reflectance and vegetation index in different bands, "invisible" mushroom rings that cannot be identified by the naked eye and ordinary cameras can be discovered through UAV imagery, thus avoiding the problems of missed detection or positioning deviation.

[0020] 2. This invention automatically filters out interference from non-chlorophyll abnormal areas such as sheep tracks, ruts, and bare soil by threshold screening of multiple indices such as NDVI, NDRE, and GNDVI. It also employs false target suppression and consistency verification to avoid the problem of false mushroom circles, resulting in higher accuracy of the obtained mushroom circle set.

[0021] 3. The output parameters are presented in a comprehensive format, facilitating practical applications in fungal resource surveys and grassland ecological indicator analysis. Attached Figure Description

[0022] Figure 1 This is a technical roadmap for a grassland mushroom circle remote sensing identification method based on multispectral threshold screening and geometric constraints, provided in an embodiment of the present invention.

[0023] Figure 2 The original aerial images taken by the drone are provided for embodiments of the present invention.

[0024] Figure 3 This is an NDVI pseudo-color image provided for an embodiment of the present invention.

[0025] Figure 4 Candidate binary images provided for embodiments of the present invention.

[0026] Figure 5 The candidate binary image after cleaning is provided in an embodiment of the present invention.

[0027] Figure 6 The final identification image provided for an embodiment of the present invention.

[0028] Figure 7 This is a schematic diagram of the output coordinates provided in an embodiment of the present invention. Detailed Implementation

[0029] The technical principle of this invention is as follows: pasture stimulated by secretions of underground fungi, such as gibberellins and nitrogen, has a significantly higher near-infrared reflectance than the background grassland. By calculating the normalized vegetation index and setting specific high-pass or band-pass filters for threshold filtering, the originally blurry image can be transformed into a clear black-and-white binary feature map, thereby filtering out the background and highlighting the mushroom ring target.

[0030] The process of this method is as follows: data acquisition - spectral index calculation - adaptive threshold filtering - morphological denoising - geometric torus fitting - geographic coordinate output. For example... Figure 1 As shown.

[0031] This invention addresses the practical problems of grassland mushroom rings (also known as fungal rings) in remote sensing imagery, including weak target contrast, sensitivity to phenology, common arc breaks, numerous false rings due to interference, and difficulty in parameterizing existing results due to patchy distribution. It proposes a remote sensing identification method for grassland mushroom rings based on multispectral threshold screening and geometric constraints. The core of this method lies in: constructing vegetation indices that characterize differences in vegetation vitality using multispectral imagery, converting the "green grass ring / inner / outer differences" of mushroom rings on the ground surface into stably calculable index anomalies; based on this, candidate ring pixels are obtained through adaptive threshold screening, and morphological processing is used to repair arc breaks and remove noise, "purifying" the candidate regions from complex backgrounds; subsequently, a method for identifying mushroom rings is introduced. The geometric priors of the morphology, including radius range, roundness, ellipticity, fitting residuals, and circumferential coverage, are used to fit and parameterize candidate regions into circles or ellipses. Finally, through false target suppression and consistency verification strategies, including ring width constraints, inner-outer ring comparison, radial gradient consistency, multi-index consistency, non-vegetation and road / water body masking removal, and optional multi-temporal stability, the false detection of "false mushroom circles" caused by factors such as road ruts, bare patch boundaries, water accumulation boundaries, saline patches, shadows, and topographic relief in grassland scenes is significantly reduced. This results in structured results that can be used for fungal resource surveys and ecological indicator analysis, including center coordinates, radius, ring width, circumferential integrity, confidence score, and vectorized boundary.

[0032] The present invention will now be described in detail with reference to embodiments and accompanying drawings. However, it should be understood that the embodiments and drawings are for illustrative purposes only and do not constitute any limitation on the scope of protection of the present invention. All reasonable modifications and combinations included within the inventive spirit of the present invention fall within the scope of protection of the present invention.

[0033] Example 1

[0034] This embodiment provides a remote sensing identification method for grassland mushroom circles based on multispectral threshold screening and geometric constraints. The method specifically includes the following steps and can be executed by a computer program: S1 first performs image acquisition and preprocessing, acquiring multispectral remote sensing images of the study area. These images can come from a UAV multispectral camera or a satellite sensor with a near-infrared band. Radiometric calibration or reflectance conversion, band registration, orthorectification, stitching, and cropping are then performed on the images. Noise suppression and shadow processing are performed when necessary to obtain multiband reflectance images. .

[0035] In the vegetation index construction stage, S2 calculates at least one vegetation index based on the multi-band reflectance image to reflect differences in vegetation physiological state, such as NDVI, where NDVI = (NIR − Red) / (NIR + Red). Further indices such as GNDVI, NDRE, and EVI can be calculated as auxiliary indices to adapt to different cover levels and phenological stages. To enhance the ring structure of the mushroom rings and reduce the influence of large-scale brightness gradients or topographic shadows, this embodiment can also perform ring enhancement processing on the index map, such as using high-pass or band-pass filtering, multi-scale difference, or local background removal, to make the "local ring anomaly" more prominent in space, thereby obtaining an enhanced index map. .

[0036] In the candidate extraction stage (S3), this embodiment employs an adaptive threshold strategy instead of a fixed threshold to adapt to changes in the index baseline under different regions, seasonal phenology, and lighting conditions. Specifically, background ROIs without mushroom circles can be manually selected, or background statistics can be estimated across the entire image using a grid / sliding window approach. Robust statistics, such as the median and MAD, can be used to reduce the impact of outliers. Based on this, the background mean is calculated. with standard deviation Or a robust alternative, constructing a lower threshold. ,in This is an adjustable coefficient, for example, from 0.5 to 2.5, and an upper limit for the threshold can be further constructed. ,For example or upper quantile A bandpass filter is used to suppress extreme noise and non-target anomalies while retaining annular anomalies; based on this, pixels that meet the threshold conditions are marked as candidates, resulting in a binary candidate map. .

[0037] S4 Because real mushroom rings often have broken arcs, local weakening, or are affected by noise, this embodiment further performs morphological processing on the candidate binary image, including opening operations to remove small speckle noise, closing operations to connect broken arc segments, hole filling, and optional skeletonization to enhance the continuity of the ring, and filters connected components, calculating area, perimeter, and roundness (e.g., Regions that clearly do not conform to the target scale and shape are removed based on indicators such as the aspect ratio of the circumscribed rectangle, resulting in a set of candidate regions. .

[0038] In the geometric constraint fitting and parameter extraction stage, this embodiment extracts a set of boundary points or a set of skeleton points for each candidate region, and uses least squares circle fitting or RANSAC circle fitting to obtain the center of the circle. With radius And, when necessary, ellipse fitting is used to adapt to non-circular shapes caused by slope, projection, or uneven growth. To reflect the prior dimensional characteristics of the mushroom ring and avoid "misfitting a small number of arc segments as circles," this embodiment introduces several geometric constraints: one of which is the radius range constraint. For example, the value can be 1–30 m or set based on experience with the sample area; secondly, there is the fitting residual constraint, which calculates the average or median distance error from the boundary points to the fitting circle to eliminate unstable candidates; thirdly, there is the circumferential coverage constraint, which divides the circumference into several sectors according to angles and statistically analyzes the proportion of sectors with effective edge point support. and demand The value should not be lower than a preset threshold, such as 0.4 to 0.8, to exclude short arc structures formed by road boundaries or shadow edges. Through the above fitting and constraints, this embodiment can not only identify "whether mushroom rings exist", but also output structured parameters that can be used for investigation and analysis, including the center coordinates, radius, optional major and minor axes of the ellipse, fitting error, and circumferential coverage.

[0039] To address the critical issue of "excessive false mushroom circles" in grassland scenes, this embodiment further incorporates false target suppression and consistency verification steps after geometric fitting. The idea is to explicitly transform the "ecological spectral logic that mushroom circles should possess" into a computable criterion and combine it with geometric criteria. Specifically, a ring width constraint can be introduced first, and the ring width can be calculated using distance transformation or the distance between inner and outer boundaries. and demand Within a reasonable range, such as 0.2–5 m, adjustable with resolution and sample area, to distinguish between “linear boundaries / ruts” and “true rings”; secondly, construct an “inner-ring-outer” sampling structure, setting a ring sampling zone near the fitting radius, and setting background sampling zones inside and outside it, comparing the index mean or quantile relationship: for typical green grass ring patterns, the ring index should be higher than the inner and outer background (or the center degradation pattern should be lower than the ring / outer), thereby eliminating pseudo-rings that only have geometric rings but lack the support of vegetation physiological differences; thirdly, a radial gradient consistency test can be performed, extracting radial profiles along the circumference in multiple directions to determine whether the index peaks or valleys appear stably near the fitting radius and whether the gradient signs are consistent, in order to reduce false detections caused by terrain shadows or gradual brightness; Furthermore, this step supports multi-index consistency verification, which involves repeating the above-mentioned internal-external comparison or radial profile judgment on multiple indices such as NDVI, NDRE, and GNDVI. Confidence is increased or results are retained only when multiple indices provide consistent support for the ring position, thereby reducing false detections caused by single-index noise, phenological differences, or local reflection anomalies. In addition, this step can introduce a mask rejection strategy, using NDVI lower limit, NDWI, etc., to construct non-vegetation and water body masks, or importing vector boundaries such as roads, buildings, and bare rocks. When the overlap ratio between the candidate ring and the mask exceeds a threshold, it is directly rejected to suppress typical false ring sources such as road boundaries and water accumulation boundaries. When multi-temporal images are available, this embodiment can also use temporal stability verification. By comparing the center offset and radius change range of adjacent temporal phases, only targets with stable spatial positions or changes conforming to a slow expansion law are retained to further improve reliability.

[0040] Based on the above consistency verification results, a confidence score can be calculated for each candidate target in this step. The score can be obtained by combining or fusing the fitting error, coverage, ring width, internal and external consistency score, gradient consistency score, multi-index consistency score and mask overlap penalty term according to weights, and outputting a set of high-confidence mushroom circles accordingly.

[0041] Finally, S7 outputs the results, including: the center point coordinates (pixel coordinates and geographic coordinates), radius, diameter, ring width, ring coverage, fitting residuals, and confidence score of each mushroom circle. It can also vectorize the recognition results into circle or ellipse boundary and center point files, such as Shapefile, GeoJSON, or CSV navigation points, for field verification and sample point deployment. At the same time, it can generate rasterized probability maps or intensity maps for spatial statistics and ecological analysis.

[0042] The technical solution provided in this embodiment is significantly different from the general grassland identification method that only outputs irregular patches: This embodiment takes "multispectral enhancement - adaptive threshold - arc repair - geometric constraint fitting - consistency verification and false detection" as the main line, which can achieve stable identification and parameterized expression of mushroom circles in the complex background of grassland, and effectively reduce false detection of mushroom circles, thus meeting the application needs of grassland ecosystem indicator analysis and fungal resource realization (investigation, location, verification, and management).

[0043] Example 2

[0044] This embodiment provides another specific implementation, and the implementation steps are as follows: S1: Multispectral Data Acquisition In the specific image data acquisition and processing stage, a drone equipped with a multispectral camera was used to acquire grassland images of the study area, such as... Figure 2The multispectral camera includes at least red and near-infrared bands, and preferably also green and red-edge bands, to support the subsequent construction of vegetation indices such as NDVI, GNDVI, and NDRE, and enhance the separability of the mushroom ring's "green grass ring." Before aerial photography, flight path planning is required. The flight altitude is set according to the mushroom ring's scale and desired recognition accuracy to obtain a ground resolution that meets the ring's distinguishability requirements, such as 5–20 cm, the specific resolution depending on the camera parameters. Simultaneously, forward and lateral overlap are set, typically at 70%–85%, to ensure stitching quality. Appropriate flight speed and exposure parameters must also be set to avoid motion blur. Image acquisition is best performed during periods of high solar altitude and stable illumination to minimize shadow effects. Reflectivity calibration plates should be photographed before and after takeoff, or a light sensor should be used in conjunction with a calibration plate to achieve radiometric consistency. If necessary, ground control points should be deployed to improve geometric accuracy.

[0045] After acquisition, the images undergo preprocessing to generate orthoreflectance products suitable for index calculation. The preprocessing process includes radiometric calibration, reflectance conversion, inter-band registration, geometric and orthorectification correction, stitching and cropping, resampling, and noise suppression. Through these preprocessing steps, a multi-band orthoreflectance image is finally output. Its geographical reference information provides high-quality data support for subsequent vegetation index construction and mushroom circle identification.

[0046] S2: Vegetation index construction and annular feature enhancement Based on the multi-band orthoreflectivity image output in step S1 This step calculates vegetation indices to characterize differences in vegetation physiological states. The preferred method is to construct the NDVI, whose calculation formula is: NDVI=(NIR−Red) / (NIR+Red) (1) In the above formula, NDVI represents the normalized vegetation index, NIR represents the near-infrared reflectance, and Red represents the red light reflectance.

[0047] To accommodate the changes in "green ring / inner and outer differences" under different coverage and phenological stages, GNDVI and NDRE indices can be constructed in parallel. The calculation formulas for GNDVI and NDRE are as follows: GNDVI =(NIR - Green) / (NIR + Green) (2) In the above formula, GNDVI represents the green normalized vegetation index, NIR represents the near-infrared reflectance, and Green represents the green reflectance. In S2, the formula for calculating EVI is: EVI=2.5×(NIR−Red) / (NIR+6×Red−7.5×Blue+1) (3) In the above formula, EVI represents the enhanced vegetation index, NIR represents the near-infrared reflectance, Red represents the red light reflectance, and Blue represents the blue light reflectance. NDRE=(NIR−RedEdge) / (NIR+RedEdge) (4) In the above formula, NDRE represents the normalized difference red edge index, NIR represents the near-infrared band reflectance, and RedEdge represents the red edge band reflectance.

[0048] Considering the presence of large-scale brightness gradients, terrain shadows, and slowly changing backgrounds in grassland images, direct thresholding on the original index map is easily affected by background drift. Therefore, this embodiment of the invention performs ring feature enhancement processing on the NDVI index map to highlight local ring-shaped anomalies and suppress slowly changing backgrounds. The enhancement processing can be one or a combination of local background removal, high-pass or band-pass filtering enhancement, and multi-scale differential enhancement. Local background removal can obtain local background reference values ​​through a sliding window mean or median method and then perform differential normalization. Through the above enhancement processing, the mushroom ring structure can be made more prominent spatially, thus providing a more stable input for subsequent NDVI-based thresholding and morphological repair.

[0049] S3: Adaptive threshold filtering generates candidate binary images.

[0050] Based on the enhancement index map obtained in step S2 Alternatively, using the original vegetation index map, this step performs adaptive threshold filtering to extract candidate ring zone pixels. Since different flight times, seasonal phenology, and grassland cover can cause overall index shifts, a fixed threshold is difficult to apply stably. Therefore, this embodiment obtains the threshold parameter through background statistics. Background statistics can be implemented in three ways: First, manually selecting grasslands that clearly do not contain mushroom rings as background ROIs in the image and calculating the mean. with standard deviation Second, the image is divided into blocks for statistical analysis using automatic gridding or sliding window methods, and extreme values ​​at the upper and lower quantiles are removed before the background statistics are summarized. Third, robust statistics such as median and MAD are used as substitutes. , To reduce the impact of outliers, a lower threshold is constructed based on background statistical results. ,in This is an adjustable coefficient, typically ranging from 0.5 to 2.5, and an upper threshold can also be constructed according to requirements. A bandpass filter is used to suppress extreme noise or non-target anomalies while preserving annular anomalies. Pixels that meet the threshold criteria are marked as candidate pixels, generating a binary candidate map. The candidate image typically contains real mushroom rings along with a number of noisy and pseudo-ring candidates. Subsequent steps will address these issues through further denoising and filtering. Figure 4 The white area represents the candidate mushroom ring arc area, and the black area represents the background.

[0051] S4: Morphological Arc Repair and Connected Region Candidate Extraction In step S4, the binary candidate image is processed. Morphological and connected component processing is performed to achieve noise reduction, arc-breaking connections, and candidate region purification. Firstly, morphological and connected component processing is applied... Opening operations are performed to remove isolated small speckle noise and fragmented interference. The structuring element can be circular or square, and the scale is set according to the ground resolution and the expected ring width. Then, closing operations are performed on the results to connect broken arc segments and enhance the continuity of the rings, making the "broken arc / weak contrast" mushroom rings more coherent in the binary domain. If necessary, hole filling can be performed to obtain stable region representation, or skeletonization can be performed to obtain the ring centerline for subsequent fitting. After morphological processing, connected components are labeled in the binary image, and morphological indicators such as area, perimeter, roundness (e.g., 4πA / P²), and aspect ratio of each connected component are calculated. Regions that clearly do not conform to the scale and shape of the mushroom rings are removed, such as noise with too small an area, thin rut road boundaries, and excessively large background blocks, thus obtaining a set of candidate regions. and the processed binary image This provides higher-quality candidate inputs for subsequent geometric fitting. For example... Figure 5 .

[0052] S5: Geometric Constraint Fitting and Parameter Extraction For candidate region set Geometric fitting and parameter extraction are performed one by one to transform the "patchwork candidates" into "circle parameters" that can be used for investigation and analysis. Specifically, boundary point sets, such as binary boundary pixels, or skeleton point sets are extracted from each candidate region. If skeletonization has been performed in step S4, skeleton points can be directly extracted as point sets. In this embodiment, the circle fitting method is preferably used to obtain the circle center. With radius For circle fitting, either least squares circle fitting or RANSAC circle fitting can be selected, with RANSAC circle fitting being more robust to outliers and broken arcs. When the mushroom circle exhibits an approximately elliptical shape due to slope or projection, an ellipse fitting method can be used to output the center and major and minor axis parameters of the ellipse. To avoid the problem of "misfitting a small number of arc segments as circles" and to fully reflect the prior scale characteristics of the mushroom circle, this embodiment applies several geometric constraints to the fitting results: First, a radius range constraint, which can be set to r∈[rmin,rmax], typically ranging from 1 to 30 m, and can be flexibly adjusted according to the actual situation of the sample area; second, a fitting residual constraint, calculating the average distance error or median distance error from the boundary points to the fitted circle, requiring this error to be less than a preset threshold; and third, a circumferential coverage constraint, dividing the circumference into several sectors according to angles and statistically analyzing the proportion of sectors supported by boundary points. and demand The threshold value must not be lower than a preset threshold, with a common threshold range of 0.4 to 0.8. Initial screening of mushroom circle candidates is achieved through the aforementioned geometric constraints. Output parameters include center coordinates, radius, fitting error, and circumferential coverage. Simultaneously, combined with image georeferenced information, the center pixel coordinates are converted into geographic coordinates, providing accurate spatial basis for subsequent field verification and resource location.

[0053] S6: False Target Suppression and Consistency Verification To address the complex disturbances in the grassland environment and the significant problem of false mushroom circles, the preliminary candidates obtained in step S5 were further subjected to false target suppression and consistency verification. The verification approach involves explicitly transforming the "ecological spectral logic that a true mushroom circle should possess" and "ring structure consistency" into computable criteria to eliminate false rings caused by road ruts, bare patch boundaries, waterlogged boundaries, saline patch boundaries, shadows, and topographical gradations. Figure 6 The final identification image, the consistency verification may include one or more of the following combinations: one is ring width constraint, which estimates the ring width through distance transformation or inner and outer boundaries. and demand The test involves three key aspects: First, ensuring the data falls within a reasonable range to distinguish linear boundaries from true ring zones. Second, ensuring consistency between the inner and outer rings by constructing sampling zones for the inner, ring, and outer regions based on the fitted radius, and calculating statistical indices such as NDVI and NDRE. These indices must meet one of the preset patterns; for example, the green grass ring pattern requires the ring index to be higher than the inner and outer background indices, while the central degradation pattern requires the central index to be lower than the ring or outer region index. Third, ensuring radial gradient consistency by extracting radial profiles along multiple directions of the circumference to verify that index peaks or valleys appear stably near the fitted radius and exhibit consistent gradient characteristics, thus reducing the impact of variations in the fitted radius. The verification process includes several key aspects: 1) False detections caused by shadow boundaries or gradations; 2) Multi-index consistency verification, which verifies the consistency of ring positions and contrast relationships across multiple indices such as NDVI, NDRE, and GNDVI to reduce interference from single-index noise; 3) Mask removal, which calculates the overlap ratio between candidate rings and masks based on non-vegetation, water bodies, roads, or other mask or vector boundaries, and removes rings if the overlap exceeds a threshold; and 4) Multi-temporal stability verification, an optional step that verifies the stability or slow evolution of candidate ring center positions and scale changes when multiple image periods are available. Based on these verification results, a confidence score can be calculated for candidate targets. The score is obtained by fusing indicators such as fitting error, ring coverage, ring width rationality, contrast consistency, multi-index consistency, and mask penalty term. This score is used to remove low-confidence false targets, ultimately obtaining a high-quality mushroom ring set {}, providing a reliable basis for subsequent structured output results. Figure 6 .

[0054] S7: Results Output and Applications As the output of the entire identification process, the core objective is to identify the mushroom circle target. The data is transformed into readily usable survey and analysis results and visualization products, realizing the transformation of technological achievements into practical applications. For each mushroom circle target, the system needs to output its core structured parameters, including center point coordinates (covering pixel coordinates and geographic coordinates), radius and diameter, ring width, ring coverage, fitting residual, and confidence score. It can also add type labels based on the characteristics determined during the identification process, such as green grass ring or central degradation type. At the same time, key metadata such as image acquisition time, data source type, and ground resolution need to be recorded synchronously to ensure the traceability and reusability of the results data.

[0055] In terms of data output format, the needs of different application scenarios must be considered: on the one hand, the results should be vectorized to output circular or elliptical boundary features and center point features, supporting saving in common vector formats such as Shapefile, GeoJSON, and KML, which facilitates importing into GIS software for spatial analysis; on the other hand, tabular files, such as CSV or Excel formats, should be generated. Figure 7Output a coordinate diagram. Clearly list the parameter information of each mushroom circle, facilitating statistical analysis and import into field navigation equipment. For visualization, an RGB true-color image or NDVI index map can be used as a base map, overlaid with fitted circles, center points, and type labels to generate a recognition result image, intuitively displaying the spatial distribution and core parameters of mushroom circles, providing intuitive support for manual verification and results report presentation. The above output results directly address the practical needs of grassland fungal resource surveys and ecological research: structured parameters and vector data accurately support field sampling point layout and field verification navigation; statistical tables and visualization charts provide a high-quality structured spatial data foundation for analyzing the spatial distribution patterns of mushroom circles and studying their ecological indicator functions, effectively realizing a closed loop of "technical identification—results output—practical application." Figure 7 .

[0056] In one exemplary embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0057] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0058] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0059] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0060] The above embodiments are merely preferred embodiments of the present invention, and the scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A remote sensing identification method for grassland mushroom circles based on multispectral threshold screening and geometric constraints, characterized in that: Includes the following steps: S1. Multispectral Remote Sensing Image Acquisition and Preprocessing Multispectral remote sensing images of the study area are acquired from UAV multispectral cameras or satellite sensors with near-infrared bands. Preprocessing involves radiometric calibration, reflectance conversion, band registration, geometric and orthorectification correction, image stitching and cropping, resampling, and noise suppression to obtain multiband reflectance images. ; S2, Vegetation Index Construction and Enhancing Zonal Features Based on the multi-band reflectivity image Calculate at least one vegetation index among NDVI, GNDVI, EVI, and NDRE to reflect differences in vegetation physiological state and obtain an index map; The index map is subjected to ring enhancement processing, which employs any one or any combination of high-pass filtering, band-pass filtering, multi-scale difference, and local background removal to make local ring anomalies more prominent in space, resulting in an enhanced index map. ; S3. Adaptive threshold filtering generates candidate binary images. Based on the enhancement index diagram obtained in step S2 An adaptive threshold is used to adapt to changes in the index baseline under different regions, seasonal phenology, and illumination conditions. The adaptive threshold is obtained through background statistics. Pixels that meet the threshold conditions are marked as candidate pixels, and candidate binary maps are generated. ; S4. Morphological arc-break repair and connected region candidate extraction In the candidate binary map Morphological processing is performed on the region, including at least opening operations to remove small speckle noise, closing operations to connect broken arc segments, and hole filling. Connected regions are then filtered, and regions that do not conform to the target scale and shape are eliminated by calculating at least one of the following indices: area, perimeter, roundness, and aspect ratio of the circumscribed rectangle, resulting in a candidate region set. ; S5. Geometric Constraint Fitting and Parameter Extraction For the candidate region set Geometric fitting and parameter extraction are performed one by one. For each candidate region, a set of boundary points or a set of skeleton points are extracted, and the center of the circle is obtained by least squares circle fitting or RANSAC circle fitting. With radius Alternatively, ellipse fitting can be used to adapt to non-circular shapes, and geometric constraints can be applied to the fitting results to obtain preliminary screening of mushroom ring candidates. The geometric constraints include at least radius range constraints, fitting residual constraints, and circumferential coverage constraints. S6. False Target Suppression and Consistency Verification The preliminary screening of mushroom circle candidates undergoes pseudo-target suppression and consistency verification to eliminate false loops. The consistency verification includes at least comparative consistency verification using an "inner-loop-outer" sampling structure and multi-index consistency verification. Based on the verification results, a confidence score is calculated for each candidate target. This confidence score is determined at least based on fitting error, loop coverage, loop width rationality, comparative consistency, multi-index consistency, and a mask penalty term. Low-confidence pseudo-targets are then eliminated to obtain a high-quality mushroom circle set. ; S7. Output of Identification Results Output the mushroom ring set The system identifies the center point coordinates, radius, diameter, ring width, circumferential coverage, fitting residuals, and confidence scores for each mushroom ring. The identification results are then vectorized into a file containing the circle or ellipse boundary and center point. Simultaneously, a rasterized probability map or intensity map is generated for spatial statistics and ecological analysis.

2. The grassland mushroom circle remote sensing identification method based on multispectral threshold screening and geometric constraints according to claim 1, characterized in that: In step S2, the NDVI calculation formula is as follows: NDVI=(NIR−Red) / (NIR+Red) (1) In the above formula, NDVI represents the normalized vegetation index, NIR represents the near-infrared reflectance, and Red represents the red light reflectance. The formula for calculating GNDVI is: GNDVI =(NIR - Green) / (NIR + Green) (2) In the above formula, GNDVI represents the green normalized vegetation index, NIR represents the near-infrared reflectance, and Green represents the green reflectance. The formula for calculating EVI is as follows: EVI=2.5×(NIR−Red) / (NIR+6×Red−7.5×Blue+1) (3) In the above formula, EVI represents the enhanced vegetation index, NIR represents the near-infrared reflectance, Red represents the red light reflectance, and Blue represents the blue light reflectance. The formula for calculating NDRE is: NDRE=(NIR−RedEdge) / (NIR+RedEdge) (4) In the above formula, NDRE represents the normalized difference red edge index, NIR represents the near-infrared band reflectance, and RedEdge represents the red edge band reflectance.

3. The grassland mushroom circle remote sensing identification method based on multispectral threshold screening and geometric constraints according to claim 2, characterized in that: In step S3, the background statistics method can be selected from one of the following three methods: First, manually, select grassy areas in the image that clearly do not contain mushroom rings as the background ROI, and calculate the mean. with standard deviation ; Second, using automatic grid or sliding window methods, the image is divided into blocks for statistical analysis, and after removing extreme values ​​at the upper and lower quantiles, the background statistics are summarized. Third, robust statistical methods are used to reduce the impact of outliers; Construct a lower threshold based on background statistical results. ,in This is an adjustable coefficient, ranging from 0.5 to 2.5, with an upper threshold constructed based on requirements. Max forms a bandpass filter.

4. The grassland mushroom circle remote sensing identification method based on multispectral threshold screening and geometric constraints according to claim 3, characterized in that: The multiple geometric constraints mentioned in step S5 are: Firstly, there is a radius range constraint. Where rmin is 1m and rmax is 30m; Secondly, fitting residual constraints: calculate the average distance error or median distance error from the boundary points to the fitting circle, and eliminate unstable candidates based on this. Third, circumferential coverage constraint: Divide the circumference into several sectors according to angles, and statistically analyze the proportion of sectors with effective edge point support. And it is required that c≥c0, where the value of c0 ranges from 0.4 to 0.

8.

5. The grassland mushroom circle remote sensing identification method based on multispectral threshold screening and geometric constraints according to claim 4, characterized in that: Step S6, the spurious target suppression and consistency verification, includes at least the following second and fourth items: First, ring width constraint Calculate the width of the ring using distance transformation or distance between inner and outer boundaries. and demand It falls within the preset range; Second, construct an "inner-loop-outer" sampling structure. Set up a ring sampling band near the fitting radius, and set up background sampling bands inside and outside the ring, and compare the exponential mean or quantile relationship: to satisfy the inner and outer contrast relationship of the green grass ring pattern or the central degradation pattern; Third, radial gradient consistency test Extract radial profiles along multiple directions of the circumference to determine whether exponential peaks or valleys appear stably near the fitting radius and whether the gradient signs are consistent. Fourth, multi-index consistency verification Repeat the internal and external comparisons or radial profile judgments on the NDVI, NDRE, GNDVI or EVI indices, and retain the results or increase the confidence level only when multiple indices give consistent support for the annular zone position. Fifth, mask removal Construct non-vegetation or water body masks using NDVI lower limit and NDWI, or import road, building, or bare rock vector boundaries, and discard candidate rings when the overlap ratio between the candidate ring and the mask exceeds a threshold. Sixth, multi-phase stability When multiple images are available, the candidate ring center position and scale change are verified to conform to a stable or slow evolution law; and the confidence score is obtained by weighting the fitting error, coverage, ring width, internal and external contrast consistency score, gradient consistency score, multi-index consistency score and mask overlap penalty term according to preset weights.

6. The grassland mushroom circle remote sensing identification method based on multispectral threshold screening and geometric constraints according to claim 5, characterized in that: In step S7, for each mushroom ring target, its structured parameters are output, including center point coordinates, radius and diameter, ring width, ring coverage, fitting residual and confidence score, and the type label is output as green grass ring type or central degradation type; at the same time, key metadata such as image acquisition time, data source type and ground resolution are recorded.

7. A remote sensing identification system for grassland mushroom circles based on multispectral threshold screening and geometric constraints, characterized in that: include: The multispectral remote sensing image acquisition and preprocessing module is used to acquire multispectral remote sensing images of the study area, which are obtained from UAV multispectral cameras or satellite sensors with near-infrared bands. The preprocessing involves radiometric calibration, reflectance conversion, band registration, geometric and orthorectification correction, mosaicking and cropping, resampling, and noise suppression of the multispectral remote sensing image to obtain a multiband reflectance image. ; The vegetation index construction and zonal feature enhancement module is used to construct vegetation indices based on the multi-band reflectance image. At least one vegetation index among NDVI, GNDVI, EVI, and NDRE is calculated to reflect differences in vegetation physiological states, resulting in an index map. The index map is then subjected to ring enhancement processing, employing any one or any combination of high-pass filtering, band-pass filtering, multi-scale difference, and local background removal to make local ring-shaped anomalies more spatially prominent, resulting in an enhanced index map. ; An adaptive threshold filtering module for generating candidate binary maps is used to generate candidate binary maps based on the obtained enhanced exponential map. An adaptive threshold is used to adapt to changes in the index baseline under different regions, seasonal phenology, and illumination conditions. The adaptive threshold is obtained through background statistics. Pixels that meet the threshold conditions are marked as candidate pixels, and candidate binary maps are generated. ; The morphological arc-break repair and connected component candidate region extraction module is used in the candidate binary map. Morphological processing is performed on the region, including at least opening operations to remove small speckle noise, closing operations to connect broken arc segments, and hole filling. Connected regions are then filtered, and regions that do not conform to the target scale and shape are eliminated by calculating at least one of the following indices: area, perimeter, roundness, and aspect ratio of the circumscribed rectangle, resulting in a candidate region set. ; The geometric constraint fitting and parameter extraction module is used to process the candidate region set. Geometric fitting and parameter extraction are performed one by one. For each candidate region, a set of boundary points or a set of skeleton points are extracted, and the center of the circle is obtained by least squares circle fitting or RANSAC circle fitting. With radius Alternatively, ellipse fitting can be used to adapt to non-circular shapes, and geometric constraints can be applied to the fitting results to obtain preliminary screening of mushroom ring candidates. The geometric constraints include at least radius range constraints, fitting residual constraints, and circumferential coverage constraints. The pseudo-target suppression and consistency verification module is used to perform pseudo-target suppression and consistency verification on the initially screened mushroom circle candidates to eliminate pseudo-loops. The consistency verification includes at least comparative consistency verification of constructing an "inner-loop-outer" sampling structure and multi-index consistency verification. Based on the verification results, a confidence score is calculated for the candidate targets. The confidence score is determined at least based on fitting error, loop coverage, loop width rationality, comparative consistency, multi-index consistency, and mask penalty term. Based on this, low-confidence pseudo-targets are eliminated to obtain a high-quality mushroom circle set. ; The identification result output module is used to output the mushroom ring set. The system identifies the center point coordinates, radius, diameter, ring width, circumferential coverage, fitting residuals, and confidence scores for each mushroom ring. The identification results are then vectorized into a file containing the circle or ellipse boundary and center point. Simultaneously, a rasterized probability map or intensity map is generated for spatial statistics and ecological analysis.

8. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the steps of the grassland mushroom circle remote sensing identification method based on multispectral threshold screening and geometric constraints as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the grassland mushroom circle remote sensing identification method based on multispectral threshold screening and geometric constraints as described in any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the steps of the grassland mushroom circle remote sensing identification method based on multispectral threshold screening and geometric constraints as described in any one of claims 1-6.

Images