Methods and Systems for Remote Sensing Monitoring and Assessment of Pests and Diseases in Field Orchards

Abstract

This invention provides a UAV remote sensing monitoring and assessment method and system for orchard pests and diseases, belonging to the field of orchard pest and disease monitoring and assessment technology. This invention effectively eliminates interference from complex background noise by constructing a ternary separation index that integrates feature enhancement, background suppression, and brightness normalization. Based on the distribution of this index, a threshold is determined to accurately classify pixels into three categories: non-canopy background, shaded canopy, and illuminated canopy, achieving fine segmentation of unstructured scenes. A gain compensation term is constructed using the local illumination benchmark mean to correct the reflectivity curve of the shaded canopy, eliminating interference from uneven illumination. A health reference benchmark is established by selecting local high quantiles, and the physiological stress index is calculated by coupling the discriminant term and the attenuation term, enabling keen capture of early, subtle disease characteristics. Abnormal pixel differences are weighted and accumulated using a power law to output a regional hazard level index, objectively quantifying the overall disaster risk of the region through nonlinear aggregation.

Description

Technical Field

[0001] This invention relates to the field of orchard pest and disease monitoring and assessment technology, specifically to a method and system for unmanned aerial vehicle (UAV) remote sensing monitoring and assessment of pests and diseases in field orchards. Background Technology

[0002] Large-scale orchards typically feature large planting areas, complex canopy structures, and irregular row distribution. In the early stages of pest and disease outbreaks, they often exhibit scattered distribution and concealed symptoms. Therefore, it is necessary to rely on large-scale, non-contact monitoring methods to assess the overall health status of the orchard. UAV remote sensing technology, with its advantages of high acquisition efficiency, high spatial resolution, and the ability to conduct repeated observations across multiple time phases, has gradually become an important technical means for monitoring pests and diseases in orchards. Under natural lighting conditions, large-scale orchard images inevitably contain both illuminated canopy areas and a large number of shaded canopy areas. Especially in high-resolution multispectral images acquired by UAV low-altitude remote sensing, the shaded canopy pixels account for a large proportion of orchard images due to factors such as solar altitude angle, tree structure, and row occupancy.

[0003] Existing methods for monitoring orchard pests and diseases based on UAV remote sensing mostly employ vegetation indices, spectral feature statistics, or machine learning classification models to uniformly analyze canopy pixels across the entire image. These methods typically assume that the lighting conditions of canopy pixels within the same orchard are relatively uniform, or simply reduce the influence of light through brightness normalization, thereby directly using reflectance or vegetation indices to determine vegetation health. Although some methods distinguish between canopy and non-canopy areas, they still treat illuminated canopy pixels and shaded canopy pixels as the same category within the canopy itself.

[0004] However, in actual orchard settings, the spectral information of shaded canopy pixels is not invalid; changes in reflectance are extremely sensitive to the physiological state of the vegetation. Current technologies, lacking independent identification and targeted modeling of shaded canopy pixels, easily confuse reduced reflectance caused by insufficient light with physiological decline due to pests and diseases, or simply ignore shaded areas. Furthermore, within orchard areas, abiotic stresses such as water and nutrients are often influenced by topography, soil type, and irrigation systems, exhibiting strong spatial continuity, and the resulting canopy spectral changes tend to be consistent within local areas. Furthermore, the spectrum of light is smooth, while pest and disease stress usually starts from individual plants or local infection centers, exhibiting highly localized discrete and patchy distribution characteristics. Its spectral changes show abrupt and discontinuous states in space. Existing unified analysis methods are difficult to effectively distinguish between these two spatially different stress signals, and easily submerge discrete pest and disease characteristics in the continuous background of abiotic stress. As a result, in orchards with complex canopy structures and extensive shadow distribution, the results of pest and disease assessment mainly depend on light conditions, with poor stability and consistency, making it difficult to accurately reflect the true degree of pest and disease in the orchard.

[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide a method and system for remote sensing monitoring and assessment of pests and diseases in field orchards using unmanned aerial vehicles (UAVs), in order to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A method for remote sensing monitoring and assessment of pests and diseases in field orchards using unmanned aerial vehicles (UAVs), comprising the following steps:

[0009] Step 1: Acquire multispectral remote sensing images of the orchard to be monitored and extract the reflectance curve of each pixel. Based on the differences in spectral response of different bands, extract the reflectance of characteristic bands and construct feature enhancement term, background suppression term and brightness normalization factor. By the ratio of the difference between feature enhancement term and background suppression term to brightness normalization factor, generate a ternary separation index that represents the confidence of each pixel.

[0010] Step 2: Calculate the ternary separation index of all pixels and determine the classification threshold. Based on the relationship between the index and the classification threshold, perform pixel-by-pixel discrimination to classify the pixels into non-canopy background pixels, shadow canopy pixels, and illuminated canopy pixels.

[0011] Step 3: Construct a brightness component representing the degree of brightness based on the reflectivity of the characteristic bands. Traverse all shadow canopy pixels and retrieve the average brightness component of the illuminated canopy pixels in their neighborhood as the local illumination baseline average.

[0012] Step 4: Based on the local illumination baseline mean and the luminance component of the corresponding shadow canopy pixel, construct a gain compensation term to quantify the degree of illumination attenuation, and use the gain compensation term to correct the reflectivity curve of the shadow canopy pixel.

[0013] Step 5: Construct a discrimination term to characterize the degree of vegetation activity based on the corrected reflectance curve. Statistically analyze the reflectance distribution within a local evaluation range of a preset spatial scale, centered on the pixel. Select the reflectance value corresponding to the set quantile as the health reference benchmark. Compare the pixel reflectance with the health reference benchmark to construct an attenuation term, and couple it with the discrimination term to calculate the physiological stress index of each pixel.

[0014] Step 6: Set a judgment threshold based on the health reference benchmark, calculate the physiological stress index below the threshold, perform power-weighted summation and normalization on the difference between the index and the judgment threshold, and output the regional hazard level index that represents the overall degree of pest and disease damage in the evaluation area.

[0015] Furthermore, by acquiring multispectral remote sensing images and extracting the reflectance curve of each pixel, the reflectance of each pixel in the set characteristic band is extracted from the reflectance curve based on the spectral response differences of different bands.

[0016] The defined characteristic bands include near-infrared, red, green, and blue light bands, and the reflectance of each characteristic band is the average reflectance within that characteristic band.

[0017] Furthermore, the steps for generating the ternary separation index for each pixel are as follows:

[0018] For each pixel, a feature enhancement term that amplifies vegetation features is constructed by the square of the reflectance in the near-infrared band, and the product of the reflectance in the red and blue bands is used as a background suppression term.

[0019] The reflectance of the green light band is selected as the visible light brightness reference, and the vector modulus constructed by the reflectance of the red and blue light bands is calculated. The product of the visible light brightness reference and the vector modulus is used as the brightness normalization factor.

[0020] Based on the ratio of the difference between the feature enhancement term and the background suppression term to the brightness normalization factor, a ternary separation index is generated to characterize the vegetation confidence of a pixel.

[0021] Furthermore, the steps for determining the classification threshold are as follows:

[0022] The ternary separation index of all pixels is calculated, and the minimum, median and maximum values ​​are selected as the initial cluster centers for the non-canopy background, the shadow canopy and the illuminated canopy, respectively.

[0023] Calculate the numerical distance between the ternary separation index of each pixel and the three initial cluster centers. Based on the numerical distance, temporarily assign the pixel to the category represented by the nearest initial cluster center. After all pixels have been assigned, calculate the average value of all pixel values ​​in the three temporary categories and update the three average values ​​as the new cluster centers.

[0024] Repeat the pixel classification and cluster center update operation until the difference between the new cluster center and the previous cluster center is less than the preset small error, which means that the numerical changes of the three cluster centers have converged and stabilized.

[0025] The arithmetic mean of the cluster centers representing the non-canopy background and the cluster centers representing the shaded canopy was used as the vegetation / background separation threshold, and the arithmetic mean of the cluster centers representing the shaded canopy and the cluster centers representing the illuminated canopy was used as the shadow / illumination separation threshold.

[0026] Furthermore, the steps for obtaining the luminance component of each pixel are as follows:

[0027] Based on the differences in sensitivity of visible light bands in the spectral response of sensors, the contribution weights of red light, green light and blue light in the near-surface light environment to visual brightness perception are set, and are defined as the corresponding spectral light efficiency weights.

[0028] The reflectance of each pixel in the red, green, and blue light bands is multiplied by the corresponding spectral luminous efficiency weight to calculate the contribution component of each band. The brightness component, which characterizes the brightness of the pixel, is constructed by summing the contribution components of each band.

[0029] Furthermore, the step of obtaining the local illumination reference mean is as follows:

[0030] Centered on the current shadow canopy pixel, a square neighborhood window of fixed size is defined. Each pixel in the square neighborhood window is traversed, and all pixels of the illuminated canopy are selected as valid reference samples.

[0031] The luminance components of the effective reference samples are statistically analyzed, summed, and divided by the total number of pixels in the effective reference samples to obtain the arithmetic mean, which is used as the local illumination benchmark mean characterizing the theoretical luminance of the region in an unobstructed state.

[0032] If no illuminated canopy pixels are found in the current window, the window range is expanded by a preset step size for a second search. If no illuminated canopy pixels are found in the second search range, the average value of the brightness components of all illuminated canopy pixels in the entire multispectral remote sensing image is calculated and used as the local illumination reference average value of the shaded canopy pixel.

[0033] Furthermore, the correction steps for the reflectivity curve are as follows:

[0034] For each shadow canopy pixel, a stabilization factor with a very small value is introduced into the corresponding luminance component and summed. The ratio between the local illumination baseline mean and the summed luminance component is calculated as the original illumination attenuation ratio.

[0035] The original illumination attenuation ratio is numerically nonlinearly compressed and smoothed using the natural logarithm function, and a basic gain modulus with continuous transition characteristics is constructed by superimposing a unit basis.

[0036] A preset adjustment index is selected to perform a power operation on the basic gain modulus, thereby nonlinearly controlling the sensitivity to shadow depth and constructing the final gain compensation term for spectral restoration.

[0037] The reflectance value of each pixel in the shadow canopy is multiplied and corrected using the constructed gain compensation term to obtain the enhanced reflectance curve.

[0038] Furthermore, the steps for constructing the discriminant term are as follows:

[0039] In the reflectance curves of the rectified shadow canopy pixels, the reflectance in the near-infrared band and the reflectance in the red band are extracted.

[0040] A difference term characterizing the spectral steepness of vegetation is constructed by using the difference between the reflectance of the near-infrared band and the reflectance of the red band. The sum of the reflectance of the near-infrared band and the reflectance of the red band is used as the normalized background base. A discrimination term characterizing the vegetation activity of a pixel is constructed based on the ratio of the difference term to the normalized background base.

[0041] The steps for obtaining the physiological stress index of the pixel are as follows:

[0042] Centered on the current pixel and according to a preset spatial scale, a local evaluation range is determined. Within this range, the reflectance distribution of the illuminated canopy pixels and the corrected shadow canopy pixels in the green light band is statistically analyzed. The statistically analyzed reflectance is then sorted from largest to smallest, and the top-ranked pixels are selected. The reflectance at that location is used as a health reference benchmark for the current pixel;

[0043] Calculate the absolute value of the deviation between the reflectance of the current pixel in the green band and the healthy reference baseline, and construct an attenuation term using the ratio of the absolute value of the deviation to the healthy reference baseline;

[0044] The physiological stress index is obtained by multiplying and coupling the discrimination term and the attenuation term. The specific calculation formula is as follows:

[0045]

[0046] In the formula, The physiological stress index for each pixel. , ,and This represents the reflectivity of the current pixel in the near-infrared, red, and green light bands. As a health reference benchmark.

[0047] Furthermore, the steps for obtaining the regional hazard level index include:

[0048] The physiological stress index corresponding to all shadow canopy pixels within the local evaluation range is statistically analyzed. The physiological stress index of all pixels that meet the health reference benchmark is selected and its arithmetic mean is calculated. The product of the preset tolerance coefficient and the arithmetic mean is set as the judgment threshold.

[0049] Physiological stress indices below the judgment threshold are selected and their differences from the judgment threshold are calculated to construct a deviation term that quantifies the degree of deviation of a single point from the health benchmark.

[0050] The deviation term is nonlinearly filtered using a linear rectification function, and the filtered result is used as the effective hazard contribution value. The hazard contribution value is then exponentially calculated using a preset hazard factor to construct a weighted hazard component that nonlinearly amplifies the disaster characteristics.

[0051] The weighted hazard components of all pixels within the local evaluation area are summed, and the summation result is normalized using the total number of pixels to generate a regional hazard level index that characterizes the overall severity of pest and disease damage in the evaluation area. The specific calculation formula is as follows:

[0052]

[0053] The regional damage level index is used to characterize and evaluate the overall severity of pest and disease damage in a region. This represents the total number of pixels in the shadow canopy within the local evaluation range. For pixel index, the value range is: , Within the scope of local evaluation Physiological stress index per pixel To determine the threshold, It is a hazardous factor.

[0054] This invention also provides a drone remote sensing monitoring and assessment system for pests and diseases in field orchards. This system is used to implement the aforementioned drone remote sensing monitoring and assessment method for pests and diseases in field orchards, and includes:

[0055] The ternary index construction module is used to acquire multispectral remote sensing images of the orchard to be monitored and extract the reflectance curve of each pixel. Based on the differences in spectral response of different bands, the reflectance of the characteristic bands is extracted, and feature enhancement term, background suppression term and brightness normalization factor are constructed. The ternary separation index, which represents the confidence of each pixel, is generated by the ratio of the difference between the feature enhancement term and the background suppression term to the brightness normalization factor.

[0056] The pixel classification and discrimination module is used to calculate the ternary separation index of all pixels and determine the classification threshold. Based on the relationship between the index and the classification threshold, it performs pixel-by-pixel discrimination and divides the pixels into non-canopy background pixels, shadow canopy pixels, and illuminated canopy pixels.

[0057] The illumination reference retrieval module is used to construct a brightness component representing the degree of brightness based on the reflectivity of the characteristic band, traverse all shadow canopy pixels, and retrieve the average brightness component of the illuminated canopy pixels in its neighborhood as the local illumination reference average.

[0058] The shadow spectral correction module is used to construct a gain compensation term that quantifies the degree of light attenuation based on the local illumination reference mean and the brightness component of the corresponding shadow canopy pixel, and to use the gain compensation term to correct the reflectance curve of the shadow canopy pixel.

[0059] The stress index calculation module is used to construct a discrimination term characterizing the degree of vegetation activity based on the corrected reflectance curve. It statistically analyzes the reflectance distribution within a local evaluation range of a preset spatial scale with the pixel as the center, selects the reflectance value corresponding to the set quantile as the health reference benchmark, compares the pixel reflectance with the health reference benchmark to construct an attenuation term, and couples it with the discrimination term to calculate the physiological stress index of each pixel.

[0060] The regional hazard assessment module is used to set a judgment threshold based on a health reference benchmark, calculate the physiological stress index below the threshold, perform power-weighted summation and normalization on the difference between the index and the judgment threshold, and output a regional hazard level index that represents the overall degree of pest and disease damage in the evaluation area.

[0061] Compared with the prior art, the beneficial effects of the present invention are:

[0062] This invention constructs a ternary separation index to clearly divide pixels in an image into non-canopy background pixels, illuminated canopy pixels, and shaded canopy pixels. This allows shaded canopy pixels to be identified and processed independently in the processing flow. In the evaluation stage, only shaded canopy pixels are used as the analysis object. Since shaded canopy pixels have relatively consistent illumination constraints in the same image, their reflectance changes mainly come from differences in the physiological state of the vegetation itself, thus naturally reducing the interference of light intensity changes on the judgment of pests and diseases.

[0063] This invention corrects the shaded canopy pixels by introducing a local illumination benchmark and constructs a physiological stress index based on this. It only performs statistics and accumulation on the shaded canopy pixels, avoiding the problem that the enhanced direct light in the illuminated canopy pixels masks the early attenuation characteristics of pests and diseases. Thus, this invention can stably utilize the information source of shaded canopy pixels, which is neglected in the prior art, without relying on complex models, to achieve more consistent, reliable and regionally representative assessment results of the severity of pests and diseases in orchards. Attached Figure Description

[0064] Figure 1 This is a schematic diagram of the overall method flow of the present invention;

[0065] Figure 2 This is a feature map showing the coupling between the discriminant term and the physiological stress index based on shadow spectral correction.

[0066] Figure 3 This is a schematic diagram of the overall system structure of the present invention. Detailed Implementation

[0067] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0068] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0069] Example:

[0070] Please see Figures 1-2 The present invention provides a technical solution:

[0071] A method for remote sensing monitoring and assessment of pests and diseases in field orchards using unmanned aerial vehicles (UAVs), comprising the following steps:

[0072] Step 1: Acquire multispectral remote sensing images of the orchard to be monitored and extract the reflectance curve of each pixel. Based on the differences in spectral response of different bands, extract the reflectance of the characteristic bands and construct a feature enhancement term, a background suppression term, and a brightness normalization factor. By the ratio of the difference between the feature enhancement term and the background suppression term to the brightness normalization factor, generate a ternary separation index that represents the confidence of each pixel.

[0073] Traditional visible light imaging can only record color information and is easily affected by changes in ambient light intensity. In contrast, reflectance curves are inherent physical properties of objects and do not undergo fundamental shifts in properties with changes in lighting conditions. Reflectance data obtained through radiometric calibration can construct an absolute physical quantity independent of the imaging device gain.

[0074] In this embodiment, the reflectance curve of each pixel is extracted from the acquired multispectral remote sensing image. Based on the spectral response differences of different bands, the reflectance of each pixel in the set characteristic band is extracted from the reflectance curve.

[0075] The defined characteristic bands include the near-infrared band. Red light band Green light band and blue light band The reflectance of each characteristic band is the average reflectance within that band, designed to suppress random noise and sensor thermal noise that may exist in single-wavelength data. By using the mean, it is ensured that the data represents the most essential spectral response characteristics of ground objects in that band, improving the tolerance to minor fluctuations in sensor readings.

[0076] In this embodiment, the step of generating the ternary separation index for each pixel is as follows:

[0077] Because the mesophyll cell structure of healthy vegetation has a strong multiple scattering effect on near-infrared light, it has a very high reflectivity in the near-infrared band, while the reflectivity of backgrounds such as soil and dead branches is low in this band. Therefore, for each pixel, a feature enhancement term that amplifies vegetation features is constructed by the square of the reflectivity in the near-infrared band.

[0078] Healthy vegetation exhibits extremely low reflectivity in the red and blue light bands due to the absorption of chlorophyll and carotenoids through photosynthesis. In contrast, bare soil and abiotic backgrounds typically have higher reflectivity in these bands. Therefore, the product of the reflectivity in the red and blue light bands is used as the background suppression term.

[0079] The reflectance of the green light band is selected as the visible light brightness benchmark to characterize the peak reflectance intensity of vegetation in the visible light range. The vector modulus of the absorption band is constructed by calculating the reflectance of the red and blue light bands. The product of the visible light brightness benchmark and the vector modulus is used as the brightness normalization factor, which effectively offsets the overall numerical drift caused by light intensity.

[0080] Based on the ratio of the difference between the feature enhancement term and the background suppression term to the brightness normalization factor, a ternary separation index representing the vegetation confidence of a pixel is generated. The specific calculation formula is as follows:

[0081]

[0082] In the formula, It is the ternary separation index. , , , The reflectance values ​​are for the near-infrared, red, green, and blue light bands, respectively. For feature enhancement terms, For background suppression, The brightness normalization factor. This is a numerical stability constant used to prevent the denominator from being zero; in this embodiment, it is set to... .

[0083] Among them, the feature enhancement term is used to enhance the vegetation signal, the background suppression term can lock and remove the high reflectance features of non-vegetation background to maximize the signal-to-noise ratio, and the brightness normalization factor can effectively offset the numerical drift caused by uneven illumination. Through this triple coupling mechanism of enhancement, suppression and normalization, the generated ternary separation index can generate a high-confidence vegetation probability distribution from the complex orchard background.

[0084] Step 2: Calculate the ternary separation index of all pixels and determine the classification threshold. Based on the relationship between the index and the classification threshold, perform pixel-by-pixel discrimination to classify the pixels into non-canopy background pixels, shadow canopy pixels, and illuminated canopy pixels.

[0085] In this embodiment, the step of determining the classification threshold is as follows:

[0086] In areas with bare soil, concrete roads, or dry weeds, the red edge effect is lacking, and the ternary separation index is the lowest, representing the non-canopy background. The top of a healthy tree canopy exposed to direct sunlight has the strongest reflected signal and the highest ternary separation index, representing the illuminated canopy. Leaves inside the canopy or on the shaded side, although they have vegetation characteristics, have a signal attenuation due to insufficient light, and the ternary separation index is at an intermediate level, representing the shaded canopy.

[0087] Therefore, in this embodiment, the minimum, median and maximum values ​​of the ternary separation index of all pixels are selected as the initial cluster centers of the non-canopy background, shadow canopy and illuminated canopy, respectively. Determining the initial cluster centers through data distribution characteristics can greatly reduce the number of iterations and ensure the stability of the classification results.

[0088] Calculate the numerical distance between the ternary separation index of each pixel and the three initial cluster centers. Based on the numerical distance, temporarily assign the pixel to the category represented by the nearest initial cluster center. After all pixels have been assigned, calculate the average value of all pixel values ​​in the three temporary categories and update the three average values ​​as the new cluster centers.

[0089] The lighting in a causal garden environment is continuously changing, making it difficult to find a perfect segmentation point in a single calculation. Through iterative approximation, the system can automatically find the gravity center of the data distribution and dynamically adapt to the overall brightness and contrast differences in images captured at different times and under different weather conditions. This scheme repeatedly performs pixel classification and cluster center updates until the difference between the new cluster center and the previous cluster center is less than a preset small error, indicating that the numerical changes of the three cluster centers have converged and stabilized.

[0090] The arithmetic mean of the cluster centers representing the non-canopy background and the cluster centers representing the shaded canopy was used as the vegetation / background separation threshold, and the arithmetic mean of the cluster centers representing the shaded canopy and the cluster centers representing the illuminated canopy was used as the shadow / illumination separation threshold.

[0091] When the ternary separation index of a pixel is less than the vegetation / background separation threshold, it is determined that its spectral features lack the red edge effect of vegetation and belong to non-canopy background pixels.

[0092] When the ternary separation index of a pixel is greater than or equal to the vegetation / background separation threshold and less than the shadow / light separation threshold, it is determined that it has vegetation characteristics but the energy intensity is suppressed, and it belongs to the shadow canopy pixel.

[0093] A pixel is considered a high-confidence canopy pixel when its ternary separation index is greater than or equal to the shadow / light separation threshold.

[0094] Among these methods, canopy classification can eliminate interference from soil and weeds, preventing background noise (such as green weeds) from being mistakenly included in pest and disease assessment and causing false warnings; secondly, it can accurately locate the shadowed canopy pixels, laying the foundation for targeted spectral repair and correction in subsequent steps; finally, the separated illuminated canopy pixels, being unobstructed, retain the most authentic reflectivity information and can serve as a physical reference for radiation correction in shadowed areas.

[0095] Step 3: Construct a brightness component representing the degree of brightness based on the reflectivity of the characteristic bands. Traverse all shadow canopy pixels and retrieve the average brightness component of the illuminated canopy pixels in their neighborhood as the local illumination baseline average.

[0096] In this embodiment, the step of obtaining the luminance component of each pixel is as follows:

[0097] Based on the differences in sensitivity of visible light bands in the spectral response of sensors, the contribution weights of red, green and blue light in the near-surface light environment to visual brightness perception are set and defined as the corresponding spectral light efficiency weights.

[0098] Healthy leaves have the highest reflectivity to green light, and the green light band contains the richest canopy structure texture information, so the green light weight is set to 0.59; the blue light band is most affected by Rayleigh scattering in atmospheric transmission and has the highest noise level, so reducing its weight can effectively suppress high-frequency noise interference caused by atmospheric scattering, so the blue light weight is set to 0.11; to ensure that the brightness component is consistent with the original reflectivity in terms of numerical magnitude, the red light weight is set to 0.3.

[0099] The reflectivity of each pixel in the red, green, and blue light bands is multiplied by the corresponding spectral light efficiency weight to calculate the contribution component of each band. By summing the contribution components of each band, a brightness component representing the brightness of the pixel is constructed, thereby preserving texture information and suppressing atmospheric scattering noise.

[0100] In this embodiment, the step of obtaining the local illumination reference mean is as follows:

[0101] Centered on the current shadow canopy pixel, define a square neighborhood window of fixed size, traverse every pixel in the square neighborhood window, and filter out all pixels of the illuminated canopy as valid reference samples.

[0102] Among them, the size of the square neighborhood window is set to The pixel matrix, if the window is too small (e.g. The entire window may fall inside the shadow, making it impossible to find a valid lighting reference point; and when the window is too large (e.g., This will cause the retrieved illumination values ​​to no longer be locally representative.

[0103] The luminance components of the effective reference samples are statistically analyzed, summed, and divided by the total number of pixels in the effective reference samples to obtain the arithmetic mean, which is used as the local illumination benchmark mean characterizing the theoretical luminance of the region under unobstructed conditions.

[0104] If no illuminated canopy pixels are found in the current window, the window range is expanded by a preset step size for a second search. If no illuminated canopy pixels are found in the second search range, the average value of the brightness components of all illuminated canopy pixels in the entire multispectral remote sensing image is calculated and used as the local illumination reference average value of the shaded canopy pixel to improve data accuracy.

[0105] Step 4: Based on the local illumination baseline mean and the luminance component of the corresponding shadow canopy pixel, construct a gain compensation term to quantify the degree of illumination attenuation, and use the gain compensation term to correct the reflectivity curve of the shadow canopy pixel.

[0106] In this embodiment, the correction step for the reflectivity curve is as follows:

[0107] For each shadow canopy pixel, a very small stabilization factor is introduced into the corresponding luminance component and summed. The ratio between the local illumination baseline mean and the summed luminance component is calculated as the original illumination attenuation ratio term for the missing multiple of quantized photon flux. This directly reflects that after the light is blocked, the energy of the pixel is only a fraction of that under normal illumination.

[0108] By using the natural logarithm function to perform numerical nonlinear compression and smoothing on the original illumination attenuation ratio, the noise amplification effect caused by the excessively large ratio in the deep shadow region can be effectively suppressed. Furthermore, by superimposing a unit substrate to construct a basic gain modulus with continuous transition characteristics, the numerical stability of the correction process can be ensured.

[0109] A preset adjustment index is selected to perform a power operation on the basic gain modulus, thereby nonlinearly controlling the sensitivity to shadow depth and constructing the final gain compensation term for spectral restoration.

[0110] Among them, the adjustment index Set to 0.85, when When using full linear compensation, it often leads to an overcorrected bright ring effect at the shadow edges; while using full linear compensation... Fine-tuning to 0.85 creates a sublinear compensation curve that effectively brightens the texture within shadows while maintaining a smooth transition at the boundary between shadow and illuminated areas.

[0111] The constructed gain compensation term is used to perform multiplicative correction on each reflectance value on the reflectance curve of the shadow canopy pixels to obtain the enhanced reflectance curve. The specific formula for calculating the correction of the reflectance curve is as follows:

[0112]

[0113] In the formula, for Band-corrected reflectivity This is a band index, with values ​​ranging from near-infrared, red, green, and blue light bands. for Reflectivity of the band The luminance component of the current pixel. This is the local illumination baseline mean value corresponding to this pixel. To prevent the stability factor from being zero in the denominator, this embodiment sets it to... , This is the original light attenuation ratio term. To adjust the index, the value is set to 0.85.

[0114] Specifically, constructing the original light attenuation ratio term ensures that the correction process does not depend on global parameters, but rather performs precise point-to-point energy compensation for each shaded point of each canopy; subsequently, a logarithmic function is introduced. Numerical compression of high-magnification losses effectively suppresses the noise amplification effect in deep shadow regions. Finally, the gain intensity is softened by adjusting the index to prevent overcorrection at the edges. Based on this correction mechanism, the distorted spectrum obscured by shadows can be restored to the true physiological reflectance curve, significantly improving the spatial coverage of pest and disease monitoring.

[0115] Step 5: Construct a discrimination term to characterize the degree of vegetation activity based on the corrected reflectance curve. Statistically analyze the reflectance distribution within a local evaluation range of a preset spatial scale, centered on the pixel. Select the reflectance value corresponding to the set quantile as the health reference benchmark. Compare the pixel reflectance with the health reference benchmark to construct an attenuation term, and couple it with the discrimination term to calculate the physiological stress index of each pixel.

[0116] In this embodiment, the steps for constructing the discrimination term are as follows:

[0117] Based on the unique red-edge effect mechanism of vegetation: healthy mesophyll cell structure strongly scatters near-infrared light, while chlorophyll strongly absorbs red light. The reflectance of the near-infrared band and the reflectance of the red band are extracted from the reflectance curves of the calibrated shaded canopy pixels.

[0118] A difference term characterizing the spectral steepness of vegetation is constructed using the difference between the reflectance of the near-infrared band and the reflectance of the red band. The sum of the reflectance of the near-infrared band and the reflectance of the red band is used as the normalized background base. A discrimination term characterizing the vegetation activity of a pixel is constructed based on the ratio of the difference term to the normalized background base.

[0119] In this embodiment, the steps for obtaining the physiological stress index are as follows:

[0120] The local evaluation range is determined centered on the current pixel and according to a preset spatial scale. In this embodiment, the preset spatial scale is [missing information]. Pixel.

[0121] Within this range, environmental factors such as soil fertility and irrigation conditions are homogeneous, making adjacent fruit trees the best reference. By utilizing the spatial continuity of abiotic stresses (such as water and nutrient shortages) and the local discrete and patchy characteristics of pest and disease stresses, by limiting the local range and selecting high quantile reflectance as the health benchmark, smooth abiotic background signals can be effectively suppressed, thereby accurately highlighting discrete abnormal characteristics of pests and diseases.

[0122] Because early-stage pests and diseases exhibit more pronounced reflectance anomalies in the visible green band than in the near-infrared band, this embodiment statistically analyzes the reflectance distribution of illuminated canopy pixels and corrected shaded canopy pixels in the green band within this range, and sorts the statistically analyzed reflectance from highest to lowest. To eliminate occasional high-frequency noise caused by high light reflection or sensor defects, the first... The reflectance at that location is used as a health reference benchmark for the current pixel.

[0123] Calculate the absolute value of the deviation between the reflectance of the current pixel's green light band and the healthy reference baseline. Use the ratio of the absolute value of the deviation to the healthy reference baseline and combine it with an exponential function to construct an attenuation term to achieve nonlinear amplification of the disease characteristics.

[0124] The physiological stress index of a pixel is obtained by multiplying and coupling the discrimination term and the attenuation term. The specific calculation formula is as follows:

[0125]

[0126] In the formula, The physiological stress index for each pixel. , ,and This represents the reflectivity of the current pixel in the near-infrared, red, and green light bands. As a health reference benchmark for the current pixel, As a discriminant to characterize the degree of vegetation activity at a pixel, This is the attenuation term for nonlinear amplification of disease characteristics.

[0127] Specifically, the basic biomass level of vegetation is established through a discrimination term, and a nonlinear penalty is applied to the local relative deviation in the green light band using an attenuation term with the best vegetation in the neighborhood as a dynamic reference frame. This operation can keenly detect early latent diseases in leaves when pigmentation has changed before biomass has significantly decreased, thus significantly improving the monitoring system's early warning capability for initial disasters.

[0128] Step 6: Set a judgment threshold based on the health reference benchmark, calculate the physiological stress index below the threshold, perform power-weighted summation and normalization on the difference between the index and the judgment threshold, and output the regional hazard level index that represents the overall degree of pest and disease damage in the evaluation area.

[0129] In this embodiment, the step of obtaining the regional hazard level index includes:

[0130] The physiological stress index corresponding to all shadow canopy pixels within the local evaluation range is statistically analyzed. The physiological stress index of all pixels that meet the health reference benchmark is selected and its arithmetic mean is calculated. The product of the preset tolerance coefficient and the arithmetic mean is set as the judgment threshold, which defines the boundary of normal growth fluctuations and actively ignores the slight spectral decrease caused by leaf angle and differences between new and old leaves.

[0131] Even healthy vegetation in nature exhibits slight fluctuations in its physiological indicators, which are normally distributed. Therefore, in this embodiment, the tolerance coefficient is set to 0.9, excluding pixels whose physiological stress index is slightly below the mean but still within the normal fluctuation range from the disaster assessment, thereby greatly reducing the false alarm rate caused by natural growth differences.

[0132] Physiological stress indices below the judgment threshold are selected and their differences from the judgment threshold are calculated to construct a deviation term that quantifies the degree of deviation of a single point from the health benchmark.

[0133] A linear rectification function is used to nonlinearly filter the deviation terms, ensuring that only when the physiological stress index is below the judgment threshold are the filtered results considered as valid hazard contribution values, thus preventing healthy pixels from lowering the overall hazard index. A power operation is then performed on the hazard contribution values ​​using preset hazard factors to construct a nonlinearly amplified weighted hazard component that amplifies the disaster-affected characteristics.

[0134] The weighted hazard components of all pixels within the local evaluation area are summed, and the summation result is normalized using the total number of pixels to generate a regional hazard level index that characterizes the overall severity of pest and disease damage in the evaluation area. The specific calculation formula is as follows:

[0135]

[0136] The regional damage level index is used to characterize and evaluate the overall severity of pest and disease damage in a region. This represents the total number of pixels in the shadow canopy within the local evaluation range. For pixel index, the value range is: , Within the scope of local evaluation Physiological stress index per pixel To determine the threshold, It is a hazardous factor.

[0137] Among them, the judgment threshold and the linear rectification function are used. A one-way filtering mechanism is constructed to accurately shield the dilution effect of healthy pixels on disaster assessment; the difference between the physiological stress index and the threshold is used to quantify the disaster depth of a single point; then, hazard factors are introduced to nonlinearly amplify the severe stress signal, highlighting the weight of the disease outbreak center; finally, the regional hazard level index is output through mean normalization, realizing the transformation of discrete pixel-level microscopic pathological features into macroscopic decision indicators that objectively reflect the overall risk level of the region.

[0138] Table 1 is an example table of physiological stress characteristic parameters of the canopy in field orchards.

[0139] Table 1: Characteristic Parameters of Canopy Physiological Stress in Field Orchards

[0140]

[0141] Table 1 shows the ability of this method to distinguish different canopy physiological states in complex orchard scenarios. Non-canopy background pixels have extremely low discriminant terms and physiological stress indices (close to 0), reflecting a lack of vegetation spectral characteristics. Healthy canopy pixels (regardless of light or shadow) exhibit high discriminant terms (above 0.75) and high physiological stress indices (0.63–0.89), indicating strong vegetation activity and weak attenuation. Canopy pixels under disease stress have higher discriminant terms, but significantly lower physiological stress indices (0.39–0.62), and even if the discriminant terms remain high, they are effectively suppressed by the increased attenuation term. This table demonstrates that single spectral discrimination can easily confuse mild stress with a healthy state. By coupling discriminant terms and attenuation terms and using a two-layer classification mechanism, it can accurately capture the subtle physiological decline caused by early pests and diseases.

[0142] Please see Figure 3 The present invention also provides a UAV remote sensing monitoring and assessment system for pests and diseases in field orchards. This system is used to implement the aforementioned UAV remote sensing monitoring and assessment method for pests and diseases in field orchards, and includes:

[0143] The ternary index construction module is used to acquire multispectral remote sensing images of the orchard to be monitored and extract the reflectance curve of each pixel. Based on the differences in spectral response of different bands, the reflectance of the characteristic bands is extracted, and feature enhancement term, background suppression term and brightness normalization factor are constructed. The ternary separation index, which represents the confidence of each pixel, is generated by the ratio of the difference between the feature enhancement term and the background suppression term to the brightness normalization factor.

[0144] The pixel classification and discrimination module is used to calculate the ternary separation index of all pixels and determine the classification threshold. Based on the relationship between the index and the classification threshold, it performs pixel-by-pixel discrimination and divides the pixels into non-canopy background pixels, shadow canopy pixels, and illuminated canopy pixels.

[0145] The illumination reference retrieval module is used to construct a brightness component representing the degree of brightness based on the reflectivity of the characteristic band, traverse all shadow canopy pixels, and retrieve the average brightness component of the illuminated canopy pixels in its neighborhood as the local illumination reference average.

[0146] The shadow spectral correction module is used to construct a gain compensation term that quantifies the degree of light attenuation based on the local illumination reference mean and the brightness component of the corresponding shadow canopy pixel, and to use the gain compensation term to correct the reflectance curve of the shadow canopy pixel.

[0147] The stress index calculation module is used to construct a discrimination term characterizing the degree of vegetation activity based on the corrected reflectance curve. It statistically analyzes the reflectance distribution within a local evaluation range of a preset spatial scale with the pixel as the center, selects the reflectance value corresponding to the set quantile as the health reference benchmark, compares the pixel reflectance with the health reference benchmark to construct an attenuation term, and couples it with the discrimination term to calculate the physiological stress index of each pixel.

[0148] The regional hazard assessment module is used to set a judgment threshold based on a health reference benchmark, calculate the physiological stress index below the threshold, perform power-weighted summation and normalization on the difference between the index and the judgment threshold, and output a regional hazard level index that represents the overall degree of pest and disease damage in the evaluation area.

[0149] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0150] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0151] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0152] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for unmanned aerial vehicle (UAV) remote sensing monitoring and assessment of pests and diseases in field orchards, characterized in that, The specific steps include: Step 1: Acquire multispectral remote sensing images of the orchard to be monitored and extract the reflectance curve of each pixel. Based on the differences in spectral response of different bands, extract the reflectance of characteristic bands and construct feature enhancement term, background suppression term and brightness normalization factor. By the ratio of the difference between feature enhancement term and background suppression term to brightness normalization factor, generate a ternary separation index that represents the confidence of each pixel. Step 2: Calculate the ternary separation index of all pixels and determine the classification threshold. Based on the relationship between the index and the classification threshold, perform pixel-by-pixel discrimination to classify the pixels into non-canopy background pixels, shadow canopy pixels, and illuminated canopy pixels. Step 3: Construct a brightness component representing the degree of brightness based on the reflectivity of the characteristic bands. Traverse all shadow canopy pixels and retrieve the average brightness component of the illuminated canopy pixels in their neighborhood as the local illumination baseline average. Step 4: Based on the local illumination baseline mean and the luminance component of the corresponding shadow canopy pixel, construct a gain compensation term to quantify the degree of illumination attenuation, and use it to correct the reflectivity curve of the shadow canopy pixel. Step 5: Construct a discrimination term to characterize the degree of vegetation activity based on the corrected reflectance curve. Statistically analyze the reflectance distribution within a local evaluation range of a preset spatial scale, centered on the pixel. Select the reflectance value corresponding to the set quantile as the health reference benchmark. Compare the pixel reflectance with the health reference benchmark to construct an attenuation term, and couple it with the discrimination term to calculate the physiological stress index of each pixel. Step 6: Set a judgment threshold based on the health reference benchmark, calculate the physiological stress index below the threshold, perform power-weighted summation and normalization on the difference between the index and the judgment threshold, and output the regional hazard level index that represents the overall degree of pest and disease damage in the evaluation area. By acquiring multispectral remote sensing images and extracting the reflectance curve of each pixel, the reflectance of each pixel in the set characteristic band is extracted from the reflectance curve based on the spectral response differences of different bands. The defined characteristic bands include near-infrared, red, green and blue light bands, and the reflectance of each characteristic band is the average reflectance within that characteristic band. The steps for generating the ternary separation index for each pixel are as follows: For each pixel, a feature enhancement term that amplifies vegetation features is constructed by the square of the reflectance in the near-infrared band, and the product of the reflectance in the red and blue bands is used as a background suppression term. The reflectance of the green light band is selected as the visible light brightness reference, and the vector modulus constructed by the reflectance of the red and blue light bands is calculated. The product of the visible light brightness reference and the vector modulus is used as the brightness normalization factor. Based on the ratio of the difference between the feature enhancement term and the background suppression term to the brightness normalization factor, a ternary separation index is generated to characterize the vegetation confidence of a pixel. The correction steps for the reflectivity curve are as follows: For each shadow canopy pixel, a stabilization factor with a very small value is introduced into the corresponding luminance component and summed. The ratio between the local illumination baseline mean and the summed luminance component is calculated as the original illumination attenuation ratio. The original illumination attenuation ratio is numerically nonlinearly compressed and smoothed using the natural logarithm function, and a basic gain modulus with continuous transition characteristics is constructed by superimposing a unit basis. A preset adjustment index is selected to perform a power operation on the basic gain modulus, thereby nonlinearly controlling the sensitivity to shadow depth and constructing the final gain compensation term for spectral restoration. The reflectance value of each pixel in the shadow canopy is multiplied and corrected using the constructed gain compensation term to obtain the enhanced reflectance curve. The steps for constructing the discriminant are as follows: In the reflectance curves of the rectified shadow canopy pixels, the reflectance in the near-infrared band and the reflectance in the red band are extracted. A difference term characterizing the spectral steepness of vegetation is constructed by using the difference between the reflectance of the near-infrared band and the reflectance of the red band. The sum of the reflectance of the near-infrared band and the reflectance of the red band is used as the normalized background base. A discrimination term characterizing the vegetation activity of a pixel is constructed based on the ratio of the difference term to the normalized background base. The steps for obtaining the physiological stress index of the pixel are as follows: Centered on the current pixel and according to a preset spatial scale, a local evaluation range is determined. Within this range, the reflectance distribution of the illuminated canopy pixels and the corrected shadow canopy pixels in the green light band is statistically analyzed. The statistically analyzed reflectance is then sorted from largest to smallest, and the top-ranked pixels are selected. The reflectance at that location is used as a health reference benchmark for the current pixel; Calculate the absolute value of the deviation between the reflectance of the current pixel in the green band and the healthy reference baseline, and construct an attenuation term using the ratio of the absolute value of the deviation to the healthy reference baseline; The physiological stress index is obtained by multiplying and coupling the discrimination term and the attenuation term. The specific calculation formula is as follows: In the formula, The physiological stress index for each pixel. , ,and This represents the reflectivity of the current pixel in the near-infrared, red, and green light bands. As a health reference benchmark.

2. The method for UAV remote sensing monitoring and assessment of pests and diseases in field orchards according to claim 1, characterized in that: The steps to determine the classification threshold are as follows: The ternary separation index of all pixels is calculated, and the minimum, median and maximum values ​​are selected as the initial cluster centers for the non-canopy background, the shadow canopy and the illuminated canopy, respectively. Calculate the numerical distance between the ternary separation index of each pixel and the three initial cluster centers. Based on the numerical distance, temporarily assign the pixel to the category represented by the nearest initial cluster center. After all pixels have been assigned, calculate the average value of all pixel values ​​in the three temporary categories and update the three average values ​​as the new cluster centers. Repeat the pixel classification and cluster center update operation until the difference between the new cluster center and the previous cluster center is less than the preset small error, which means that the numerical changes of the three cluster centers have converged and stabilized. The arithmetic mean of the cluster centers representing the non-canopy background and the cluster centers representing the shaded canopy was used as the vegetation / background separation threshold, and the arithmetic mean of the cluster centers representing the shaded canopy and the cluster centers representing the illuminated canopy was used as the shadow / illumination separation threshold.

3. The method for unmanned aerial vehicle (UAV) remote sensing monitoring and assessment of pests and diseases in field orchards according to claim 2, characterized in that: The steps to obtain the luminance component of each pixel are as follows: Based on the differences in sensitivity of visible light bands in the spectral response of sensors, the contribution weights of red light, green light and blue light in the near-surface light environment to visual brightness perception are set, and are defined as the corresponding spectral light efficiency weights. The reflectance of each pixel in the red, green, and blue light bands is multiplied by the corresponding spectral luminous efficiency weight to calculate the contribution component of each band. The brightness component, which characterizes the brightness of the pixel, is constructed by summing the contribution components of each band.

4. The method for UAV remote sensing monitoring and assessment of pests and diseases in field orchards according to claim 3, characterized in that: The steps for obtaining the local illumination baseline mean are as follows: Centered on the current shadow canopy pixel, a square neighborhood window of fixed size is defined. Each pixel in the square neighborhood window is traversed, and all pixels of the illuminated canopy are selected as valid reference samples. The luminance components of the effective reference samples are statistically analyzed, summed, and divided by the total number of pixels in the effective reference samples to obtain the arithmetic mean, which is used as the local illumination benchmark mean characterizing the theoretical luminance of the region in an unobstructed state. If no illuminated canopy pixels are found in the current window, the window range is expanded by a preset step size for a second search. If no illuminated canopy pixels are found in the second search range, the average value of the brightness components of all illuminated canopy pixels in the entire multispectral remote sensing image is calculated and used as the local illumination reference average value of the shaded canopy pixel.

5. The method for UAV remote sensing monitoring and assessment of pests and diseases in field orchards according to claim 4, characterized in that: The steps for obtaining the regional hazard level index include: The physiological stress index corresponding to all shadow canopy pixels within the local evaluation range is statistically analyzed. The physiological stress index of all pixels that meet the health reference benchmark is selected and its arithmetic mean is calculated. The product of the preset tolerance coefficient and the arithmetic mean is set as the judgment threshold. Physiological stress indices below the judgment threshold are selected and their differences from the judgment threshold are calculated to construct a deviation term that quantifies the degree of deviation of a single point from the health benchmark. The deviation term is nonlinearly filtered using a linear rectification function, and the filtered result is used as the effective hazard contribution value. The hazard contribution value is then exponentially calculated using a preset hazard factor to construct a weighted hazard component that nonlinearly amplifies the disaster characteristics. The weighted hazard components of all pixels within the local evaluation area are summed, and the summation result is normalized using the total number of pixels to generate a regional hazard level index that characterizes the overall severity of pest and disease damage in the evaluation area. The specific calculation formula is as follows: The regional damage level index is used to characterize and evaluate the overall severity of pest and disease damage in a region. This represents the total number of pixels in the shadow canopy within the local evaluation range. For pixel index, the value range is: , Within the scope of local evaluation Physiological stress index per pixel To determine the threshold, It is a hazardous factor.

6. A drone-based remote sensing monitoring and assessment system for pests and diseases in field orchards, characterized in that: The aforementioned UAV remote sensing monitoring and assessment system for pests and diseases in field orchards is used to implement the UAV remote sensing monitoring and assessment method for pests and diseases in field orchards as described in any one of claims 1-5, comprising: The ternary index construction module is used to acquire multispectral remote sensing images of the orchard to be monitored and extract the reflectance curve of each pixel. Based on the differences in spectral response of different bands, the reflectance of the characteristic bands is extracted, and feature enhancement term, background suppression term and brightness normalization factor are constructed. The ternary separation index, which represents the confidence of each pixel, is generated by the ratio of the difference between the feature enhancement term and the background suppression term to the brightness normalization factor. The pixel classification and discrimination module is used to calculate the ternary separation index of all pixels and determine the classification threshold. Based on the relationship between the index and the classification threshold, it performs pixel-by-pixel discrimination and divides the pixels into non-canopy background pixels, shadow canopy pixels, and illuminated canopy pixels. The illumination reference retrieval module is used to construct a brightness component representing the degree of brightness based on the reflectivity of the characteristic band, traverse all shadow canopy pixels, and retrieve the average brightness component of the illuminated canopy pixels in its neighborhood as the local illumination reference average. The shadow spectral correction module is used to construct a gain compensation term that quantifies the degree of light attenuation based on the local illumination reference mean and the brightness component of the corresponding shadow canopy pixel, and to use the gain compensation term to correct the reflectance curve of the shadow canopy pixel. The stress index calculation module is used to construct a discrimination term characterizing the degree of vegetation activity based on the corrected reflectance curve. It statistically analyzes the reflectance distribution within a local evaluation range of a preset spatial scale with the pixel as the center, selects the reflectance value corresponding to the set quantile as the health reference benchmark, compares the pixel reflectance with the health reference benchmark to construct an attenuation term, and couples it with the discrimination term to calculate the physiological stress index of each pixel. The regional hazard assessment module is used to set a judgment threshold based on a health reference benchmark, calculate the physiological stress index below the threshold, perform power-weighted summation and normalization on the difference between the index and the judgment threshold, and output a regional hazard level index that represents the overall degree of pest and disease damage in the evaluation area.

Images