A complex bacterial community effect identification method for blade disease spot expansion inflection point detection

Abstract

The present application belongs to the field of machine vision and data analysis, and proposes a complex bacterial community effect identification method for detection of leaf disease spot expansion inflection point, specifically: first, after applying the complex bacterial community, daily image acquisition is performed to obtain leaf images, the leaf images are preprocessed to obtain comparison images, then the comparison images are converted into disease feature vectors, including disease spot expansion boundary intensity and chlorosis index, finally, through inflection point analysis of the disease feature vectors, an inflection point factor is obtained, and whether the complex bacterial community is effective is dynamically judged according to the inflection point factor. Quantitative interpretation of the occurrence trend of the disease spot from slow rise to short-term transition expansion provides a unified reference index for subsequent screening and sequencing of the complex bacterial community scheme, inference of the supplement window and statistical comparison of the difference in the effective time node, improves the timeliness and accuracy of the identification of the complex bacterial community effect state, and provides efficient and accurate theoretical support for the screening and effect mechanism research of the complex bacterial community for leaf disease spots.

Description

Technical Field

[0001] This invention belongs to the fields of machine vision and data analysis technology, specifically relating to a method for identifying the effectiveness of a complex microbial community in detecting the inflection point of leaf lesion expansion. Background Technology

[0002] In the practice of green control of Amomum leaf spot disease, practitioners generally use compound microbial communities such as antagonistic Bacillus to spray or drench the roots in order to inhibit pathogen infection and promote growth. However, in actual operation, whether the compound microbial community is effective usually depends on visual field inspection or the results of a small number of samples. In the same plot or the same batch of potted plants, a common phenomenon is that some leaves show a sudden acceleration in the expansion of lesions within a few days, with water-soaked boundaries or brown rings rapidly expanding outwards, while other leaves show a significant slowdown or even stagnation in the expansion of lesions. Moreover, the above differences often occur under the same application scheme, making it impossible for the applicator to determine in the early stage whether the microbial community has taken effect and entered the inhibition stage, or whether the concentration is still insufficient and about to enter the stage of uncontrolled expansion. This leads to problems such as missing the opportunity for reapplication and ineffective intervention due to late application. The cause of leaf spot disease is that it is caused by pathogenic fungi such as Fusarium, and the disease course has a typical non-linear chain of stages. After germination and attachment under suitable temperature and humidity conditions and a water film on the leaf surface, spores penetrate the cuticle, invade through stomata, and enter the leaf tissue through cell wall degradation enzymes and toxins. After an incubation period, they form a hyphal network and necrotic patches. During the expansion stage, the outer edge of the lesion is the growth front with the strongest pathogen activity, and the expansion rate often shows an accelerated characteristic, driven by the leaf microenvironment and the host's nutritional status. The effect of the complex Bacillus flora also has a lag and staged nature. Its effectiveness depends on colonization in the foliar and rhizosphere, accumulation of antagonistic metabolites, and induction and amplification of the plant's defense response. Therefore, its effectiveness often corresponds to the inflection point where the expansion of the lesion changes from accelerated to decelerated or stagnant, rather than a simple change in the absolute value of the lesion area. The shortcomings of commonly used techniques such as manual disease grading, disease index statistics, or single-image recognition are that they only depict the area or color characteristics of lesions at a certain moment, failing to depict dynamic information such as the curvature and boundary activity of lesion expansion. Furthermore, they cannot reliably identify expansion inflection points from continuous images in the early, low-cost manner to determine whether the microbial community has taken effect and guide the reapplication window. In addition, relying on application strategies with fixed days or fixed frequencies ignores the pathogen incubation period and the lag in microbial community effectiveness, which drift with environmental fluctuations, resulting in large dispersion of control efficacy, delayed decision-making, and uncontrollable investment. Therefore, a composite microbial community effectiveness identification method based on leaf lesion expansion inflection point detection is urgently needed. Summary of the Invention

[0003] The purpose of this invention is to propose a method for identifying the effectiveness of a complex microbial community in detecting the inflection point of leaf lesion expansion, so as to solve one or more technical problems existing in the prior art, and at least provide a beneficial option or create conditions.

[0004] To achieve the above objectives, according to one aspect of the present invention, a method for identifying the effectiveness of a complex microbial community in detecting the inflection point of leaf lesion expansion is provided, the method comprising the following steps: S100, after the application of compound microbial community, image acquisition is carried out daily to obtain leaf images; S200, preprocess the blade image to obtain the comparison image; S300 converts the compared images into disease feature vectors, including the intensity of lesion expansion boundaries and the chlorosis index; S400 obtains inflection point factors through inflection point analysis of disease feature vectors; S500 dynamically determines whether the complex microbial community is effective based on inflection point factors.

[0005] Furthermore, in step S100, the method for obtaining leaf images by daily image acquisition after applying the compound microbial community is as follows: after applying the compound microbial community, at least one image acquisition is performed on the same leaf or the same leaf position area of ​​the target plant on a daily sampling cycle; a standardized parameter set for image acquisition is preset, which includes at least the shooting distance, shooting angle, imaging resolution, and shooting time period.

[0006] The goal of this step is to make leaf images acquired on different dates comparable and to avoid drift in lesion area and color indicators caused by changes in natural light, shooting angle and scale, which would affect the stability and repeatability of subsequent lesion expansion inflection point detection. After applying the compound microbial community, a standardized image acquisition process is performed on the target plant to form a consistent leaf image sequence across days.

[0007] Further, in step S200, the method for preprocessing the leaf image to obtain the comparison image is as follows: performing denoising and brightness equalization processing on the comparison image; detecting reference markers in the leaf image and obtaining the color reference value or grayscale reference value of the reference markers; performing color correction and brightness normalization on the leaf image based on the color reference value or grayscale reference value to obtain a corrected image; locating and segmenting the leaf region on the corrected image to obtain a leaf mask; setting or removing non-leaf region pixels to zero based on the leaf mask to form a leaf foreground image; and performing scale alignment on the leaf foreground image to obtain the comparison image.

[0008] Because daily image acquisition is often affected by ambient light intensity, the original images contain salt-and-pepper noise, compression noise, and local shadows, causing broken or false boundary phenomena at the edges of lesions. These factors all affect the stable calculation of features such as lesion boundary intensity. Therefore, this step performs denoising and brightness equalization processing on the leaf images. Denoising is achieved using median filtering, Gaussian filtering, or bilateral filtering. Brightness equalization employs histogram equalization, local contrast enhancement, or shadow compensation based on the lighting model. The aim is to reduce brightness gradient variations caused by local shading, ensuring a consistent overall brightness distribution for the same leaf on different days.

[0009] Routine monitoring often involves white balance shifts caused by changes in color temperature between morning and evening and differences in weather conditions. Therefore, all reference markers within the frame need to be corrected for color and brightness. Reference markers are detected in leaf images, and their color or grayscale reference values ​​are obtained. Based on these values, the leaf images are color-corrected and brightness normalized to obtain a corrected image. 1) When the reference marker is a grayscale marker, the measured grayscale value of the grayscale block in the image is mapped to its preset grayscale reference. A brightness scaling factor is calculated, and the entire image is normalized for brightness. 2) When the reference marker is a preset color block marker, the measured color value of the color block is mapped to the preset color reference. White balance and color cast are corrected through channel gain adjustment or a color transformation matrix. 3) When the reference marker is a scale containing the preset color reference, a scale reference is simultaneously provided for subsequent scale alignment. Reference marker correction distinguishes between yellowing and greening of leaves caused by light variations and chlorosis and browning caused by diseases. This allows changes in color-related indicators such as the chlorosis index over time to better reflect the true disease progression rather than the shooting conditions.

[0010] Although the shooting distance is constrained, slight distance fluctuations or compositional shifts may still exist, causing the pixel scale of leaves on different dates to be inconsistent in the images. Directly comparing geometric quantities such as lesion area across days will be affected by scale drift, leading to false inflection points or inflection point drift in further inflection point detection. Therefore, this step performs scale alignment on the leaf foreground image. The scale alignment method includes: 1) pixel scale conversion based on the scale benchmark provided by the reference marker; 2) scaling and normalization based on geometric features such as leaf length and midrib length of the leaf foreground image.

[0011] Further, in step S300, the method for converting the compared image into a disease feature vector, including the lesion expansion boundary strength and chlorosis index, is as follows: lesion mask is obtained by identifying the lesion area in the compared image; the lesion boundary is calculated based on the lesion mask to obtain the lesion boundary length and lesion area; the lesion expansion boundary strength is constructed based on the lesion boundary length and lesion area, which is used to characterize the degree of active expansion of the outer edge of the lesion; the green percentage or green intensity is calculated under the constraint of the leaf mask, and the difference is made with the preset health benchmark or the same leaf historical benchmark to obtain the chlorosis index, which is used to characterize the degree of decay of leaf browning; the lesion expansion boundary strength and the chlorosis index are combined to form a disease feature vector.

[0012] Furthermore, morphological opening and closing operations, hole filling, and small connected region removal are performed on the lesion mask to eliminate noise points and improve the coherence of the lesion area, thereby making the subsequent lesion boundary extraction more stable.

[0013] The lesion boundary is obtained using any one of the following methods: contour extraction operator, edge tracking, or mask morphological gradient. The lesion boundary length is taken as the pixel length of the lesion contour, and the lesion area is the number of pixels within the lesion mask. By default, the method for constructing the lesion expansion boundary strength is to use the ratio of the lesion boundary length to the lesion area as the lesion expansion boundary strength.

[0014] The intensity of lesion expansion boundary is used to characterize the activity level of lesion outer edge expansion. When lesions are in a rapid expansion phase, the proportion of newly formed expansion boundaries at the lesion's outer edge is high, the boundary morphology tends to be complex, and the boundary length per unit area is larger, leading to an increase in this index value. Conversely, when lesion expansion is suppressed and tends to stabilize, the boundary extrapolation speed shows a decreasing trend, the proportion of active boundary areas decreases, causing this index value to decrease or remain at a low level. Therefore, compared to using only lesion area, lesion expansion boundary intensity can more sensitively reflect changes in edge activity during lesion expansion and is suitable for subsequent time series inflection point detection.

[0015] In addition to the geometric expansion of lesions, leaf spot disease is often accompanied by color changes such as browning and necrosis of leaves. To characterize this type of change, this method calculates the green percentage or the mean green intensity under leaf mask constraints and constructs a chlorosis index by difference with a preset healthy baseline or a historical baseline of the same leaf. The green percentage represents the proportion of pixels in the leaf area that meet the green pixel judgment criteria, which is a value higher than the preset threshold of the green channel. The mean green intensity represents the mean of the green channel in the leaf area, or the mean of the green-related components in color spaces such as HSV / Lab. Either the green percentage or the mean green intensity can be selected. The preset healthy baseline can be derived from a baseline sample of healthy leaves or the initial baseline frame before fungicide application. The historical baseline of the same leaf can be the moving average of the previous day or several days. The chlorosis index focuses on relative changes by differencing it with the baseline. It should be noted that although this index is called the chlorosis index, it can be used to characterize the degree of decay from green to yellowing, browning, and even necrosis. In some implementations, the proportion of dark brown pixels can be combined and weighted to make it more sensitive to browning and necrosis, but essentially, they all belong to the quantification of the leaf color degradation trend.

[0016] Furthermore, in step S400, the method for obtaining the inflection point factor through inflection point analysis using the disease feature vector is as follows: Using a natural day as a test, the first few tests of any test are defined as a disease retrospective window, where the disease retrospective window is a sequence of tests as objects, and the disease retrospective window includes any test. The value range of the several tests is NUM=[10,20]. The spatial restraint component DGC is obtained by performing a second-order difference on the intensity of the lesion expansion boundary within the disease retrospective window, and the first-order difference of all chlorosis indices is used as the physiological restraint component GTC. If the product of the spatial restraint component and the physiological restraint component corresponding to a test is negative, the test is defined as a trend deviation point, and the ratio of the lesion expansion boundary intensity and the chlorosis index at the trend deviation point is calculated as the penalty coefficient LY. Trend divergence points are anomalous markers proposed for intervention in the time-varying biological system of leaf spot disease complex microbial communities. Because the apparent progression of leaf diseases is simultaneously governed by two pathways: the pathogen's expansion at the lesion's outer edge and the spread of physiological degradation within the host tissue, these two pathways often exhibit phase differences and decoupling after the intervention of the complex microbial community. Therefore, the second-order difference of the lesion expansion boundary intensity is used to characterize the change in the net expansion acceleration of the pathogen at the lesion edge. This expansion is driven by the pathogen hyphae's invasive ability at the leaf tissue interface, the release of cell wall degrading enzymes / toxins, and the positive feedback formed by the leaf surface micro-water film and nutrient leakage. When the complex microbial community gradually colonizes and produces antagonistic metabolites or triggers induced resistance, it often first acts on the edge expansion front, manifesting as a decrease in the acceleration of boundary activity or even a shift from expansion to inhibition. Meanwhile, the first difference of the chlorophyll loss index mainly reflects the net rate of change of physiological degradation such as destruction of chloroplast structure in mesophyll cells, degradation of chlorophyll, peroxidation of membrane lipids, and necrosis and browning. In addition to being affected by local lesion infection, this degradation is also significantly affected by pathogen toxins, vascular bundle water and metabolic consumption caused by excessive host defense response. Therefore, its spread has a stronger intra-tissue diffusion property and often lags behind or is independent of edge expansion.

[0017] Based on this natural law, when the DGC and GTC signs are opposite in a certain measurement, it suggests that the kinetics of peripheral expansion and the rate of tissue degradation are inconsistent within the same time window. For example, when boundary expansion begins to be contained, the spatial level tends to stabilize, but browning is still accelerating, and the physiological level continues to deteriorate, corresponding to a typical metastable state of controlled surface stress with sustained internal stress. Conversely, when boundary expansion is still accelerating while chlorosis and degradation slow down briefly, it may correspond to a latent transition state from enhanced invasion front to before tissue degradation has manifested. Such divergent states are of crucial significance in actual biocontrol screening. They often appear in the window period when the system is about to undergo a critical transformation and are rare but high-value state signals that are difficult to capture by conventional single-indicator monotonic trends. This method defines such measurements as trend divergence points and further calculates the ratio of lesion expansion boundary intensity to chlorosis index at the divergence point as a penalty coefficient. This coefficient is used to quantify the mismatch between spatial expansion pressure and physiological degradation load, thereby suppressing false improvement judgments caused by temporary slowing of the boundary but continuous internal deterioration. This enhances the sensitivity and discriminative power of time series analysis to complex states such as metastable states, stress reversals, and critical fluctuations, and further improves the early warning and explanatory power of complex microbial community effectiveness identification and screening ranking for early failure risks.

[0018] Within the disease retrospective window, the spatial confidence weight THJ and physiological confidence weight TKJ for each measurement are calculated using a variance-based confidence weighting method. The spatial and physiological containment components are then weighted and summed to obtain the containment synergy degree FJP. The variance-based confidence weighting method involves calculating the variances of the lesion expansion boundary strength and chlorosis index for all lesion expansions within the disease retrospective window. The variances of the lesion expansion boundary strength and chlorosis index are then calculated, and 1 is added to each variance, with the reciprocal taken to obtain the spatial confidence weight THJ and physiological confidence weight TKJ. The specific formula for the weighted summation is: FJP = THJ × DGC + TKJ × GTC. If the measurement is a trend divergence point, a penalty coefficient LY is applied to correct the containment synergy degree. The specific correction formula is: FJP = (THJ × DGC + TKJ × GTC) × LY. The inhibition synergy index incorporates the second-order difference of the lesion expansion boundary intensity and the first-order difference of the chlorosis index into a variance-based confidence-weighted fusion to form a composite index characterizing the dynamics of disease inhibition. Unlike existing techniques that assign fixed weights or linear summation to single features, this method uses the degree of variation of each feature as a confidence signal, dynamically and adaptively adjusting the contributions of the two types of inhibition components at different time periods. This gives greater weight to features with smaller fluctuations and higher reliability, thereby improving the accuracy of characterizing the true inhibition trend. This mechanism can suppress non-physical contributions introduced by abnormal fluctuations, acquisition drift, or short-term spurious changes, preventing unreasonable results from entering the decision-making process and improving the stability and professional reliability of subsequent effectiveness identification and inflection point detection.

[0019] The average value of the containment synergy of all trend divergence points within the disease retrospective window corresponding to a test is taken as the net inhibition increment. If there is no trend divergence point, the net inhibition increment is set as the median of the containment synergy of each disease retrospective window. The difference between the containment synergy of the test and the net inhibition increment is normalized by sigmoid to obtain the inflection point factor IFP. The process of constructing a disease retrospective window and quantifying the degree of containment synergy using a variance-based confidence weighting method is based on fitting the temporal coupled evolution of lesion expansion boundary intensity and chlorosis index using disease feature vectors from consecutive historical measurements. Therefore, it can effectively quantify the disease inflection point and its accompanying staged inhibition risk caused by the combined effect of spatial and physiological containment components, under the premise that lesion spatial expansion intensifies step by step while physiological degradation changes synchronously. However, this process is highly dependent on difference features and window variance estimation, and is prone to numerical distortion in short-period fluctuations or when observation noise is high, especially when disease development enters a period of oscillation, the stability of weights and synergy decreases. Therefore, to further improve the stability in oscillating periods and trend deviation scenarios, and to provide more robust inflection point prediction conclusions, this method proposes a better scheme as follows: Preferably, in step S400, the method for obtaining the inflection point factor through inflection point analysis using the disease feature vector is as follows: Using a natural day as a test, the first few tests of any test are defined as a disease retrospective window, where the disease retrospective window is a sequence of tests as objects, and the disease retrospective window includes any test. The value range of the several tests is NUM=[10,20]. Within the time window of the lesion corresponding to any measurement, the mean vector of all disease feature vectors is calculated as the equilibrium centroid. The Euclidean distance between the disease feature vector of each measurement and the equilibrium centroid is calculated and defined as the radial offset. The Euclidean distance between the disease feature vector of each measurement and the disease feature vector of the previous measurement is calculated and defined as the evolution step distance. The ratio of the radial offset to the evolution step distance of each measurement is used as the radial step deviation rate KTT. The radial step deviation rate combines the absolute offset of disease characteristic changes with the rate of adjacent dynamic changes to form a sensitive quantification of abnormal trends in the micro-evolution of lesions. Unlike existing technologies that typically capture abnormal fluctuations in lesions using only a single distance metric or mean drift, this method introduces a proportional relationship into the radial step deviation rate, dynamically linking the dispersion of the measurement state in a globally balanced temporal environment with its local evolutionary stride. This enhances the sensitivity to sudden changes and anomalies in lesions, significantly improving the accuracy and robustness of detecting abnormal lesion changes.

[0020] For each measurement, the intensity of the lesion expansion boundary and the chlorosis index were used to determine the local extreme values, and the bimodal and bivalve points were marked. Within the lesion time window, the in-window ranking weight of any measurement is defined as its sequence number within the lesion time window. If a measurement is a bimodal point, the distance between that measurement and the first bimodal point in the reverse time direction is calculated and recorded as the bimodal distance. The ratio of the in-window ranking weight to the bimodal distance is used as the peak-valley adaptive weight. If a measurement is a bivalve point, the peak-valley adaptive weight is defined as the ratio of the in-window ranking weight to the number of bivalve points within the lesion time window. The peak-valley adaptive weights of the remaining measurements are constructed as the reciprocal of the in-window ranking weights. The peak-valley adaptive weights are used as weights to perform a weighted average of the radial deviation rate of each measurement within the lesion time window to obtain the weighted radial deviation WPB. The larger the weight value of the position within the window, the farther away it is from the current position. After obtaining the lesion time window in each measurement, a weighted radial step offset can be obtained, and it will not be replaced or updated in the future.

[0021] Unlike existing technologies that use fixed-window weighting or simple bilateral sequential weighting, the peak-valley adaptive weighting method considers both the relative order of the measurement within the time window and its proximity to local extreme points. This achieves adaptive focusing and weight enhancement of key information points, while actively weakening the influence of noise points and window edge points on diagnostic conclusions. The rationale is that the expansion and suppression of leaf spot disease is not a uniform and monotonic process, but rather exhibits a non-stationary evolution characterized by accelerated expansion after incubation, deceleration under suppression, and the possibility of rebound. Furthermore, the accumulation of antagonistic metabolites and the induction of resistance by the complex microbial community have a lag, meaning that state transitions often concentrate near a few critical peaks and valleys. These peaks and valleys carry information about dynamic switching, while ordinary uniform weighting dilutes the transition signal in a large number of stable measurements. Therefore, this method enhances the response sensitivity to critical transition moments by assigning higher weights to measurements near peaks and valleys, enabling the algorithm to stably capture key changes in lesion expansion even in complex fluctuating environments. The peak-valley adaptive weighting and subsequent judgment results form a closed-loop feedback, which is used to dynamically adjust the area of ​​interest within the time window, making the diagnosis more environmentally adaptable and evolutionary time-series sensitive to non-stationary disturbances caused by different temperatures, humidity, light, and individual leaf differences.

[0022] The difference vector of the disease feature vector between any measurement and its previous measurement is used as the current incremental direction, and the difference vector of the disease feature vector between the previous measurement and the two measurements before this measurement is used as the preceding incremental direction; the cosine value between the current incremental direction and the preceding incremental direction is used as the turning change degree TGM; the cosine value of the angle refers to the cosine value of the angle between the current incremental direction and the preceding incremental direction. For any measurement, the inflection point factor is calculated based on the weighted path step deviation and the degree of abrupt change in direction. : IFP = WPB / (TGM+WPB)×[1-exp(-TGM)]; Beneficial effects: Since the inflection point factor is obtained from the time-series analysis of the intensity of the lesion expansion boundary and the chlorosis index after the release of the compound microbial community, it can effectively quantify the failure risk corresponding to the continuous dominance of the accelerated expansion phase and the crossing of the reversible threshold in scenarios where the leaf lesion disease process has nonlinear phase switching between accelerated expansion and suppressed deceleration, including special conditions where the compound microbial community's effectiveness has a lag in establishing colonization and inducing resistance. That is, it quantifies the degree to which candidate compound microbial communities cannot detect the transition from accelerated to decelerated lesion expansion in the early stage, resulting in the continuous high activity of the outer edge of the lesion and the simultaneous accumulation of chlorosis and degradation. This can explain the trend of lesions changing from slow climbing to short-term leap expansion, providing a unified reference indicator for subsequent screening and ranking of compound microbial community programs, inference of reapplication windows, and statistical comparison of differences in the effective time nodes. This improves the timeliness and accuracy of identifying the effectiveness status of the compound microbial community, and provides an efficient and standardized time anchor and theoretical support for the screening and effectiveness mechanism research of compound microbial communities for leaf lesions.

[0023] Further, in step S500, the method for dynamically determining whether the complex microbial community is effective based on the inflection point factors is as follows: Each historical inflection point factor is constructed as a factor sequence. When the length of the factor sequence is greater than 10, a dynamic judgment is performed. The dynamic judgment process is as follows: An integer variable is set as the recently labeled quantity NREF; if all of the recently labeled NREF inflection point factors are less than the median of the factor sequence, the first effective condition is satisfied; the step difference is calculated for each element of the factor sequence to obtain the factor step difference. If the product of a factor step difference and its adjacent factor step differences is negative, it is marked as a lag factor. When the proportion of lag factors in the recently labeled NREF measurements is more than half, the second effective condition is satisfied; if both the first and second effective conditions are satisfied, the inflection point is determined and the complex microbial community is effective; if only the first effective condition is satisfied but the second effective condition is not, the inflection point is not determined but the complex microbial community is effective; other states are determined as the inflection point being undetermined and the complex microbial community not being effective.

[0024] Since the inflection point factor of a single measurement is easily affected by random noise, resulting in false inflection points or inflection point drift, this invention adopts a dual-condition determination mechanism based on historical sequences: on the one hand, it determines whether the overall level of the inflection point factor has entered the suppressed range, and on the other hand, it determines whether the local changes of the inflection point factor show a lag or reversal structure consistent with the inflection point, thereby improving the robustness and repeatability of the effective determination.

[0025] Dynamic judgment is only performed when the factor sequence length exceeds a preset threshold (greater than 10). This threshold can be adjusted based on the disease duration, sampling frequency, and study design. For example, it can be 10–14 under daily sampling conditions; under higher frequency sampling conditions, the threshold can be increased accordingly to maintain coverage of the same time span. An integer variable is set as the recent marker quantity to define the recent window, that is, dynamic judgment performs sliding evaluation on the most recently occurring data to ensure sensitivity to recent states.

[0026] This approach combines the first and second effective conditions to output a more granular dynamic state: When both conditions are met simultaneously, the inflection point is determined and the compound microbial community is effective. Mathematically, the recent inflection point factor is generally at a low level, and local changes exhibit a consistent inflection point reversal structure, suggesting that the compound microbial community has caused a shift in the lesion expansion dynamics and entered a relatively stable, suppressed phase. When only the first effective condition is met but the second is not, the inflection point is determined but the compound microbial community is effective. This state covers a common scenario: the compound microbial community has reduced the overall disease indicators, but the inflection point structure is not yet obvious or the reversal structure is insufficient due to sampling noise. This represents an unstable, unconfirmed stage where the disease is effective but the shift is not significant, prompting continued observation or additional sampling, as there may still be a risk of disease deterioration. Other states are determined as an inflection point not determined and the compound microbial community not effective, indicating that the recent inflection point factor has not entered the suppressed range, or local changes do not show a stable reversal structure, and the disease may still be in an expansion and enhancement phase or an uncontrolled fluctuation phase.

[0027] Preferably, all undefined variables in this invention, if not explicitly defined, can be manually set thresholds.

[0028] This invention also provides a system for identifying the effectiveness of a complex microbial community in detecting the inflection point of leaf lesion expansion. This system includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the method for identifying the effectiveness of a complex microbial community in detecting the inflection point of leaf lesion expansion. This system can run on computing devices such as desktop computers, laptops, handheld computers, and cloud data centers. The runnable system may include, but is not limited to, processors, memory, and server clusters. The processor executes the computer program within the following system units: The image acquisition unit is used to acquire leaf images daily after the application of the compound microbial community. The image preprocessing unit is used to preprocess the leaf image to obtain a comparison image; The feature vector calculation unit is used to convert the compared image into a disease feature vector, including the intensity of the lesion expansion boundary and the chlorosis index; The inflection point analysis unit is used to obtain inflection point factors through inflection point analysis of disease feature vectors.

[0029] The efficacy determination unit is used to dynamically determine whether the complex microbial community has taken effect based on the inflection point factor.

[0030] The beneficial effects of this invention are as follows: This invention provides a method for identifying the effectiveness of a complex microbial community in detecting the inflection point of leaf lesion expansion. It quantifies the extent to which candidate complex microbial communities cannot detect the transition from accelerated to decelerated lesion expansion in the early stages, leading to persistently high activity at the outer edge of the lesion and simultaneous accumulation of chlorosis and degradation. This explains the trend of lesions transitioning from slow climbing to short-term rapid expansion. It provides a unified reference indicator for subsequent screening and ranking of complex microbial community programs, inference of reapplication windows, and statistical comparison of differences in the time of effectiveness. This improves the timeliness and accuracy of identifying the effectiveness status of complex microbial communities and provides an efficient and standardized time anchor and theoretical support for the screening of complex microbial communities and the study of their effectiveness mechanisms for leaf lesions.

[0031] It is beneficial to apply the compound microbial community in the early stage when the area of ​​the lesion is not significantly changing but the activity of the outer edge has already changed. It can give an early assessment of the effectiveness of the compound microbial community, and the application personnel can decide in time whether to apply more, adjust the application frequency, or change the compound microbial community scheme without waiting for the endpoint results. Conversely, in continuous monitoring, it can identify the difference between the inflection point and the overall improvement when the inflection point is not determined. It can avoid the waste of excessive application and repeated application caused by short-term fluctuations, and also avoid the delay in intervention caused by passive treatment after the lesion enters the accelerated expansion stage. To address different plots and varying temperature and humidity conditions, leaf photographs are standardized into comparable comparative images, generating consistent inflection point indicators. This makes the results of multi-point field monitoring more interpretable, facilitating the formation of actionable management indicators such as the onset time distribution, stability scores, or rebound risk warnings for compound microbial formulations. Simultaneously, as a unified quantitative tool for screening and validating compound microbial formulations, it transforms the effective time from empirical judgment into a statistically significant key performance indicator. This enables R&D teams to conduct high-throughput comparative experiments more efficiently, and allows application teams to establish standardized operating procedures based on image evidence and time-series trends. Ultimately, this improves the timeliness, stability, and controllability of compound microbial biocontrol solutions from greenhouse to field implementation. Attached Figure Description

[0032] The above and other features of the present invention will become more apparent from the detailed description of the embodiments shown in conjunction with the accompanying drawings. In the accompanying drawings, the same reference numerals denote the same or similar elements. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without any creative effort. In the drawings: Figure 1 The diagram shows a flowchart of a method for identifying the effectiveness of a complex microbial community in detecting the inflection point of leaf lesion expansion. Figure 2 The diagram shows the structure of a complex microbial community effectiveness identification system for detecting the inflection point of leaf lesion expansion. Detailed Implementation

[0033] The following will provide a clear and complete description of the concept, specific structure, and technical effects of the present invention in conjunction with the embodiments and accompanying drawings, so as to fully understand the purpose, solution, and effects of the present invention. It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other.

[0034] like Figure 1 The diagram shows a flowchart of a method for identifying the effectiveness of a complex microbial community in detecting the inflection point of leaf lesion expansion. The following section will combine... Figure 1 This invention describes a method for identifying the effectiveness of a complex microbial community in detecting the inflection point of leaf lesion expansion according to an embodiment of the present invention. The method includes the following steps: S100, after the application of compound microbial community, image acquisition is carried out daily to obtain leaf images; S200, preprocess the blade image to obtain the comparison image; S300 converts the compared images into disease feature vectors, including the intensity of lesion expansion boundaries and the chlorosis index; S400 obtains inflection point factors through inflection point analysis of disease feature vectors; S500 dynamically determines whether the complex microbial community is effective based on inflection point factors.

[0035] Furthermore, in step S100, the method for obtaining leaf images by daily image acquisition after applying the compound microbial community is as follows: after applying the compound microbial community, at least one image acquisition is performed on the same leaf or the same leaf position area of ​​the target plant on a daily sampling cycle; a standardized parameter set for image acquisition is preset, which includes at least the shooting distance, shooting angle, imaging resolution, and shooting time period.

[0036] During each image acquisition, the relative distance and orientation between the leaf and the imaging device are made to meet a standardized parameter set, and the leaf to be photographed is fully within the viewfinder. The viewfinder also includes reference markers for cross-day correction. The reference markers are any one or more of the following: preset color block markers, grayscale markers, or a scale containing a preset color reference. The leaf images containing the reference markers are associated with their corresponding acquisition dates and stored to form a leaf image sequence for subsequent cross-day comparison.

[0037] Optionally, the monitoring period can cover the incubation and spread stages after pathogen infection, or be dynamically set according to the actual disease progression and research design. For each target plant, it is preferable to continuously track and photograph the same leaf; when leaf obstruction, angle deviation, or natural leaf curling makes stable tracking impossible, it is preferable to switch to a tracking strategy of the same leaf position area, that is, selecting leaf areas of the same node, the same leaf sequence, or the same canopy height as the sampling object to ensure the consistency of the sampling objects in terms of spatial location and physiological maturity.

[0038] The purpose of the standardized parameter set for preset image acquisition is to achieve consistency in acquisition across days. Shooting distance limits the scale variation of leaves in the image, preventing changes in the pixel ratio of lesions due to distance fluctuations. Shooting angle reduces leaf reflection, texture projection, and perspective distortion, preventing lesion boundaries from being destroyed by highlights or shadows. Imaging resolution ensures consistent sampling accuracy of lesion boundary details across consecutive days. Shooting time period mitigates white balance and brightness drift caused by changes in solar altitude angle. Optionally, the standardized parameter set also includes parameters such as exposure mode, white balance mode, focal length or lens equivalent field of view, whether to enable flash or fill light, and whether to enable image stabilization. More standardized settings can further improve consistency across days.

[0039] During each image acquisition, the relative distance and orientation between the leaf and the imaging device should meet a standardized parameter set to ensure the leaf is fully within the frame. Simple positioning aids, such as rulers, fixed supports, or graduated shooting frames, are typically used to constrain the shooting distance and composition range; alternatively, alignment frames and angle indicators on the image acquisition device can assist in maintaining consistent shooting posture. For the same leaf, the frame should ideally include stable geometric features such as the midrib, leaf tip, and leaf base to facilitate automatic alignment or scale normalization of the leaf area in subsequent steps.

[0040] The viewfinder image must include reference markers for cross-day correction. These reference markers can be any of the following: preset color block markers, grayscale markers, or a scale containing a preset color reference. Ideally, the reference markers should be in the same lighting environment as the leaves and located in a nearby area, ensuring that the light intensity and color temperature changes received by the reference markers are consistent with those of the leaves. These reference markers are needed in subsequent steps to achieve brightness normalization, color correction, and scale correction. In this embodiment, when the reference marker is a grayscale marker, the image brightness is normalized based on the known reflectance of the grayscale block; when the reference marker is a preset color block marker, the white balance and color gain of the image are corrected based on the target color value of the color block; when the reference marker is a scale containing a preset color reference, a scale reference is also provided to correct pixel scale changes caused by slight fluctuations in shooting distance. The goal of introducing reference markers is to reduce yellowing and greening artifacts of leaves caused by changes in lighting in cross-day images, and to avoid misinterpreting changes in shooting conditions as changes in disease progression.

[0041] Optionally, leaf images can be linked to target plant identifiers, leaf position identifiers, batches of compound microbial treatments, and pathogen inoculation or application times to form a complete time-series sample set. This facilitates subsequent statistical comparison of the inflection point times of different compound microbial treatment schemes. To suppress random bias from a single image capture, multiple images of the same leaf can be captured on the same collection date, and median or confidence level screening strategies can be used to determine the valid images for that day, thereby improving the robustness of the leaf image sequence.

[0042] Further, in step S200, the method for preprocessing the leaf image to obtain the comparison image is as follows: performing denoising and brightness equalization processing on the comparison image, detecting reference markers in the leaf image and obtaining the color reference value or grayscale reference value of the reference markers, performing color correction and brightness normalization on the leaf image based on the color reference value or grayscale reference value to obtain a corrected image; locating and segmenting the leaf region on the corrected image to obtain a leaf mask, setting or removing non-leaf region pixels to zero based on the leaf mask to form a leaf foreground image, and performing scale alignment on the leaf foreground image to obtain the comparison image.

[0043] The noise reduction process uses any one of median filtering, Gaussian filtering, or bilateral filtering; the brightness equalization process uses histogram equalization, local contrast enhancement, or shadow compensation based on the lighting model. In the leaf image, a reference marker is detected and its color or grayscale reference value is obtained. Based on the color or grayscale reference value, the leaf image is color-corrected and brightness-normalized to obtain a corrected image. 1) When the reference marker is a grayscale marker, the measured grayscale value of the grayscale block in the image is mapped to its preset grayscale reference, the brightness scaling factor is calculated, and the brightness of the entire image is normalized; 2) When the reference marker is a preset color block marker, the measured color value of the color block is mapped to the preset color reference, and the white balance and color cast are corrected through channel gain adjustment or color transformation matrix; 3) When the reference marker is a scale containing the preset color reference, a scale reference is provided simultaneously for subsequent scale alignment.

[0044] The scale alignment methods include 1) pixel scale conversion based on the scale benchmark provided by the reference marker; and 2) scaling and normalization based on geometric features such as leaf length and main vein length of the leaf foreground image.

[0045] Further, in step S300, the method for converting the compared image into a disease feature vector, including the lesion expansion boundary strength and chlorosis index, is as follows: lesion mask is obtained by identifying the lesion area in the compared image; the lesion boundary is calculated based on the lesion mask to obtain the lesion boundary length and lesion area; the lesion expansion boundary strength is constructed based on the lesion boundary length and lesion area, which is used to characterize the degree of active expansion of the outer edge of the lesion; the green percentage or green intensity is calculated under the constraint of the leaf mask, and the difference is made with the preset health benchmark or the same leaf historical benchmark to obtain the chlorosis index, which is used to characterize the degree of decay of leaf browning; the lesion expansion boundary strength and the chlorosis index are combined to form a disease feature vector.

[0046] In contrast to the steps of obtaining lesion masks through image lesion region identification, lesion region identification employs methods such as threshold segmentation based on color space, cluster segmentation, and texture features, or uses a pixel-level segmentation model to output a lesion probability map and then thresholds it for identification. The generation of the lesion mask is performed under the constraint of a leaf mask, meaning that lesions are identified only within the foreground area of ​​the leaf.

[0047] Furthermore, morphological opening and closing operations, hole filling, and small connected region removal are performed on the lesion mask to eliminate noise points and improve the coherence of the lesion area, thereby making the subsequent lesion boundary extraction more stable.

[0048] The lesion boundary is obtained using any one of the following methods: contour extraction operator, edge tracking, or mask morphological gradient. The lesion boundary length is taken as the pixel length of the lesion contour, and the lesion area is the number of pixels within the lesion mask. By default, the method for constructing the lesion expansion boundary strength is to use the ratio of the lesion boundary length to the lesion area as the lesion expansion boundary strength.

[0049] In some implementations, the strength of the lesion expansion boundary can also be constructed in an equivalent form, such as normalizing the boundary length to the leaf area or the lesion perimeter to the lesion area, or introducing boundary gradient weights in the calculation of the lesion boundary length to improve the identification of weak boundaries. However, these are all variations of the same kind that characterize the degree of expansion activity by the relative relationship between boundary length and area.

[0050] The green percentage represents the proportion of pixels in the leaf area that meet the criteria for green pixels, which is a value higher than the preset threshold of the green channel. The average green intensity represents the average green channel value in the leaf area, or the average value of the green-related components in color spaces such as HSV / Lab. Either the green percentage or the average green intensity can be selected. The preset health benchmark can be derived from a benchmark sample of healthy leaves or the initial benchmark frame before fungicide application. The historical benchmark for the same leaf can be the moving average of the previous day or several days. The chlorosis index focuses on relative changes by differing from the benchmark.

[0051] Furthermore, in step S400, the method for obtaining the inflection point factor through inflection point analysis using the disease feature vector is as follows: Using a natural day as a test, the first few tests of any test are defined as a disease retrospective window, where the disease retrospective window is a sequence of tests as objects, and the disease retrospective window includes any test. The value range of the several tests is NUM=[10,20]. The spatial restraint component DGC is obtained by performing a second-order difference on the intensity of the lesion expansion boundary within the disease retrospective window, and the first-order difference of all chlorosis indices is used as the physiological restraint component GTC. If the product of the spatial restraint component and the physiological restraint component corresponding to a test is negative, the test is defined as a trend deviation point, and the ratio of the lesion expansion boundary intensity and the chlorosis index at the trend deviation point is calculated as the penalty coefficient LY. Within the disease retrospective window, the spatial confidence weight THJ and physiological confidence weight TKJ for each measurement are calculated using a variance-based confidence weighting method. The spatial and physiological containment components are then weighted and summed to obtain the containment synergy degree FJP. The variance-based confidence weighting method involves calculating the variances of the lesion expansion boundary strength and chlorosis index for all lesion expansions within the disease retrospective window. The variances of the lesion expansion boundary strength and chlorosis index are then calculated, and 1 is added to each variance, with the reciprocal taken to obtain the spatial confidence weight THJ and physiological confidence weight TKJ. The specific formula for the weighted summation is: FJP = THJ × DGC + TKJ × GTC. If the measurement is a trend divergence point, a penalty coefficient LY is applied to correct the containment synergy degree. The specific correction formula is: FJP = (THJ × DGC + TKJ × GTC) × LY. The average value of the containment synergy of all trend divergence points within the disease retrospective window corresponding to a test is taken as the net inhibition increment. If there is no trend divergence point, the net inhibition increment is set as the median of the containment synergy of each disease retrospective window. The difference between the containment synergy of the test and the net inhibition increment is normalized by sigmoid to obtain the inflection point factor IFP. Preferably, in step S400, the method for obtaining the inflection point factor through inflection point analysis using the disease feature vector is as follows: Using a natural day as a test, the first few tests of any test are defined as a disease retrospective window, where the disease retrospective window is a sequence of tests as objects, and the disease retrospective window includes any test. The value range of the several tests is NUM=[10,20]. Within the time window of the lesion corresponding to any measurement, the mean vector of all disease feature vectors is calculated as the equilibrium centroid. The Euclidean distance between the disease feature vector of each measurement and the equilibrium centroid is calculated and defined as the radial offset. The Euclidean distance between the disease feature vector of each measurement and the disease feature vector of the previous measurement is calculated and defined as the evolution step distance. The ratio of the radial offset to the evolution step distance of each measurement is used as the radial step deviation rate KTT. For each measurement, the intensity of the lesion expansion boundary and the chlorosis index are used to determine local extreme values, and bimodal and bivalve points are identified. The specific method for determining local extreme values ​​is as follows: if the intensity of the lesion expansion boundary in the current measurement is greater than both the previous and subsequent measurements, it is recorded as a local peak with strong boundary; if it is less than both the previous and subsequent measurements, it is recorded as a local valley with strong boundary. If the chlorosis index in the current measurement is greater than both the previous and subsequent measurements, it is recorded as a local peak with chlorosis; if it is less than both the previous and subsequent measurements, it is recorded as a local valley with chlorosis. If the current measurement is marked as both a local peak with strong boundary and a local peak with chlorosis, it is determined to be a bimodal point; if it is marked as both a local valley with strong boundary and a local valley with chlorosis, it is determined to be a bivalve point. Within the lesion time window, the in-window ranking weight of any measurement is defined as its sequence number within the lesion time window. If a measurement is a bimodal point, the distance between that measurement and the first bimodal point in the reverse time direction is calculated and recorded as the bimodal distance. The ratio of the in-window ranking weight to the bimodal distance is used as the peak-valley adaptive weight. If a measurement is a bivalve point, the peak-valley adaptive weight is defined as the ratio of the in-window ranking weight to the number of bivalve points within the lesion time window. The peak-valley adaptive weights of the remaining measurements are constructed as the reciprocal of the in-window ranking weights. The peak-valley adaptive weights are used as weights to perform a weighted average of the radial deviation rate of each measurement within the lesion time window to obtain the weighted radial deviation WPB. The larger the weight value of the position within the window, the farther away it is from the current position. After obtaining the lesion time window in each measurement, a weighted radial step offset can be obtained, and it will not be replaced or updated in the future.

[0052] The difference vector of the disease feature vector between any measurement and its previous measurement is used as the current incremental direction, and the difference vector of the disease feature vector between the previous measurement and the two measurements before this measurement is used as the preceding incremental direction; the cosine value between the current incremental direction and the preceding incremental direction is used as the turning change degree TGM; the cosine value of the angle refers to the cosine value of the angle between the current incremental direction and the preceding incremental direction. For any measurement, the inflection point factor is calculated based on the weighted path step deviation and the degree of abrupt change in direction. : IFP = WPB / (TGM+WPB)×[1-exp(-TGM)]; Further, in step S500, the method for dynamically determining whether the complex microbial community is effective based on the inflection point factors is as follows: Each historical inflection point factor is constructed as a factor sequence. When the length of the factor sequence is greater than 10, a dynamic judgment is performed. The dynamic judgment process is as follows: An integer variable is set as the recently labeled quantity NREF; if all of the recently labeled NREF inflection point factors are less than the median of the factor sequence, the first effective condition is satisfied; the step difference is calculated for each element of the factor sequence to obtain the factor step difference. If the product of a factor step difference and its adjacent factor step differences is negative, it is marked as a lag factor. When the proportion of lag factors in the recently labeled NREF measurements is more than half, the second effective condition is satisfied; if both the first and second effective conditions are satisfied, the inflection point is determined and the complex microbial community is effective; if only the first effective condition is satisfied but the second effective condition is not, the inflection point is not determined but the complex microbial community is effective; other states are determined as the inflection point being undetermined and the complex microbial community not being effective.

[0053] Dynamic judgment is only performed when the factor sequence length is greater than a preset threshold (greater than 10). This threshold can be adjusted according to the disease duration, sampling frequency, and study design. For example, it can be set to 10-14 under daily sampling conditions; under higher frequency sampling conditions, the threshold can be increased accordingly to maintain coverage of the same time span. NREF is set to an integer between 3 and 7.

[0054] It should be noted that the lag factor determination is based on the step difference sign reversal structure, which is one way to realize the inflection point pattern. Without deviating from the overall idea of ​​this invention, the second effective condition can also be achieved by using equivalent local curvature, second-order difference sign change or trend reversal test. The technical effects are the same, and they are all used to determine the confirmability of the inflection point structure.

[0055] An embodiment of the present invention provides a composite microbial community effectiveness identification system for detecting the inflection point of leaf lesion expansion, such as... Figure 2 The diagram shows a structural diagram of a complex microbial community effectiveness identification system for detecting the inflection point of leaf lesion expansion according to the present invention. This embodiment of the complex microbial community effectiveness identification system for detecting the inflection point of leaf lesion expansion includes: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps in the above embodiment of the complex microbial community effectiveness identification method for detecting the inflection point of leaf lesion expansion.

[0056] The system includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program in units of the following system: The image acquisition unit is used to acquire leaf images daily after the application of the compound microbial community. The image preprocessing unit is used to preprocess the leaf image to obtain a comparison image; The feature vector calculation unit is used to convert the compared image into a disease feature vector, including the intensity of the lesion expansion boundary and the chlorosis index; The inflection point analysis unit is used to obtain inflection point factors by performing inflection point analysis through disease feature vectors. The efficacy determination unit is used to dynamically determine whether the complex microbial community has taken effect based on the inflection point factor.

[0057] The aforementioned system for identifying the effectiveness of a complex microbial community in detecting the inflection point of leaf lesion expansion can run on computing devices such as desktop computers, laptops, handheld computers, and cloud servers. The system that can run on this system may include, but is not limited to, processors and memory. Those skilled in the art will understand that the example described is merely an illustration of a system for identifying the effectiveness of a complex microbial community in detecting the inflection point of leaf lesion expansion and does not constitute a limitation on such a system. It may include more or fewer components, or a combination of certain components, or different components. For example, the system may also include input / output devices, network access devices, buses, etc.

[0058] The processor referred to can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. This processor is the control center of the composite microbial community effectiveness identification system for detecting the inflection point of leaf lesion expansion, and it connects various parts of the system via various interfaces and lines.

[0059] The memory can be used to store the computer program and / or modules. The processor, by running or executing the computer program and / or modules stored in the memory, and by calling the data stored in the memory, realizes various functions of the composite microbial community effectiveness identification system for detecting leaf lesion expansion inflection points. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created based on the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0060] Although the invention has been described in considerable detail and particularly with regard to several of the described embodiments, it is not intended to limit itself to any of these details or embodiments or any particular embodiment, thereby effectively covering the intended scope of the invention. Furthermore, the invention has been described above with respect to embodiments foreseeable by the inventors in order to provide a useful description, and non-substantial modifications to the invention that have not yet been foreseen may still represent equivalent modifications.

Claims

1. A method for identifying the effectiveness of a complex microbial community in detecting the inflection point of leaf lesion expansion, characterized in that, The method includes the following steps: S100, after the compound microbial community is released, leaf images are collected daily; S200, preprocess the blade image to obtain the comparison image; S300 converts the compared images into disease feature vectors, including the intensity of lesion expansion boundaries and the chlorosis index; S400 obtains inflection point factors through inflection point analysis of disease feature vectors; S500, dynamically judges whether the complex microbial community is effective based on inflection point factors; In step S400, the method for obtaining the inflection point factor through inflection point analysis using disease feature vectors is as follows: A natural day is considered as one measurement; the first few measurements of any given measurement are defined as the disease backtracking window; the second-order difference of the lesion expansion boundary intensity within the disease backtracking window is used to obtain the spatial restraint component; the first-order difference of all chlorosis indices is used as the physiological restraint component; if the product of the spatial restraint component and the physiological restraint component corresponding to a measurement is negative, then that measurement is defined as a trend deviation point; the ratio of the lesion expansion boundary intensity to the chlorosis index at the trend deviation point is calculated as the penalty coefficient; within the disease backtracking window, through... The spatial confidence weight and physiological confidence weight of each test are calculated using a variance-based confidence weighting method. The spatial and physiological containment components are then weighted and summed to obtain the containment synergy. If the test is a trend deviation point, a penalty coefficient is applied to correct the containment synergy. The average containment synergy of all trend deviation points within the disease retrospective window corresponding to a test is taken as the net inhibition increment. If there is no trend deviation point, the net inhibition increment is set as the median of the containment synergy within the disease retrospective window. The difference between the containment synergy and the net inhibition increment of the test is sigmoid normalized to obtain the inflection point factor.

2. The method for identifying the effectiveness of a complex microbial community in detecting the inflection point of leaf lesion expansion according to claim 1, characterized in that, In step S100, the method for obtaining leaf images by daily image acquisition after the application of compound microbial community is as follows: after the application of compound microbial community, at least one image acquisition is performed on the same leaf or the same leaf position area of ​​the target plant on a daily sampling cycle. A standardized set of parameters for image acquisition is preset, which includes at least the shooting distance, shooting angle, imaging resolution, and shooting time period.

3. The method for identifying the effectiveness of a complex microbial community in detecting the inflection point of leaf lesion expansion according to claim 1, characterized in that, In step S200, the method for preprocessing the leaf image to obtain the comparison image is as follows: performing denoising and brightness equalization processing on the comparison image, detecting reference markers in the leaf image and obtaining the color reference value or grayscale reference value of the reference markers, and performing color correction and brightness normalization on the leaf image based on the color reference value or grayscale reference value to obtain the corrected image. The leaf region is located and segmented in the corrected image to obtain a leaf mask. Based on the leaf mask, the pixels in the non-leaf region are set to zero or removed to form a leaf foreground image. The leaf foreground image is scaled to obtain a comparison image.

4. The method for identifying the effectiveness of a complex microbial community in detecting the inflection point of leaf lesion expansion according to claim 1, characterized in that, In step S300, the method for converting the comparison image into a disease feature vector, including the lesion expansion boundary strength and chlorosis index, is as follows: lesion mask is obtained by identifying the lesion area in the comparison image; the lesion boundary is calculated based on the lesion mask to obtain the lesion boundary length and lesion area; the lesion expansion boundary strength is constructed based on the lesion boundary length and lesion area; the green percentage or green intensity is calculated under the constraint of the leaf mask, and the difference is made with the preset health benchmark or the historical benchmark of the same leaf to obtain the chlorosis index; the lesion expansion boundary strength and chlorosis index constitute the disease feature vector.

5. The method for identifying the effectiveness of a complex microbial community in detecting the inflection point of leaf lesion expansion according to claim 1, characterized in that, In step S500, the method for dynamically determining whether the complex microbial community is effective based on the inflection point factors is as follows: each historical inflection point factor is constructed into a factor sequence. When the length of the factor sequence is greater than 10, dynamic judgment is performed. The dynamic judgment process is as follows: an integer variable is set as the recent label quantity NREF; if all the recent NREF inflection point factors are less than the median of the factor sequence, the first effective condition is determined to be met. The step difference is calculated for each element of the factor sequence. If the product of a factor step difference and its adjacent factor step differences is negative, it is marked as a lag factor. When the proportion of lag factors in the most recent NREF measurements is more than half, the second effective condition is determined to be met. If both the first and second conditions for effectiveness are met simultaneously, the inflection point is determined and the complex microbial community becomes effective. If only the first condition for effectiveness is met but the second condition for effectiveness is not met, it is determined that the inflection point is not determined but the compound microbial community is effective; otherwise, it is determined that the inflection point is not determined and the compound microbial community is not effective.

6. A composite microbial community effectiveness identification system for detecting the inflection point of leaf lesion expansion, characterized in that, The composite microbial community effectiveness identification system for detecting the inflection point of leaf lesion expansion includes: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the composite microbial community effectiveness identification method for detecting the inflection point of leaf lesion expansion as described in any one of claims 1-5. The composite microbial community effectiveness identification system for detecting the inflection point of leaf lesion expansion runs in a computing device.

Images