AI visual model-based pest and disease identification method and system

Abstract

The invention provides a pest and disease identification method and system based on an AI visual model, and the method comprises the steps: firstly obtaining visual collection data of different growth parts of a to-be-identified plant at different collection moments; constructing a disease and pest feature dynamic association chain containing core visual feature units, association strength parameters and cross-stage feature evolution path information of various diseases and pests in different development stages, and then calling an AI visual model to perform multi-dimensional feature interaction processing on the plant visual data set based on the association chain; the method comprises the steps of generating a feature interaction result and a preliminary pest recognition result, performing multi-round iterative optimization on interaction matching parameters based on historical feedback data to obtain optimized parameters and a corrected recognition result, and finally generating a recognition report containing pest type identification, development stage evaluation and feature matching description and sending the recognition report to a target terminal. Therefore, the accuracy, comprehensiveness and adaptability of disease and pest identification are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer vision, in particular to a pest and disease identification method and system based on an AI vision model. BACKGROUND

[0002] In the field of agriculture, accurate identification and timely prevention of plant pests and diseases are crucial for ensuring crop yield and quality. Traditional methods of plant pest and disease identification mainly rely on manual observation and experience-based judgment. These methods are not only inefficient, but also susceptible to the professional level and subjective factors of the observer, resulting in low identification accuracy. Especially in the early stages of pest and disease occurrence, symptoms are often not obvious, making manual identification more difficult and delaying the best prevention opportunity.

[0003] With the development of computer vision technology, some pest and disease identification methods based on image recognition have emerged. These methods usually involve collecting plant images, extracting features using image processing algorithms, and then comparing them with a pre-established pest and disease feature library to achieve identification. However, existing pest and disease identification methods based on image recognition have many limitations. On the one hand, they often only consider the image features of a plant at a certain moment or a certain part, ignoring the dynamic changes of pests and diseases in different growth parts and different development stages, resulting in inaccurate and incomplete identification results. On the other hand, these methods lack in-depth exploration and analysis of the relationship between features, making it difficult to effectively handle complex and variable pest and disease characteristics. When faced with new pests and diseases or similar pests and diseases, the recognition ability is significantly insufficient. In addition, existing methods lack a self-optimization and improvement mechanism in the identification process, making it difficult to continuously improve recognition accuracy based on feedback data from actual applications. SUMMARY

[0004] In view of the above-mentioned problems, in combination with the first aspect of the present application, the embodiments of the present application provide a pest and disease identification method based on an AI vision model, which comprises:

[0005] Obtaining a set of plant visual data to be identified, which contains visual collection data of different growth parts of the plant to be identified at different collection times;

[0006] Constructing a dynamic correlation chain of pest and disease characteristics, which contains core visual feature units of various pests and diseases at different development stages, correlation strength parameters between each core visual feature unit, and cross-stage feature evolution path information;

[0007] The AI vision model is called to perform multi-dimensional feature interaction processing on the plant visual data set based on the dynamic association chain of pest and disease characteristics, generate a feature interaction result and a preliminary pest and disease identification result corresponding to the to-be-identified plant visual data set, and the feature interaction result includes the interaction matching parameters of each growth part visual feature and the core visual feature unit in the association chain.

[0008] Based on the parameter adjustment rule in the historical pest and disease identification feedback data, the interaction matching parameters in the feature interaction result are processed by multiple rounds of iteration optimization to obtain optimized interaction matching parameters and a corrected pest and disease identification result.

[0009] Based on the corrected pest and disease identification result and the optimized interaction matching parameters, a pest and disease identification report containing pest and disease type identification, development stage evaluation and feature matching description is generated, and the pest and disease identification report is sent to the target terminal.

[0010] In another aspect, the embodiments of the present application also provide a pest and disease identification system based on an AI vision model, comprising:

[0011] A processor; a machine readable storage medium for storing machine executable instructions of the processor; wherein the processor is configured to execute the above-mentioned pest and disease identification method based on an AI vision model by executing the machine executable instructions.

[0012] In another aspect, the embodiments of the present application also provide a computer program product, which comprises machine executable instructions stored in a computer readable storage medium, and a processor of a computer device reads the machine executable instructions from the computer readable storage medium, and the processor executes the machine executable instructions, so that the computer device executes the above-mentioned pest and disease identification method based on an AI vision model.

[0013] Based on the above aspects, by acquiring visual collection data of different growth parts of the plant to be identified at different collection time points, and then constructing a disease and pest feature dynamic correlation chain, the core visual feature units, correlation strength parameters and cross-stage feature evolution path information of multiple diseases and pests at different development stages are integrated, which can deeply depict the dynamic change rule and internal correlation of the disease and pest features, and effectively solve the problem of ignoring the dynamic characteristics and feature correlation of the existing method. The AI vision model is called to perform multi-dimensional feature interaction processing on the plant visual data set based on the correlation chain, and the generated feature interaction result and preliminary disease and pest identification result fully consider the matching of the plant visual features and the core features of the disease and pest, which improves the accuracy and comprehensiveness of the identification. Based on the parameter adjustment rule in the historical disease and pest identification feedback data, the interaction matching parameters in the feature interaction result are iteratively optimized for multiple rounds, so that the model can continuously adjust and improve itself according to the feedback in the actual application, further improving the accuracy and adaptability of the identification. Finally, the disease and pest identification report containing the disease and pest type identification, development stage evaluation and feature matching description is generated, which provides detailed and accurate disease and pest information for agricultural practitioners, and helps them to take effective prevention and control measures in time, to ensure the healthy growth of crops and improve the efficiency of agricultural production. BRIEF DESCRIPTION OF DRAWINGS

[0014] Figure 1 is the execution flow diagram of the disease and pest identification method based on the AI vision model provided by an embodiment of the application.

[0015] Figure 2 is a schematic diagram of exemplary hardware and software components of the disease and pest identification system based on the AI vision model provided by an embodiment of the application. DETAILED DESCRIPTION

[0016] The application will be described in detail below with reference to the accompanying drawings, Figure 1 is a flow diagram of the disease and pest identification method based on the AI vision model provided by an embodiment of the application, and the disease and pest identification method based on the AI vision model will be described in detail below.

[0017] Step S110: acquiring a plant visual data set to be identified, the plant visual data set to be identified containing visual collection data of different growth parts of the plant to be identified at different collection time points.

[0018] In this embodiment, the tomato plants in the greenhouse are taken as an example for disease and pest identification. When collecting the visual data set of the plant to be identified, systematic visual data collection needs to be performed for different growth parts of the tomato plant. The growth parts of the tomato plant mainly include leaves (including new leaves, old leaves and tender leaves), stems (including main stems and lateral branches), fruits (including green fruits, ripe fruits and turning fruits) and flowers. In order to fully reflect the growth state of the tomato plant and the characteristics of possible diseases and pests, the visual data collection needs to be performed at different times, such as different time periods in the morning, noon and evening, and different weather conditions, such as sunny, cloudy and overcast days, so as to avoid the interference of light conditions on the visual features. The collection device is selected as an industrial camera with fixed focal length and resolution. The camera is installed on a movable track bracket and can be shot from different angles (such as front, side and top) for each growth part of the tomato plant. During the collection process, the image definition of each growth part needs to meet the subsequent feature extraction requirements. The leaves need to fully show the leaf surface, back and edge parts, the stems need to shoot the surface texture of different height sections, and the fruits need to shoot the overall shape and fruit surface details. For flowers, targeted collection needs to be performed during the flowering period.

[0019] In addition, when collecting private sensitive data such as location information of the greenhouse and personal information of the grower, privacy protection needs to be performed. The specific technical means include deleting the geographic location label in the image data, anonymizing the metadata related to personal information, encrypting the data in the transmission and storage process by using data encryption technology, and ensuring that the data does not leak in the collection, transmission and storage links.

[0020] Step S120: constructing a disease and pest feature dynamic correlation chain, the disease and pest feature dynamic correlation chain containing core visual feature units of multiple diseases and pests at different development stages, correlation strength parameters between each core visual feature unit and cross-stage feature evolution path information.

[0021] Step S121: acquiring whole-cycle growth monitoring data of multiple diseases and pests, the whole-cycle growth monitoring data containing continuous visual collection data and corresponding growth environment parameter records of each disease and pest from the initial occurrence stage, the spread stage to the serious damage stage.

[0022] For example, common diseases and pests of tomatoes include tomato early blight, tomato late blight, tomato gray mold, tomato leaf mold, tomato aphids, tomato whitefly, etc. For each of the above diseases and pests, a special monitoring sample library needs to be established to collect continuous visual collection data from the initial occurrence stage, the spread stage, to the serious damage stage. Taking tomato early blight as an example, the initial occurrence stage is usually manifested as the appearance of a needle tip size brown spot on the leaf. As the disease develops, the spot gradually expands to a circular or irregular shape with obvious concentric rings. In the serious damage stage, a large number of leaves are yellow and shed, and the fruits appear brown hard spots. During the entire monitoring period, the continuity of the visual collection data needs to be ensured, and the collection frequency is not less than three times a week until the disease and pest development reaches the serious damage stage or is effectively controlled. At the same time, the corresponding growth environment parameters need to be recorded simultaneously, including air temperature, air relative humidity, soil temperature, soil humidity, light duration, light intensity, CO2 concentration, etc. The above environmental parameters are collected in real time by sensors installed in the greenhouse, and the sensors are evenly distributed to ensure the representativeness of the collected data. For the obtained whole-cycle growth monitoring data, data quality assessment is needed to eliminate blurred, out-of-focus, overexposed or insufficient image data, and abnormal environmental parameter records to ensure data reliability.

[0023] Step S122: Extracting visual feature units of each development stage from the whole-cycle growth monitoring data of each disease and pest to obtain a multi-stage feature unit set corresponding to each disease and pest, wherein the visual feature units include color feature units, morphological feature units, texture feature units, and lesion area distribution feature units.

[0024] Step S1221: Segmenting the whole-cycle growth monitoring data of each disease and pest in time sequence, determining the segmentation interval according to the growth rate and morphological change amplitude of the disease and pest, and dividing the continuous monitoring data into multiple time periods, each time period corresponding to a development stage.

[0025] For example, taking tomato early blight as an example, the interval between segments is determined according to the growth rate and the range of morphological changes. In the initial occurrence stage, the development of the disease is slow and the range of morphological changes is small, so the interval between segments can be set to be longer, for example, three days for a time period; after entering the spread stage, the development of the disease is accelerated and the range of morphological changes is increased, so the interval between segments can be shortened to two days for a time period; to the serious damage stage, the disease develops rapidly, and the interval between segments is further shortened to one day for a time period. Through the above segmentation method, the continuous monitoring data of tomato early blight is divided into multiple time periods, and each time period corresponds to a clear development stage. In the segmentation process, the starting time and ending time of each time period need to be accurately marked to ensure that the time boundaries between stages are clear. For different types of diseases and pests, the interval between segments needs to be adjusted according to their respective growth characteristics, for example, the spread speed of tomato gray mold may be faster than that of tomato early blight, so the interval between segments needs to be shortened accordingly.

[0026] Step S1222: The visual acquisition data in each time period is preprocessed, and the color feature of the preprocessed lesion area visual data is extracted. The color space conversion method is used to convert the visual data from the RGB color space to the HSV color space, and the hue mean, saturation variance and brightness range of the lesion area are extracted as the color feature unit.

[0027] For example, the preprocessing step includes image denoising, image enhancement, contrast adjustment, etc. The image denoising adopts the Gaussian filtering method to remove the Gaussian noise in the image; the image enhancement enhances the contrast of the image through the histogram equalization method, so that the lesion area is more obvious. After preprocessing, the color feature of the lesion area visual data is extracted. First, the color space conversion method is used to convert the image data in the RGB color space to the HSV color space. In the HSV color space, the hue (H) component can more intuitively reflect the color of the species, the saturation (S) component reflects the purity of the color, and the brightness (V) component reflects the brightness of the color. For the lesion area, the hue mean value is extracted, which represents the overall color tendency of the lesion area; the saturation variance is extracted, which reflects the uniformity of the color purity of the lesion area; and the brightness range is extracted, which represents the change interval of the brightness of the lesion area. The hue mean value, saturation variance and brightness range are combined to form a color feature unit. For example, the lesion area of tomato early blight in the initial occurrence stage has a hue mean value in a certain range, a small saturation variance, and a narrow brightness range; as the disease develops, the hue mean value changes, the saturation variance increases, and the brightness range widens.

[0028] Step S1223: The morphological feature of the preprocessed lesion area visual data is extracted, the contour information of the lesion area is obtained through the edge detection algorithm, and the perimeter, area and circularity of the contour are calculated as the morphological feature unit.

[0029] In this embodiment, first, an edge detection algorithm such as the Canny edge detection algorithm is used to perform edge detection on the lesion area to obtain contour information of the lesion area. In the edge detection process, the threshold parameter needs to be reasonably set to ensure that the contour edge of the lesion area can be accurately detected, while avoiding the background noise from being mistakenly detected as an edge. After obtaining the contour information, the perimeter of the lesion area is calculated based on the contour, the perimeter representing the total length of the contour of the lesion area; the area surrounded by the contour is calculated, the area reflecting the size of the lesion area; the circularity is calculated through the formula (4 x π x area) / (perimeter x perimeter), which is used to describe the closeness of the contour of the lesion area to a circle. The above three, i.e., the perimeter, the area and the circularity, are taken as the morphological feature units. For example, the initial lesion area contour of the tomato early blight may be relatively regular, with a high circularity, a small perimeter and a small area; after entering the spreading stage, the contour gradually becomes irregular, the circularity decreases, and the perimeter and the area increase.

[0030] Step S1224: Texture feature extraction is performed on the preprocessed lesion area visual data, and the energy, the entropy and the contrast of the lesion area are calculated as the texture feature units by using the gray level co-occurrence matrix method.

[0031] For example, first, the color image is converted into a gray-scale image, and then the texture feature is calculated by using the gray level co-occurrence matrix method. The gray level co-occurrence matrix is a matrix constructed by counting the occurrence probability of different gray level pixel pairs in the image in a specific direction and distance. Based on the matrix, the energy, the entropy and the contrast of the lesion area are calculated. The energy reflects the uniformity and regularity of the image texture, the greater the energy value, the more uniform and regular the texture; the entropy reflects the complexity of the image texture, the greater the entropy value, the more complex the texture; the contrast reflects the definition and groove depth of the image texture, the greater the contrast value, the deeper the texture groove. The above three, i.e., the energy, the entropy and the contrast, are taken as the texture feature units. For example, the lesion area texture of the tomato leaf mildew may be relatively complex, with a large entropy value and a high contrast; while the lesion area texture of the tomato powdery mildew may be relatively uniform, with a large energy value and a small entropy value.

[0032] Step S1225: Lesion area distribution feature extraction is performed on the preprocessed lesion area visual data, and the position coordinates of the lesion area in the overall plant visual data, the distribution density and the distance between adjacent lesion areas are marked as the lesion area distribution feature units.

[0033] For example, first mark the position coordinates of each lesion area in the overall visual data of the plant. The coordinate system takes the upper left corner of the image as the origin, the horizontal direction as the X-axis, and the vertical direction as the Y-axis. The position of each lesion area is represented by the upper left corner and the lower right corner coordinates of its circumscribed rectangle. Then calculate the distribution density of the lesion area, which is measured by the number of lesion areas per unit area, i.e. the ratio of the total number of lesion areas in the image to the total area of the image. Finally, calculate the distance between adjacent lesion areas. For each lesion area in the image, find the other lesion area closest to it, calculate the Euclidean distance between the two, and calculate the mean and variance of the distances of all adjacent lesion areas. Combine the above position coordinates, distribution density, and distance between adjacent lesion areas as the lesion area distribution feature unit. For example, the lesion area of tomato aphid disease may be more scattered, with a lower distribution density and a larger distance between adjacent lesion areas; while the lesion area of tomato late blight may show a clustered distribution, with a higher distribution density and a smaller distance between adjacent lesion areas.

[0034] Step S1226: integrate the color feature unit, the morphological feature unit, the texture feature unit, and the lesion area distribution feature unit extracted in each time period to form a stage feature unit group corresponding to the time period; arrange the stage feature unit groups corresponding to all time periods of each disease and pest in chronological order to form a multi-stage feature unit set corresponding to each disease and pest.

[0035] In this embodiment, each feature unit is represented in the form of a feature vector. For example, the color feature unit can be represented as a three-dimensional vector containing the mean hue, the variance of saturation, and the range of brightness; the morphological feature unit is represented as a three-dimensional vector containing the perimeter, the area, and the circularity; the texture feature unit is represented as a three-dimensional vector containing the energy, the entropy, and the contrast; and the lesion area distribution feature unit is represented as a multi-dimensional vector containing the position coordinates (multiple coordinate points), the distribution density, and the adjacent distance (mean and variance). The above feature vectors are sequentially spliced to form a stage feature unit group corresponding to the time period. The stage feature unit group contains comprehensive visual feature information of the lesion area in the time period. For each disease and pest, arrange the stage feature unit groups corresponding to all time periods in chronological order, i.e. from the first time period of the initial occurrence stage to the last time period of the severe damage stage, to form a multi-stage feature unit set corresponding to the disease and pest. The multi-stage feature unit set completely records the changes in visual features during the entire process from the occurrence to the development of the disease and pest.

[0036] Step S123: perform core feature screening on the multi-stage feature unit set of each disease and pest, and retain the visual feature units that appear more frequently in any development stage of the disease and pest than in other stages and can uniquely identify the stage as core visual feature units to form a subset of core visual feature units for each stage of each disease and pest.

[0037] For example, taking the multi-stage characteristic unit set of tomato early blight as an example, for each development stage (initial occurrence stage, spread stage, severe damage stage), the frequency of occurrence of each visual characteristic unit in the stage characteristic unit group is statistically analyzed and compared with the frequency of occurrence in other stages. For example, in the initial occurrence stage, if a certain hue mean range occurs more frequently than in the spread and severe damage stages, and this hue mean range only appears in the initial occurrence stage, uniquely identifying this stage, then the color characteristic unit corresponding to this hue mean range is retained as the core visual characteristic unit of the initial occurrence stage. For morphological characteristic units, if a certain roundness range occurs most frequently in the spread stage and is significantly different from the roundness range in other stages, uniquely identifying the spread stage, then it is taken as the core morphological characteristic unit of the spread stage. Similarly, a similar screening is performed on texture characteristic units and lesion area distribution characteristic units. Through the above screening process, core visual characteristic units are determined for each development stage of each disease and pest, and these core visual characteristic units together constitute the core visual characteristic unit subset of that stage. There should be clear differences between the subsets of core visual feature units at each stage to ensure that different development stages can be accurately distinguished through the core visual feature units.

[0038] Step S124: Analyze the correlation between core visual feature units in adjacent development stages of each disease and pest, and calculate the co-occurrence frequency, morphological similarity, and evolutionary dependence of core visual feature units in adjacent stages respectively; normalize the co-occurrence frequency, morphological similarity, and evolutionary dependence to the same numerical range, and determine the correlation strength parameter between each core visual feature unit based on the weighted calculation results of the normalized three.

[0039] When analyzing the correlation between core visual feature units in adjacent developmental stages of each pest or disease, the core visual feature units of two adjacent developmental stages, such as the initial occurrence stage and the spread stage, are extracted first. The co-occurrence frequency is calculated, i.e., the frequency with which core visual feature units in two adjacent stages appear simultaneously; a higher co-occurrence frequency indicates a stronger correlation between the two. Morphological similarity is calculated by comparing various feature parameters of the core visual feature units, such as the mean hue of color feature units and the area of ​​morphological feature units, using methods such as cosine similarity or Euclidean distance. Higher morphological similarity indicates a stronger morphological correlation between the two. Evolutionary dependency is calculated, i.e., the degree to which the appearance of a later-stage core visual feature unit depends on the earlier-stage core visual feature unit. This can be determined by analyzing the causal relationship between the two over time; a higher evolutionary dependency indicates that the earlier-stage core visual feature unit is an important foundation for the evolution of the later-stage core visual feature unit.

[0040] After calculating co-occurrence frequency, morphological similarity, and evolutionary dependency, all three need to be normalized to the same numerical range, for example, to between 0 and 1. The normalization process uses the min-max normalization method: for a given parameter, its actual value is subtracted from its minimum value across all samples, and then divided by the difference between its maximum and minimum values ​​to obtain the normalized parameter value. After normalization, different weight coefficients are assigned to each parameter, determined based on their importance to the association relationship. For example, the weight coefficient for co-occurrence frequency can be set higher, followed by morphological similarity, and then evolutionary dependency. A weighted calculation is performed based on the normalized parameter values ​​and their corresponding weight coefficients. The weighted results are then summed to obtain the association strength parameter between each core visual feature unit. The association strength parameter ranges from 0 to 1; a higher value indicates a stronger association between core visual feature units at adjacent stages.

[0041] Step S125: Track the changes in the core visual feature units of each disease and pest from the initial occurrence stage to the severe damage stage, record the order of appearance, morphological transition mode and duration of each core visual feature unit, and form cross-stage feature evolution path information for each disease and pest.

[0042] Step S1251: Assign time labels to each development stage in the full-cycle growth monitoring data of each pest and disease, marking the start and end times of each development stage.

[0043] For example, taking tomato late blight as an example, its full-cycle growth monitoring data includes the initial occurrence stage, the spread stage, and the severe damage stage. A unique time label is assigned to each stage, such as using a stage number combined with the start and end times, like "Stage 1 - Start Time T1 - End Time T2", "Stage 2 - Start Time T2 - End Time T3", etc.

[0044] Step S1252: Extract the occurrence time of each core visual feature unit from the subset of core visual feature units of each stage of each disease and pest, wherein the occurrence time refers to the moment when the core visual feature unit first appears in the visual acquisition data.

[0045] For example, for each core visual feature unit, in the time period corresponding to the development stage to which it belongs, find the image frame in which the core visual feature unit first appears in the visual collection data, and the timestamp corresponding to the image frame is the appearance time of the core visual feature unit. For example, a certain core morphological feature unit of tomato late blight spread stage first appears in an image frame in the second time period of the stage, and the timestamp of the image frame is the appearance time of the core visual feature unit. The extraction of the appearance time needs to be accurate to seconds to ensure the accuracy of the time sequence. For core visual feature units that last for multiple time periods, only the first appearance time is recorded.

[0046] Step S1253: Sort all core visual feature units of each pest and disease in chronological order of appearance time to form a core visual feature unit time sequence.

[0047] For example, the appearance times of all core visual feature units can be compared and arranged in ascending order to form a core visual feature unit time sequence. In the time sequence, each core visual feature unit is arranged in order of its appearance in the development process of pests and diseases, for example, the core visual feature unit of the initial occurrence stage is arranged at the front, then the core visual feature unit of the spread stage, and finally the core visual feature unit of the severe damage stage. For multiple core visual feature units appearing in the same development stage, they are arranged in the corresponding position of the time sequence in order of their appearance time. The core visual feature unit time sequence clearly shows the appearance order of the core visual features in the development process of pests and diseases, reflecting the evolution process of pests and diseases.

[0048] Step S1254: Analyze the morphological differences between adjacent two core visual feature units in the core visual feature unit time sequence to determine the morphological transition mode between them, which includes gradual transition, sudden transition and superposition transition, where gradual transition refers to gradual change of morphological parameters, sudden transition refers to sudden change of morphological parameters, and superposition transition refers to addition of new morphological parameters to the previous core visual feature unit.

[0049] For example, taking two adjacent core visual feature units A and B as an example, A is the previous core visual feature unit and B is the next core visual feature unit. The morphological parameters of the two are compared, such as the circumference, area, circularity, etc. If the morphological parameters of B show a gradual change trend relative to the morphological parameters of A, that is, the parameters change continuously and smoothly in the time sequence without a jump change, it is determined that the morphological transition mode between the two is gradual transition. If the morphological parameters of B change significantly and jumpily in a short time relative to the morphological parameters of A, and the difference with the morphological parameters of A is large, it is determined to be a sudden transition. If the morphological parameters of B contain all the morphological parameters of A and add new morphological parameters on the basis of A, for example, add special contour shape features, it is determined to be superimposed transition. Through the above analysis, the morphological transition mode of each pair of adjacent core visual feature units in the time sequence is determined, and recorded in the transition mode list.

[0050] Step S1255: Calculate the time span of each core visual feature unit from the appearance time to the disappearance time, and take the time span as the duration of the core visual feature unit. If the core visual feature unit lasts until the end of the serious damage stage, the duration of the core visual feature unit is the time span from the appearance time to the end time of the serious damage stage.

[0051] In this embodiment, the disappearance time refers to the time when the core visual feature unit appears for the last time in the visual acquisition data. By traversing the visual acquisition data of the development stage and the subsequent stage to which the core visual feature unit belongs, the image frame in which the core visual feature unit no longer appears is found, and the previous time of the image frame is the disappearance time. If the core visual feature unit continues to appear until the end of the serious damage stage, the disappearance time is the end time of the serious damage stage. The duration is obtained by subtracting the appearance time from the disappearance time, that is, the time span experienced by the core visual feature unit from the first appearance to the last disappearance. The duration reflects the existence time of the core visual feature unit in the development process of the disease and pest.

[0052] Step S1256: Integrate the time sequence of the core visual feature unit, the morphological transition mode of each adjacent core visual feature unit, and the duration of each core visual feature unit to form the basic information of the cross-stage feature evolution path; draw the cross-stage feature evolution path graph of each disease and pest based on the basic information of the cross-stage feature evolution path, and mark the position, appearance time, duration and morphological transition mode of each core visual feature unit; supplement the change trend of the correlation strength parameter corresponding to each core visual feature unit in the cross-stage feature evolution path graph, and mark the change curve of the correlation strength parameter with time.

[0053] In this embodiment, the basic information is stored in the form of a data structure, including a time sequence linked list, a transition mode array, and a duration dictionary, etc. Based on the above basic information, a cross-stage feature evolution path diagram is drawn. The path diagram takes time as the horizontal axis and core visual feature units as the vertical axis. The position of each core visual feature unit on the horizontal axis corresponds to its occurrence time and duration, forming a time interval. In the path diagram, different shapes or colors are used to identify the types of core visual feature units, such as circular shapes for color feature units and square shapes for morphological feature units. Adjacent core visual feature units are connected by directed line segments, and the types or colors of the line segments represent the morphological transition modes, such as solid lines for gradual transition, dashed lines for sudden transition, and dotted lines for superimposed transition. The associated strength parameter values are annotated beside the line segments. In addition, in the cross-stage feature evolution path diagram, the trend of the associated strength parameter for each core visual feature unit is supplemented. By selecting multiple time points on the time axis, the associated strength parameters of the core visual feature unit and its adjacent core visual feature units at different time points are calculated, and the parameter values are connected to form a curve of the associated strength parameter over time, which can intuitively show the dynamic changes of the associated strength.

[0054] Step S1257: Integrate the cross-stage feature evolution path diagram and the corresponding textual description information to form the cross-stage feature evolution path information of each pest.

[0055] In this embodiment, the textual description information includes detailed explanations of each core visual feature unit in the cross-stage feature evolution path diagram, such as the type of feature unit, specific feature parameters, occurrence time, duration, etc.; specific manifestations of each morphological transition mode, such as the change process of morphological parameters in gradual transition and the specific change amplitude of sudden transition; analysis of the trend of the associated strength parameter, such as when the associated strength is enhanced or weakened and possible reasons, etc. The cross-stage feature evolution path diagram and the above textual description information are integrated to form the complete cross-stage feature evolution path information of each pest. The cross-stage feature evolution path information comprehensively records the evolution process of core visual feature units from the initial occurrence stage to the severe damage stage of pests.

[0056] Step S126: Integrate the core visual feature unit subset of each stage, the corresponding associated strength parameter, and the cross-stage feature evolution path information of each pest to form the feature association sub-chain of a single pest.

[0057] For example, taking tomato powdery mildew as an example, the feature association sub-chain contains the initial occurrence stage core visual feature unit subset, the spread stage core visual feature unit subset, and the severe damage stage core visual feature unit subset. The core visual feature unit subset of each stage is connected with the core visual feature unit subset of the adjacent stage through an association strength parameter, which reflects the association closeness between the core visual feature units of different stages. The cross-stage feature evolution path information describes the evolution order, transition mode and duration of each core visual feature unit in the entire disease and pest development process. Through the above integration, each part of the information is organically organized together to form a complete and structured single disease and pest feature association sub-chain. The sub-chain can clearly show the core features of the disease and pest at different development stages and their mutual relationship, and is a basic component of constructing the dynamic feature association chain of diseases and pests.

[0058] Step S127: Collect all single disease and pest feature association sub-chains, sort them by disease and pest type, mark similar core visual feature units between different disease and pest types, and supplement cross-disease and pest type feature association reference information.

[0059] Collect all single disease and pest feature association sub-chains, such as tomato early blight feature association sub-chain, tomato late blight feature association sub-chain, tomato gray mold feature association sub-chain, etc. Sort the above sub-chains by disease and pest type, such as fungal diseases, bacterial diseases, viral diseases, and pests. Each category contains the corresponding disease and pest feature association sub-chain. On the basis of classification and sorting, similar core visual feature units between different disease and pest types are marked. For example, tomato early blight and tomato late blight may have similar color feature units at a certain development stage. By comparing the hue mean value, saturation variance and other parameters of the two, the similarity is determined. For similar core visual feature units, add association marks in each feature association sub-chain, indicating the similar core visual feature units of other diseases and pests and their belonging disease and pest types. At the same time, supplement cross-disease and pest type feature association reference information, including similarity value of similar feature units, possible confusion reason analysis, and distinguishing points. The above reference information helps to avoid misjudgment caused by similar features in the subsequent disease and pest identification process.

[0060] Step S128: Based on the cross-disease and pest type feature association reference information, adjust the association strength parameters of the core visual feature units in each disease and pest feature association sub-chain, so that the feature association logic between different disease and pest feature association sub-chains remains consistent.

[0061] In this embodiment, for different pests and diseases with similar core visual feature units, the differences in their correlation strength parameters are analyzed. If the correlation strength parameters of similar feature units in different sub-chains differ greatly, it may lead to inconsistent feature correlation logic. In this case, the correlation strength parameters need to be adjusted according to the similarity values and distinguishing points in the reference information. For example, for core visual feature units with high similarity but belonging to different pests and diseases, their correlation strength parameters with adjacent feature units can be appropriately reduced to highlight their differences from other pests and diseases. During the adjustment process, the logic of the correlation strength parameters within each pest and disease feature correlation sub-chain must be self-consistent, and the feature correlation logic between different sub-chains must be consistent to avoid contradictory or conflicting correlation relationships. The adjusted correlation strength parameters need to be recalculated and updated to the feature correlation sub-chains.

[0062] Step S129: Integrate all the adjusted pest and disease feature correlation sub-chains to construct a pest and disease feature dynamic correlation chain containing feature correlation information of multiple pests and diseases, and store it in the feature correlation database of the AI vision model.

[0063] In this embodiment, during the integration process, a unified indexing mechanism needs to be established to uniquely identify all the core visual feature units of pests and diseases, ensuring accurate correlation between feature units in different sub-chains. The pest and disease feature dynamic correlation chain is stored in a mesh structure, with each node representing a core visual feature unit and the connections between nodes representing correlation relationships. The weight of the connection is the correlation strength parameter. The chain also contains cross-stage feature evolution path information and cross-pest and disease type feature correlation reference information. After construction, the pest and disease feature dynamic correlation chain is stored in the feature correlation database of the AI vision model. The feature correlation database uses a distributed storage architecture to ensure data security and access efficiency. At the same time, an update mechanism for the database is established, which can update the pest and disease feature dynamic correlation chain in a timely manner when new pest and disease monitoring data is collected or new pest and disease types are discovered, to ensure the timeliness and accuracy of the correlation chain.

[0064] Step S130: Call the AI vision model to perform multi-dimensional feature interaction processing on the plant visual data set based on the pest and disease feature dynamic correlation chain, generate feature interaction results and preliminary pest and disease identification results corresponding to the to-be-identified plant visual data set, and the feature interaction results include the interaction matching parameters of each growth part visual feature and the core visual feature unit in the correlation chain.

[0065] Step S131: Input the to-be-identified plant visual data set into the feature analysis module of the AI vision model, and perform zoning processing on the visual collection data according to the growth parts to obtain leaf area visual data, stem area visual data, and fruit area visual data of the to-be-identified plant.

[0066] In this embodiment, the visual data set of tomato plants is taken as an example. The primary task of the feature analysis module is to process the visual data by growth site. The input visual data is segmented by image segmentation algorithms, such as semantic segmentation algorithms based on deep learning. The semantic segmentation algorithm classifies each pixel in the image through a trained neural network model, identifying leaf regions, stem regions, fruit regions, and background regions, etc. For the leaf region, further differentiation is needed between different types of leaves such as new leaves, old leaves, and tender leaves. For the stem region, differentiation is needed between the main stem and lateral branch regions. For the fruit region, differentiation is needed between green fruit, ripe fruit, and turning fruit regions. After segmentation, the regions of each growth site are extracted from the original image to obtain leaf region visual data, stem region visual data, and fruit region visual data. The visual data of each region includes image information and corresponding position coordinate information, which is used for subsequent analysis of the distribution characteristics of the diseased area.

[0067] Step S132: Feature extraction is performed on each region visual data to obtain a leaf region visual feature unit set, a stem region visual feature unit set, and a fruit region visual feature unit set. Each region's feature unit set includes color feature units, shape feature units, and texture feature units.

[0068] In this embodiment, the feature extraction process is similar to the method described in step S122, but here it is for the region visual data of the plant to be identified, rather than the historical monitoring data of the plant diseases and pests. Taking the leaf region visual data as an example, pre-processing is first performed, including denoising, enhancement, contrast adjustment, etc., and then color feature units, shape feature units, and texture feature units are extracted. The color feature units are obtained by converting the RGB color space to the HSV color space, extracting the hue mean, saturation variance, and brightness range; the shape feature units are obtained by edge detection to obtain contour information, calculating the perimeter, area, and circularity; and the texture feature units are obtained by calculating the energy, entropy, and contrast of the gray level co-occurrence matrix. The feature extraction process for the stem region and the fruit region is the same as that for the leaf region, and their respective color feature units, shape feature units, and texture feature units are obtained. All the feature units extracted from each region are integrated to form the leaf region visual feature unit set, the stem region visual feature unit set, and the fruit region visual feature unit set. Each region's feature unit set contains comprehensive visual feature information of the region.

[0069] Step S133: The disease and pest feature dynamic association chain is retrieved from the feature association database of the AI visual model, and the leaf region visual feature unit set, the stem region visual feature unit set, and the fruit region visual feature unit set are preliminarily matched with the core visual feature units at each stage in the association chain.

[0070] In this embodiment, the dynamic correlation chain of pest and disease characteristics contains core visual feature units, correlation strength parameters and cross-stage feature evolution path information of various pests and diseases at different development stages. The leaf area visual feature unit set, the stem area visual feature unit set and the fruit area visual feature unit set are preliminarily matched with the core visual feature units at each stage in the correlation chain. The preliminary matching adopts a feature vector similarity calculation method, compares each feature unit in the visual feature unit set of the to-be-identified area with the feature vectors of all core visual feature units in the correlation chain, and calculates a similarity value. The similarity value can be calculated by using a cosine similarity, a Euclidean distance or the like. A preliminary matching degree threshold is set, and when the similarity value of a certain feature unit of the to-be-identified area and a certain core visual feature unit in the correlation chain is higher than the threshold, it is considered that the two are preliminarily matched successfully. Through the preliminary matching, the core visual feature units possibly related to the visual features of the to-be-identified plant are screened out.

[0071] Step S134: According to the preliminary matching result, the core visual feature unit group whose matching degree exceeds the preset matching degree threshold is screened out from each area visual feature unit set, the disease type and development stage corresponding to each core visual feature unit group are determined, and a candidate combination list is formed.

[0072] In this embodiment, the matching degree is obtained by calculating the average value of the similarity values of all preliminarily matched feature units. A preset matching degree threshold is set, and the core visual feature unit group whose matching degree exceeds the threshold is screened out. The core visual feature unit group refers to the combination of multiple core visual feature units belonging to the same development stage of the same pest and disease. The disease type and development stage corresponding to each core visual feature unit group are determined, for example, a certain core visual feature unit group belongs to the spread stage of tomato early blight. The above core visual feature unit group and the corresponding disease type and development stage information are integrated to form a candidate combination list. Each candidate combination in the candidate combination list represents a possible pest and disease and its development stage. When the candidate combination list is formed, it is necessary to ensure that the candidate combinations in the list have a certain diversity to avoid missing potential disease types.

[0073] Step S135: The multi-dimensional feature interaction module of the AI vision model is called, and based on the pest and disease information in the candidate combination list, an interaction correlation model between the visual feature units in each area is constructed, and the interaction correlation model is used to analyze the cooperative matching relationship between the visual feature units in different areas.

[0074] Step S1351: Select a candidate combination from the candidate combination list, extract the disease and pest type information and development stage information corresponding to the candidate combination; according to the disease and pest type information and development stage information, retrieve the leaf core visual feature unit, stem core visual feature unit and fruit core visual feature unit of the disease and pest at the corresponding stage from the disease and pest feature dynamic association chain.

[0075] In this embodiment, first, a candidate combination is selected from the candidate combination list, for example, a candidate combination of the tomato early blight at the spread stage is selected. The disease and pest type information (tomato early blight) and development stage information (spread stage) corresponding to the candidate combination are extracted. According to these information, the leaf core visual feature unit, stem core visual feature unit and fruit core visual feature unit of the disease and pest at the corresponding stage are retrieved from the disease and pest feature dynamic association chain. The above core visual feature units are the typical features of each growth part of the disease and pest at the spread stage, such as the round and ring-shaped lesion feature on the leaf, the brown and sunken lesion feature on the stem, and the brown and hard lesion feature on the fruit. During the retrieval process, the integrity and accuracy of the core visual feature unit need to be ensured, including various feature units such as color, shape and texture.

[0076] Step S1352: Determine the association weight between the leaf core visual feature unit and the stem core visual feature unit, the association weight between the leaf core visual feature unit and the fruit core visual feature unit, and the association weight between the stem core visual feature unit and the fruit core visual feature unit, which are set based on the cross-regional lesion transmission rule of the disease and pest at the corresponding stage.

[0077] In this embodiment, for example, the tomato early blight at the spread stage usually starts from the leaf, and then gradually spreads to the stem and fruit, so the association weight between the leaf core visual feature unit and the stem core visual feature unit can be set to be relatively high, the association weight between the leaf core visual feature unit and the fruit core visual feature unit is the second, and the association weight between the stem core visual feature unit and the fruit core visual feature unit is relatively low. The specific value of the association weight is determined by analyzing the historical incidence data and cross-regional transmission path of the disease and pest, and can be optimized by expert experience method or machine learning method. The value range of the association weight is between 0 and 1, and the sum of the weights can or can not be 1, which is determined according to the importance of the transmission rule.

[0078] Step S1353: constructing an input layer of the interactive correlation model, taking the leaf region visual feature unit set, the stem region visual feature unit set, and the fruit region visual feature unit set of the plant to be identified as input vectors; constructing a hidden layer of the interactive correlation model, and setting three feature processing sub-modules corresponding to the leaf region, the stem region, and the fruit region respectively, each feature processing sub-module performing feature mapping processing on the input region feature unit set to obtain a region feature mapping vector.

[0079] In this embodiment, each feature processing sub-module is composed of multiple fully connected layers and activation functions, and the number of neurons of the fully connected layer is determined according to the dimension of the input vector and the feature mapping requirement. The feature processing sub-module performs feature mapping processing on the input region feature unit set, maps the original feature vector to a higher-dimensional feature space through nonlinear transformation, and obtains a region feature mapping vector. The region feature mapping vector can more effectively represent the essential features of the region visual feature unit, and improve the accuracy of subsequent interactive matching. The activation function adopts a ReLU function or other nonlinear activation function to increase the nonlinear expression ability of the model.

[0080] Step S1354: adding a cross-region interactive module in the hidden layer, performing weighted fusion processing on the region feature mapping vectors of the three regions based on the preset correlation weight, and calculating the feature synergy value of the leaf region and the stem region, the feature synergy value of the leaf region and the fruit region, and the feature synergy value of the stem region and the fruit region.

[0081] In this embodiment, the cross-region interactive module performs weighted fusion processing on the region feature mapping vectors of the leaf region, the stem region, and the fruit region based on the preset correlation weight. For example, when calculating the feature synergy value of the leaf region and the stem region, the leaf region feature mapping vector and the stem region feature mapping vector are weighted and summed according to the correlation weight between them to obtain a fusion vector, and then a fully connected layer and an activation function are used to output the feature synergy value. Similarly, the feature synergy value of the leaf region and the fruit region, and the feature synergy value of the stem region and the fruit region are calculated. The feature synergy value reflects the degree of cooperative matching between the visual features of the two regions, and the higher the value, the more consistent the features of the two regions with the cross-region lesion characteristics of the disease and insect pests in the corresponding stage.

[0082] Step S1355: Construct the output layer of the interaction correlation model, and take the intra-regional feature matching degree and the cross-regional feature synergy value as output parameters, where the intra-regional feature matching degree refers to the matching degree of the region feature mapping vector of each region and the corresponding core visual feature unit, and the cross-regional feature synergy value refers to the mean value of the three cross-regional synergy values; based on historical interaction case data, the hidden layer parameters and the cross-regional correlation weight of the interaction correlation model are adjusted, so that the intra-regional feature matching degree and the cross-regional feature synergy value output by the model can accurately reflect the synergistic matching relationship between the different region visual feature units and the core visual feature unit.

[0083] In this embodiment, the intra-regional feature matching degree is obtained by calculating the similarity between the region feature mapping vector of each region and the feature vector of the corresponding core visual feature unit, and the similarity calculation method is the same as that in the preliminary matching. The cross-regional feature synergy value is the mean value of the three feature synergy values of the leaf region and the stem region, the leaf region and the fruit region, and the stem region and the fruit region. Based on historical interaction case data, the interaction correlation model is trained and adjusted. The historical interaction case data contains a large amount of known plant visual data to be identified of pest types and development stages, and the true values of the corresponding intra-regional feature matching degree and cross-regional feature synergy value. Through the back propagation algorithm, the weights and biases of the hidden layer of the model and the cross-regional correlation weight are adjusted, so that the error between the intra-regional feature matching degree and the cross-regional feature synergy value output by the model and the true value is minimized. In the training process, the cross-validation method is used to ensure the generalization ability of the model.

[0084] Step S1356: Construct the corresponding interaction correlation model for each candidate combination in the candidate combination list according to the above steps, and form a multi-model parallel processing framework.

[0085] In this embodiment, each candidate combination corresponds to a pest type and a development stage, so an independent interaction correlation model needs to be constructed. All models form a multi-model parallel processing framework, which can simultaneously process feature interaction of multiple candidate combinations to improve processing efficiency. The multi-model parallel processing framework adopts a distributed computing architecture, and different models are distributed to different computing nodes for running. The nodes communicate and share data through the network. The framework also includes a model scheduling and load balancing mechanism, which reasonably allocates computing tasks according to the computing complexity of each model and the computing resources of the nodes, ensures the load balancing of each node, and avoids the situation that a node is overloaded.

[0086] Step S136: Calculate the interaction matching parameters of each region visual feature unit and the corresponding core visual feature unit group through the interaction correlation model, and the interaction matching parameters include the intra-regional feature matching degree, the cross-regional feature synergy degree, and the stage feature fitting degree; and integrate the interaction matching parameters of each region to form the feature interaction result corresponding to the plant visual data set to be identified.

[0087] In this embodiment, the interaction matching parameters include intra-regional feature matching degree, cross-regional feature synergy degree, and stage feature matching degree. The intra-regional feature matching degree is the average of the intra-regional feature matching degrees of the leaf region, the stem region, and the fruit region. The cross-regional feature synergy degree is the average of the three cross-regional feature synergy values. The stage feature matching degree is obtained by comparing the visual feature unit set of the to-be-identified region with the cross-stage feature evolution path information of the corresponding development stage of the pest and disease, and calculating the overall matching degree. The calculation of the stage feature matching degree takes into account the evolution information such as the order of appearance and the transition mode of the feature unit. The interaction matching parameters corresponding to each candidate combination are integrated, including the intra-regional feature matching degree, the cross-regional feature synergy degree, and the stage feature matching degree, to form a feature interaction result corresponding to the to-be-identified plant visual data set. The feature interaction result is stored in the form of a data matrix, with each row representing a candidate combination and each column representing an interaction matching parameter.

[0088] Step S137: Based on the distribution of the interaction matching parameters in the feature interaction result, the intra-regional feature matching degree, the cross-regional feature synergy degree, and the stage feature matching degree are normalized to the same numerical interval, respectively. From the candidate combination list, the candidate combination with the highest normalized intra-regional feature matching degree, the optimal normalized cross-regional feature synergy degree, and the best normalized stage feature matching degree is selected as the type and development stage of the pest and disease of the to-be-identified plant. A preliminary pest and disease identification result is generated, which includes the identified pest and disease type identifier, the development stage identifier, and the corresponding interaction matching parameters.

[0089] Step S1371: Extract the interaction matching parameters in the feature interaction result, and for each candidate combination in the candidate combination list, organize the corresponding intra-regional feature matching degree, cross-regional feature synergy degree, and stage feature matching degree. The intra-regional feature matching degree refers to the average of the matching degrees of the leaf region, the stem region, and the fruit region. The cross-regional feature synergy degree refers to the average of the three cross-regional synergy values. The stage feature matching degree refers to the overall matching degree with the corresponding stage core feature in the association chain.

[0090] In this embodiment, for each candidate combination, the intra-regional feature matching degree is calculated, which is the arithmetic mean of the intra-regional feature matching degrees of the leaf region, the stem region, and the fruit region. The cross-regional feature synergy degree is the arithmetic mean of the three cross-regional feature synergy values. The stage feature matching degree refers to the overall matching degree of the visual feature unit set of the to-be-identified plant and the corresponding stage core feature in the association chain, which is calculated by considering factors such as similarity, order of appearance, and shape transition of the feature unit. The above parameter values and the pest and disease type and development stage information of the candidate combination are stored in association.

[0091] Step S1372: Normalize the intra-regional feature matching degree, the cross-regional feature synergy degree and the stage feature fitting degree of each candidate combination to the same numerical interval respectively; set weight coefficients for the normalized intra-regional feature matching degree, the normalized cross-regional feature synergy degree and the normalized stage feature fitting degree, wherein the weight coefficient of the normalized intra-regional feature matching degree is the highest; based on the weight coefficients, perform weighted calculation on the three normalized parameters of each candidate combination to obtain the comprehensive matching score of each candidate combination.

[0092] In this embodiment, the normalization adopts the min-max normalization method, that is, for a certain parameter, the minimum value and the maximum value of the parameter in all candidate combinations are taken as the reference, and the parameter value of each candidate combination is converted to between 0 and 1. After normalization, weight coefficients are set for the three parameters. The weight coefficients are set according to the priority of pest and disease recognition. Generally, the intra-regional feature matching degree has the greatest influence on the recognition result, so its weight coefficient is the highest, the cross-regional feature synergy degree is the second, and the stage feature fitting degree is the lowest. The specific values of the weight coefficients can be determined by expert experience or machine learning method, and the sum of the weights is 1. Based on the weight coefficients, the three normalized parameters of each candidate combination are weighted calculated. The normalized intra-regional feature matching degree is multiplied by its weight coefficient, and the normalized cross-regional feature synergy degree is multiplied by its weight coefficient, and the normalized stage feature fitting degree is multiplied by its weight coefficient, to obtain the comprehensive matching score of each candidate combination. The comprehensive matching score reflects the overall matching degree of the candidate combination and the visual features of the plant to be identified.

[0093] Step S1373: Sort the candidate combinations in the candidate combination list in order from high to low according to the comprehensive matching scores to obtain a sorted candidate combination list; select the top K candidate combinations with the highest comprehensive matching scores in the sorted candidate combination list as target candidate combinations.

[0094] In this embodiment, in the sorted list, the candidate combination with the highest comprehensive matching score is ranked first, and so on. The top K candidate combinations with the highest comprehensive matching scores in the sorted candidate combination list are selected as target candidate combinations. The value of K can be set according to actual conditions, for example, K=3 or K=5, the top 3 or top 5 candidate combinations are selected. The target candidate combinations are the objects of further analysis and screening in the subsequent process, which can improve the accuracy and reliability of the recognition result.

[0095] Step S1374: Analyze the common occurrence of the disease and pest type corresponding to the target candidate combination on the plant variety to be identified. If the difference between the occurrence frequency of the disease and pest type corresponding to any one target candidate combination on the plant variety and the occurrence frequency of the disease and pest type corresponding to the other two target candidate combinations on the plant variety exceeds the preset frequency difference range, the target candidate combination with the higher occurrence frequency is given priority.

[0096] For example, the plant to be identified is a tomato, and the disease and pest types corresponding to the target candidate combinations are tomato early blight, tomato late blight, and tomato gray mold, respectively. By querying the tomato disease and pest database, the occurrence frequencies of the three diseases and pests on the tomato variety are obtained. The occurrence frequency refers to the ratio of the historical occurrence times of the disease and pest on the plant variety to the total occurrence times of the disease and pest. By comparing the occurrence frequencies corresponding to the target candidate combinations, if the difference between the occurrence frequency corresponding to any one target candidate combination and the occurrence frequencies corresponding to the other two target candidate combinations exceeds the preset frequency difference range, for example, the difference is greater than a preset percentage, the target candidate combination with the higher occurrence frequency is given priority. The disease and pest type with the higher occurrence frequency is more likely to occur on the plant variety, and thus has a higher priority.

[0097] Step S1375: If the differences between the occurrence frequencies of the disease and pest types corresponding to the target candidate combinations on the plant variety are within the preset frequency difference range, the stage feature matching degree of each target candidate combination is further analyzed, and the target candidate combination with the highest stage feature matching degree is selected.

[0098] If the differences between the occurrence frequencies of the disease and pest types corresponding to the target candidate combinations on the plant variety are within the preset frequency difference range, that is, the occurrence frequencies of the disease and pest types are relatively close, the stage feature matching degree of each target candidate combination is further analyzed. The stage feature matching degree reflects the overall matching degree between the visual features of the plant to be identified and the core features of the corresponding development stage of the disease and pest. The target candidate combination with the highest stage feature matching degree is selected, and the development stage of the disease and pest of the combination is most matched with the actual development stage of the plant to be identified.

[0099] Step S1376: If the differences between the stage feature matching degrees of the target candidate combinations are within the preset matching degree difference range, the cross-regional feature coordination degree of each target candidate combination is analyzed, and the target candidate combination with the highest cross-regional feature coordination degree is selected. The disease and pest type and development stage corresponding to the finally selected target candidate combination are determined as the disease and pest type and development stage of the plant to be identified.

[0100] If the stage feature matching degree difference of each target candidate combination is within the preset matching degree difference range, that is, the stage feature matching degree is relatively close, the cross-region feature coordination degree of each target candidate combination is analyzed. The cross-region feature coordination degree reflects the coordination matching degree between the visual features of different growth parts. The higher the coordination degree, the more the disease characteristics of each region conform to the cross-region transmission rule of the disease and pest. The target candidate combination with the highest cross-region feature coordination degree is selected. Through the above layer-by-layer screening, the disease and pest type and development stage corresponding to the finally selected target candidate combination are the disease and pest type and development stage of the plant to be identified. A preliminary disease and pest identification result is generated, including the determined disease and pest type identifier, development stage identifier and corresponding interactive matching parameters, such as intra-regional feature matching degree, cross-region feature coordination degree, stage feature matching degree, etc.

[0101] Step S140: Based on the parameter adjustment rule in the historical disease and pest identification feedback data, the interactive matching parameters in the feature interaction result are subjected to multi-round iterative optimization processing to obtain the optimized interactive matching parameters and the corrected disease and pest identification result.

[0102] Step S141: Obtain historical disease and pest identification feedback data, which includes a set of historical plant visual data to be identified, historical feature interaction results, historical preliminary disease and pest identification results, and historical corrected disease and pest identification results.

[0103] In this embodiment, the historical disease and pest identification feedback data is first obtained, which comes from past disease and pest identification cases. The historical disease and pest identification feedback data includes a set of historical plant visual data to be identified, that is, the visual collection data of each growth part of the past plant to be identified; the historical feature interaction result, that is, the interactive matching parameter in the feature interaction result at that time; the historical preliminary disease and pest identification result, that is, the preliminary identification result obtained by the AI visual model at that time; and the historical corrected disease and pest identification result, that is, the correct identification result after artificial review or actual verification. The collection of historical data needs to cover multiple disease and pest types, different development stages and different plant varieties to ensure the diversity and representativeness of the data.

[0104] Step S142: Extract the historical interactive matching parameters corresponding to the historical feature interaction results and the target interactive matching parameters corresponding to the historical corrected disease and pest identification results from the historical disease and pest identification feedback data; calculate the parameter difference between the historical interactive matching parameters and the target interactive matching parameters in each historical case, and classify the parameter difference according to the feature attributes of the set of historical plant visual data to be identified, including plant variety attributes, collection season attributes and lesion region attributes.

[0105] In this embodiment, the historical interaction matching parameters are intra-regional feature matching degree, cross-regional feature coordination degree, and stage feature matching degree in the historical feature interaction results; the target interaction matching parameters are correct interaction matching parameters corresponding to the historical correction plant disease and pest recognition results, which are accurate parameter values verified in historical cases. The parameter difference between the historical interaction matching parameters and the target interaction matching parameters in each historical case is calculated, i.e., the target interaction matching parameters minus the historical interaction matching parameters. The parameter difference is classified according to the feature attributes of the historical plant visual data set. The feature attributes include plant variety attributes, such as tomatoes, apples, and rice; collection season attributes, such as spring, summer, autumn, and winter; and lesion region attributes, such as leaf region, stem region, and fruit region. The parameter difference is classified according to different combinations of plant varieties, collection seasons, and lesion regions, for example, “tomato-summer-leaf region” is a feature attribute combination category, and the parameter difference of all historical cases in this category is collected.

[0106] Step S143: The parameter difference under each type of feature attribute is normalized respectively, the variation trend of the parameter difference under each type of feature attribute is analyzed, the parameter adjustment direction and parameter adjustment amplitude corresponding to different feature attribute combinations are determined, and a mapping relationship table is formed.

[0107] Step S1431: The feature attributes of the historical plant visual data set to be recognized are divided into multiple variety categories according to plant variety attributes, multiple season categories according to collection season attributes, and multiple lesion region categories according to lesion region attributes; for each combination of variety categories, season categories, and lesion region categories, all historical cases under the combination are filtered out to form a feature attribute combination case set.

[0108] In this embodiment, the plant variety attributes can be divided into multiple variety categories, such as tomato category, apple category, and rice category; the collection season attributes can be divided into multiple season categories, such as spring, summer, autumn, and winter; and the lesion region attributes can be divided into multiple lesion region categories, such as leaf region, stem region, and fruit region. For each combination of variety categories, season categories, and lesion region categories, all historical cases under the combination are filtered out to form a feature attribute combination case set. For example, the combination of variety category as tomato category, season category as summer, and lesion region category as leaf region has a case set containing all historical cases of summer-collected tomato leaf lesions.

[0109] Step S1432: Calculate the parameter difference mean value of the historical interaction matching parameter and the target interaction matching parameter in each feature attribute combination case set, which includes the intra-region feature matching degree difference mean value, the cross-region feature synergy degree difference mean value, and the stage feature fit degree difference mean value; analyze the positive and negative of the parameter difference mean value, if the intra-region feature matching degree difference mean value is positive, it is determined that the parameter adjustment direction of the intra-region feature matching degree under the feature attribute combination is to be reduced; if it is negative, it is determined that the parameter adjustment direction of the intra-region feature matching degree under the feature attribute combination is to be increased; the parameter adjustment direction of the cross-region feature synergy degree and the stage feature fit degree is determined according to the same logic.

[0110] In this embodiment, for the intra-region feature matching degree difference mean value, if it is positive, it means that the historical interaction matching parameter is generally lower than the target interaction matching parameter, so the parameter adjustment direction is to be increased; if it is negative, it means that the historical interaction matching parameter is generally higher than the target interaction matching parameter, so the adjustment direction is to be reduced. The parameter adjustment direction determination logic of the cross-region feature synergy degree and the stage feature fit degree is the same as that of the intra-region feature matching degree. By analyzing the positive and negative of the parameter difference mean value, the adjustment direction of each interaction matching parameter under each feature attribute combination is determined.

[0111] Step S1433: Calculate the standard deviation of the parameter difference in each feature attribute combination case set, determine the parameter adjustment amplitude level according to the standard deviation, the larger the standard deviation, the higher the parameter adjustment amplitude level, and the higher the parameter adjustment amplitude level, the larger the corresponding parameter adjustment amplitude; assign the corresponding intra-region feature matching degree parameter adjustment direction, cross-region feature synergy degree parameter adjustment direction, stage feature fit degree parameter adjustment direction, and corresponding parameter adjustment amplitude to each feature attribute combination.

[0112] In this embodiment, the standard deviation reflects the dispersion degree of the parameter difference. According to the size of the standard deviation, the parameter adjustment amplitude is divided into multiple levels, for example, three levels of low, medium, and high. The larger the standard deviation, the greater the fluctuation of the parameter difference, and the greater the adjustment amplitude is needed to correct, so the higher the parameter adjustment amplitude level, and the larger the corresponding parameter adjustment amplitude. The specific value of the parameter adjustment amplitude is determined according to the adjustment amplitude level and the difference mean value, for example, the low-level adjustment amplitude can be set to a small fixed value, the medium-level to a medium fixed value, and the high-level to a large fixed value, or calculated by the product of the standard deviation and the difference mean value. Assign the corresponding intra-region feature matching degree, cross-region feature synergy degree, stage feature fit degree parameter adjustment direction and parameter adjustment amplitude to each feature attribute combination to form a mapping relationship table entry containing feature attribute combination, adjustment direction, and adjustment amplitude.

[0113] Step S1434: The feature attribute combination and the corresponding parameter adjustment direction and parameter adjustment range are arranged into a table form to form a mapping relationship table, wherein the feature attribute combination is composed of the variety category, the season category, and the lesion area category; the data in the mapping relationship table is supplemented and perfected, if the number of historical cases corresponding to any one feature attribute combination is less than the preset case number threshold, the parameter adjustment direction and the parameter adjustment range of the similar feature attribute combination are referred to, and the particularity of the feature attribute combination is fine-tuned to make the mapping relationship table cover all possible feature attribute combinations.

[0114] In the embodiment, the mapping relationship table takes the variety category, the season category, and the lesion area category as a composite primary key, each primary key corresponds to a row of records, and contains the intra-area feature matching degree adjustment direction, the intra-area feature matching degree adjustment range, the cross-area feature synergy degree adjustment direction, the cross-area feature synergy degree adjustment range, the stage feature fitness degree adjustment direction, and the stage feature fitness degree adjustment range. The data in the mapping relationship table is supplemented and perfected, for the feature attribute combination with the number of historical cases less than the preset case number threshold, due to insufficient data, the parameter difference mean and the standard deviation cannot be accurately calculated. At this time, the parameter adjustment direction and the adjustment range of the similar feature attribute combination are referred to, the similar feature attribute combination refers to the combination with similar variety category, same season category, or same lesion area category. The particularity of the feature attribute combination is fine-tuned, such as the growth characteristics of the variety in a specific season, the special structure of the lesion area, and the like, to ensure that the mapping relationship table can cover all possible feature attribute combinations and avoid data missing.

[0115] Step S144: The plant variety attribute, the collection season attribute, and the lesion area attribute of the plant visual data set to be recognized are extracted, the corresponding parameter adjustment direction and parameter adjustment range are found in the mapping relationship table; and the intra-area feature matching degree, the cross-area feature synergy degree, and the stage feature fitness degree in the feature interaction result are respectively adjusted for the first time according to the found parameter adjustment direction and parameter adjustment range to obtain the first adjusted interaction matching parameter.

[0116] The feature attributes of the plant visual data set to be identified are extracted, including plant variety attributes such as tomatoes, collection season attributes such as summer, and lesion area attributes such as leaf area. In the mapping relationship table, the corresponding parameter adjustment direction and parameter adjustment amplitude are found according to the combination of the above three attributes. For example, it is found that the combination of "tomato-summer-leaf area" corresponds to the adjustment direction of the intra-region feature matching degree being improved, the adjustment amplitude being a certain value; the adjustment direction of the cross-region feature coordination degree being improved, the adjustment amplitude being another value; the adjustment direction of the stage feature fit degree being reduced, and the adjustment amplitude being another value. According to the adjustment direction and adjustment amplitude found, the interaction matching parameters in the feature interaction result are adjusted for the first time. The adjustment method is: if the adjustment direction is to improve, the current interaction matching parameter is added to the adjustment amplitude; if the adjustment direction is to reduce, the current interaction matching parameter is subtracted by the adjustment amplitude. The value of the adjustment amplitude is determined according to the mapping relationship table, so as to ensure that the adjusted parameter is closer to the target interaction matching parameter. After the first adjustment, the first adjusted interaction matching parameter is obtained.

[0117] Step S145: Based on the first adjusted interaction matching parameter, the candidate combination is re-screened to determine the type and development stage of the plant to be identified, and the first adjusted pest and disease identification result is obtained; the historical corrected pest and disease identification result of the same type and stage as the first adjusted pest and disease identification result in the historical feedback data is extracted, and the corresponding target interaction matching parameter range is obtained.

[0118] In this embodiment, the screening process is similar to step S137, including steps of calculating a comprehensive matching score, sorting, selecting a target candidate combination, etc., to obtain the first adjusted pest and disease identification result, i.e., a new pest and disease type and development stage. The historical corrected pest and disease identification result of the same type and stage as the first adjusted pest and disease identification result in the historical feedback data is extracted, i.e., the historical cases of the same pest and disease type and development stage. The corresponding target interaction matching parameters are obtained from these historical cases, the value range of these parameters is counted, for example, the minimum and maximum values of the target interaction matching parameter are calculated, and the target interaction matching parameter range is determined. The target interaction matching parameter range reflects the reasonable value interval of the interaction matching parameter for the pest and disease type at the development stage.

[0119] Step S146: Determine whether the first adjusted interaction matching parameter is within the target interaction matching parameter range, if not, further adjust the parameter adjustment amplitude according to the mapping relationship table; based on the adjusted parameter adjustment amplitude, the first adjusted interaction matching parameter is adjusted for the second time to obtain the second adjusted interaction matching parameter and the second adjusted pest and disease identification result.

[0120] In this embodiment, for each interaction matching parameter, such as intra-regional feature matching degree, it is checked whether its value is between the minimum value and the maximum value of the target interaction matching parameter range. If all parameters are within the range, the adjustment ends; if there is at least one parameter that is not within the range, further adjustment is needed. According to the mapping table, the adjustment amplitude correction coefficient corresponding to the parameter not within the range is found under the combination of the feature attributes. The adjustment amplitude correction coefficient is determined according to the degree of exceeding the range of the parameter difference, and the greater the exceeding range, the greater the correction coefficient. Based on the adjustment amplitude correction coefficient, the original parameter adjustment amplitude is adjusted, for example, the original adjustment amplitude is multiplied by the correction coefficient to obtain a new adjustment amplitude. The first adjusted interaction matching parameter is adjusted for the second time using the new adjustment amplitude, the adjustment direction is unchanged, and the second adjusted interaction matching parameter is obtained. Based on the second adjusted interaction matching parameter, the candidate combination is re-screened, and the second adjusted pest and disease identification result is obtained.

[0121] Step S147: Repeat the above steps until the adjusted interaction matching parameter is within the target interaction matching parameter range, and the interaction matching parameter at this time is taken as the optimized interaction matching parameter, and the corresponding pest and disease identification result is taken as the corrected pest and disease identification result.

[0122] For example, step S1471: record the current adjustment number and the corresponding interaction matching parameter, pest and disease type-stage combination; normalize the intra-regional feature matching degree, cross-regional feature synergy degree and stage feature fitting degree in the current adjusted interaction matching parameter; determine whether the normalized intra-regional feature matching degree, the normalized cross-regional feature synergy degree and the normalized stage feature fitting degree are within the normalized target interaction matching parameter range, wherein the normalized target interaction matching parameter range refers to the average value of the corresponding normalized parameters in the historical corrected pest and disease identification result plus or minus a preset fluctuation range.

[0123] In this embodiment, the current adjusted interaction matching parameter is normalized, and the normalization method is the same as step S1372, and is normalized to between 0 and 1. At the same time, the target interaction matching parameter range is also normalized to obtain the normalized target interaction matching parameter range. The normalized target interaction matching parameter range is obtained by calculating the average value of the corresponding normalized parameters in the historical corrected pest and disease identification result, and then adding or subtracting a preset fluctuation range, and the preset fluctuation range is determined according to the dispersion degree of the historical data. It is determined whether the normalized intra-regional feature matching degree, the cross-regional feature synergy degree and the stage feature fitting degree are within the normalized target interaction matching parameter range.

[0124] Step S1472: If the intra-region feature matching degree, the cross-region feature coordination degree, and the stage feature matching degree are all within the target interaction matching parameter range, stop adjusting, mark the current interaction matching parameter as the optimized interaction matching parameter, and mark the corresponding pest and disease recognition result as the corrected pest and disease recognition result.

[0125] If all the normalized interaction matching parameters are within the normalized target interaction matching parameter range, it indicates that the current interaction matching parameter is relatively accurate and can reflect the real pest and disease situation of the plant to be identified. At this time, the adjusting process is stopped, the current interaction matching parameter is marked as the optimized interaction matching parameter, and the corresponding pest and disease recognition result is marked as the corrected pest and disease recognition result. Compared with the preliminary recognition result, the corrected pest and disease recognition result has higher accuracy and reliability.

[0126] Step S1473: If at least one of the intra-region feature matching degree, the cross-region feature coordination degree, and the stage feature matching degree is not within the target interaction matching parameter range, analyze the type of the parameter that is not within the target interaction matching parameter range. If the type of the parameter that is not within the target interaction matching parameter range is the intra-region feature matching degree, find the adjustment amplitude correction coefficient of the intra-region feature matching degree under the feature attribute combination in the mapping relationship table, and adjust the parameter adjustment amplitude of the intra-region feature matching degree based on the adjustment amplitude correction coefficient to adjust the current intra-region feature matching degree again.

[0127] In this embodiment, for the type of parameter that is not within the range, the corresponding adjustment amplitude correction coefficient under the feature attribute combination is found in the mapping relationship table. The adjustment amplitude correction coefficient is determined according to the adjustment experience of the type of parameter in historical cases, and is used to further optimize the adjustment amplitude. For example, if the intra-region feature matching degree is not within the range and the current value is lower than the lower limit of the range, the adjustment amplitude needs to be increased, and the original adjustment amplitude is multiplied by a correction coefficient greater than 1; if the current value is higher than the upper limit of the range, the adjustment amplitude needs to be reduced, and the original adjustment amplitude is multiplied by a correction coefficient less than 1. Based on the adjusted adjustment amplitude, the current intra-region feature matching degree is adjusted again.

[0128] Step S1474: If the type of parameter that is not within the target interaction matching parameter range is the cross-region feature coordination degree, find the adjustment amplitude correction coefficient of the cross-region feature coordination degree under the feature attribute combination in the mapping relationship table, and adjust the parameter adjustment amplitude of the cross-region feature coordination degree based on the adjustment amplitude correction coefficient; if the type of parameter that is not within the target interaction matching parameter range is the stage feature matching degree, find the adjustment amplitude correction coefficient of the stage feature matching degree under the feature attribute combination in the mapping relationship table, and adjust the parameter adjustment amplitude of the stage feature matching degree based on the adjustment amplitude correction coefficient.

[0129] For the parameter type of cross-region feature coordination degree or stage feature fit degree not in the range, the processing method is similar to the intra-region feature matching degree. The corresponding adjustment amplitude correction coefficient is looked up in the mapping relationship table, and the parameter adjustment amplitude of the cross-region feature coordination degree or the stage feature fit degree is adjusted according to the deviation direction and size of the current value of the parameter and the target range. For example, if the current value of the cross-region feature coordination degree is lower than the lower limit of the target range, the adjustment amplitude correction coefficient is greater than 1, the original adjustment amplitude is multiplied by the coefficient, and the cross-region feature coordination degree is adjusted again.

[0130] Step S1475: Based on the adjusted interaction matching parameters, the candidate combinations are re-screened to determine a new pest and disease identification result; the new adjustment number, the new interaction matching parameters and the new pest and disease identification result are recorded, and it is returned to step S1471 to determine whether the new interaction matching parameters are in the target interaction matching parameter range.

[0131] In this embodiment, the new adjustment number, the new interaction matching parameters and the new pest and disease identification result can be recorded, and then returned to step S1471 to determine whether the new interaction matching parameters are in the target interaction matching parameter range. The above adjustment, screening and determination steps are repeatedly executed until all the interaction matching parameters are in the target range or the maximum adjustment number is reached.

[0132] Step S1476: The above steps are continuously executed until all the interaction matching parameters are in the target interaction matching parameter range or the adjustment number reaches the preset maximum number; if the adjustment number reaches the preset maximum number and some parameters are not in the target interaction matching parameter range, the adjusted intra-region feature matching degree, cross-region feature coordination degree and stage feature fit degree are normalized; the absolute difference between each normalized parameter and the median value of the normalized target interaction matching parameter range is calculated; the interaction matching parameter with the smallest sum of absolute differences is selected as the optimized interaction matching parameter, and the corresponding pest and disease identification result is selected as the corrected pest and disease identification result.

[0133] The adjustment process is continuously executed, and if some parameters are not in the target interaction matching parameter range when the adjustment number reaches the preset maximum number, the adjustment is stopped and a compromise solution is adopted. The adjusted interaction matching parameters are normalized, and the absolute difference between each normalized parameter and the median value of the normalized target interaction matching parameter range is calculated. The median value of the target interaction matching parameter range is the arithmetic mean of the minimum value and the maximum value. The absolute differences of the parameters are added to obtain the total absolute difference. The interaction matching parameter with the smallest total absolute difference is selected as the optimized interaction matching parameter, and the corresponding pest and disease identification result is selected as the corrected pest and disease identification result. In the above method, when the target range cannot be completely met, the parameter combination with the smallest deviation from the median value of the target range is selected to improve the accuracy of the identification result as much as possible.

[0134] Step S150: generating a pest and disease identification report containing pest and disease type identification, development stage evaluation and feature matching description based on the corrected pest and disease identification result and the optimized interactive matching parameters, and sending the pest and disease identification report to the target terminal.

[0135] In this embodiment, the pest and disease identification report contains multiple parts: pest and disease type identification, which clearly indicates the type of pest and disease suffered by the plant to be identified, such as tomato early blight; development stage evaluation, which determines the current development stage of the pest and disease, such as the spreading stage; feature matching description, which describes the situation of the optimized interactive matching parameters in detail, including the specific values of the regional feature matching degree, the cross-regional feature coordination degree and the stage feature fitting degree, and the comparison of these parameters with the target interactive matching parameter range, which explains the matching details of the visual features of each growth part with the core visual feature unit, such as the matching situation of the color feature unit of the leaf area with the core color feature unit of the tomato early blight spreading stage. The report can also contain prevention and control suggestions for the pest and disease, which provide targeted prevention and control methods and measures based on the type and development stage of the pest and disease. The generated pest and disease identification report is stored in text format or PDF format and sent to the target terminal, such as the grower's mobile phone, computer or agricultural management platform, through the network communication module. During the sending process, encrypted transmission is used to ensure the security of the report content. After receiving the report, the target terminal can display or print it for relevant personnel to view and reference, so as to take effective pest and disease control measures in a timely manner.

[0136] Based on the same inventive concept, please refer to Figure 2 , which shows the structure schematic block diagram of the AI vision model-based pest and disease identification system 100 provided by the embodiments of the present application for executing the above-mentioned AI vision model-based pest and disease identification method. The AI vision model-based pest and disease identification system 100 can include a communication unit 110, a machine-readable storage medium 120 and a processor 130.

[0137] In this embodiment, the machine-readable storage medium 120 and the processor 130 are both located in the AI vision model-based pest and disease identification system 100 and are separately arranged. However, it should be understood that the machine-readable storage medium 120 can also be independent of the AI vision model-based pest and disease identification system 100 and can be accessed by the processor 130 through a bus interface. Alternatively, the machine-readable storage medium 120 can also be integrated into the processor 130 and can communicate and interact with external systems through the communication unit 110.

[0138] The processor 130 is the control center of the AI vision model-based pest and disease identification system 100, connects various parts of the AI vision model-based pest and disease identification system 100 through various interfaces and lines, executes various functions of the AI vision model-based pest and disease identification system 100 and processes data by running or executing software programs and / or modules stored in the machine readable storage medium 120 and calling data stored in the machine readable storage medium 120, thereby monitoring the AI vision model-based pest and disease identification system 100 as a whole. Optionally, the processor 130 can include one or more processing cores; for example, the processor 130 can integrate an application processor and a modem processor, wherein the application processor mainly processes operating systems, user interfaces, and application programs, etc., and the modem processor mainly processes wireless communication. It can be understood that the above-mentioned modem processor can also not be integrated into the processor. Among them, the machine readable storage medium 120 is used to store machine executable instructions for executing the scheme of the present application, and the processor 130 is used to execute the machine executable instructions stored in the machine readable storage medium 120 to realize the AI vision model-based pest and disease identification method provided by the foregoing method embodiment.

[0139] It should be noted that, in order to simplify the expression of the present disclosure and to help the understanding of one or more embodiments of the present application, in the foregoing description of the embodiments of the present application, various features are sometimes combined into one embodiment, figure or description thereof.

Claims

1.A method for pest and disease identification based on an AI vision model, characterized in that, The method comprises: acquiring a plant visual data set to be identified, the plant visual data set to be identified containing visual collection data of different growth parts of a plant to be identified at different collection times; constructing a pest and disease feature dynamic correlation chain, the pest and disease feature dynamic correlation chain containing core visual feature units of multiple pests and diseases at different development stages, correlation strength parameters between the core visual feature units, and cross-stage feature evolution path information; calling an AI vision model to perform multi-dimensional feature interaction processing on the plant visual data set based on the pest and disease feature dynamic correlation chain, to generate a feature interaction result corresponding to the plant visual data set to be identified and a preliminary pest and disease identification result, the feature interaction result containing interaction matching parameters of visual features of each growth part and core visual feature units in the correlation chain; based on parameter adjustment rules in historical pest and disease identification feedback data, performing multiple rounds of iterative optimization processing on the interaction matching parameters in the feature interaction result to obtain optimized interaction matching parameters and a corrected pest and disease identification result; based on the corrected pest and disease identification result and the optimized interaction matching parameters, generating a pest and disease identification report containing pest and disease type identification, development stage evaluation, and feature matching instructions, and sending the pest and disease identification report to a target terminal. 2.The AI vision model-based pest and disease identification method of claim 1, wherein, The construction of the pest and disease feature dynamic correlation chain comprises: acquiring full-cycle growth monitoring data of multiple pests and diseases, the full-cycle growth monitoring data containing continuous visual collection data and corresponding growth environment parameter records of each pest and disease from an initial occurrence stage, a spread stage, to a severe damage stage; extracting visual feature units of each development stage from the full-cycle growth monitoring data of each pest and disease to obtain a multi-stage feature unit set corresponding to each pest and disease, the visual feature unit containing a color feature unit, a morphology feature unit, a texture feature unit, and a lesion area distribution feature unit; performing core feature screening on the multi-stage feature unit set of each pest and disease, retaining visual feature units that appear more frequently than other stages and can uniquely identify the stage as core visual feature units in any one development stage of the pest and disease, to form a core visual feature unit subset of each stage of each pest and disease; analyzing the correlation between core visual feature units of adjacent development stages of each pest and disease, and calculating the co-occurrence frequency, morphological similarity, and evolution dependency of the core visual feature units of adjacent stages; normalize the co-occurrence frequency, morphological similarity, and evolution dependency to the same numerical interval, and determine the correlation strength parameters between the core visual feature units based on the weighted calculation results of the three normalized parameters; tracking the change process of the core visual feature units of each pest and disease from the initial occurrence stage to the severe damage stage, recording the order of appearance, morphological transition mode, and duration of each core visual feature unit, to form cross-stage feature evolution path information of each pest and disease; integrate the core visual feature unit subset of each stage, the corresponding correlation strength parameters, and the cross-stage feature evolution path information of each pest and disease to form a feature correlation sub-chain of a single pest and disease; Collect all pest and disease characteristics associated sub-chain, sort by pest and disease type, mark similar core visual feature units between different pest and disease types, and supplement cross-pest and disease type feature association reference information; Based on the cross-pest and disease type feature association reference information, adjust the association strength parameters of the core visual feature units in each pest and disease characteristic association sub-chain, so that the feature association logic between different pest and disease characteristic association sub-chains remains consistent; Integrate all adjusted pest and disease characteristic association sub-chains to build a pest and disease characteristic dynamic association chain containing multiple pest and disease characteristic association information, and store it in the feature association database of the AI vision model. 3.The AI vision model-based pest and disease identification method of claim 2, wherein, The visual feature units of each development stage are extracted from the whole cycle growth monitoring data of each pest and disease, and a multi-stage feature unit set corresponding to each pest and disease is obtained, including: The whole cycle growth monitoring data of each pest and disease is segmented in time sequence, the segmentation interval is determined according to the growth rate and morphological change amplitude of the pest and disease, and the continuous monitoring data is divided into multiple time periods, each time period corresponding to a development stage; The visual collection data in each time period is preprocessed, the color feature of the preprocessed lesion area visual data is extracted, the visual data is converted from RGB color space to HSV color space by color space conversion method, and the hue mean, saturation variance and brightness range of the lesion area are extracted as color feature units; The morphological feature of the preprocessed lesion area visual data is extracted, the contour information of the lesion area is obtained by edge detection algorithm, and the perimeter, area and circularity of the contour are calculated as morphological feature units; The texture feature of the preprocessed lesion area visual data is extracted, and the energy, entropy and contrast of the lesion area are calculated as texture feature units by using gray level co-occurrence matrix method; The lesion area distribution feature of the preprocessed lesion area visual data is extracted, and the position coordinates, distribution density and distance of adjacent lesion areas of the lesion area in the whole plant visual data are marked as lesion area distribution feature units; The color feature units, morphological feature units, texture feature units and lesion area distribution feature units extracted in each time period are integrated to form a stage feature unit group corresponding to the time period; The stage feature unit groups corresponding to all time periods of each pest and disease are arranged in time sequence to form a multi-stage feature unit set corresponding to each pest and disease. 4.The AI vision model-based pest and disease identification method of claim 2, wherein, The change process of the core visual feature units of each pest and disease from the initial occurrence stage to the serious damage stage is tracked, the order of appearance, morphological transition mode and duration of each core visual feature unit are recorded, and the cross-stage feature evolution path information of each pest and disease is formed, including: Each development stage in the whole cycle growth monitoring data of each pest and disease is assigned a time label, and the starting time and ending time of each development stage are marked; The appearance time of each core visual feature unit is extracted from the core visual feature unit set of each stage of each pest and disease, and the appearance time refers to the time when the core visual feature unit first appears in the visual collection data; sequencing the core visual feature units of each pest and disease in chronological order to form a core visual feature unit time sequence; analyzing the morphological differences between two adjacent core visual feature units in the core visual feature unit time sequence to determine the morphological transition mode therebetween, the morphological transition mode including gradual transition, abrupt transition, and superimposed transition, wherein gradual transition refers to gradual change in morphological parameters, abrupt transition refers to sudden change in morphological parameters, and superimposed transition refers to the addition of morphological parameters to the previous core visual feature unit by the subsequent core visual feature unit; calculating the time span of each core visual feature unit from the appearance time to the disappearance time, taking the time span as the duration of the core visual feature unit, and if the core visual feature unit lasts until the end of the severe damage stage, the duration of the core visual feature unit is the time span from the appearance time to the end time of the severe damage stage; integrating the core visual feature unit time sequence, the morphological transition mode of each adjacent core visual feature unit, and the duration of each core visual feature unit to form the basic information of the cross-stage feature evolution path; based on the basic information of the cross-stage feature evolution path, drawing the cross-stage feature evolution path graph of each pest and disease, and labeling the position, appearance time, duration, and morphological transition mode of each core visual feature unit; supplementing the change trend of the corresponding correlation strength parameters of each core visual feature unit in the cross-stage feature evolution path graph, and labeling the change curve of the correlation strength parameters over time; integrating the cross-stage feature evolution path graph and the corresponding textual description information to form the cross-stage feature evolution path information of each pest and disease. 5.The AI vision model-based pest and disease identification method of claim 1, wherein, The calling of the AI vision model performs multi-dimensional feature interaction processing on the plant vision data set based on the dynamic correlation chain of pest and disease characteristics, generates the feature interaction result and the preliminary pest and disease recognition result corresponding to the to-be-identified plant vision data set, including: inputting the to-be-identified plant vision data set into the feature analysis module of the AI vision model, performing zoning processing on the vision collection data according to the growth parts to obtain leaf area vision data, stem area vision data, and fruit area vision data of the to-be-identified plant; performing feature extraction on each area vision data to obtain a leaf area visual feature unit set, a stem area visual feature unit set, and a fruit area visual feature unit set, and each area feature unit set includes color feature units, morphological feature units, and texture feature units; calling the dynamic correlation chain of pest and disease characteristics from the feature correlation database of the AI vision model, and preliminarily matching the leaf area visual feature unit set, the stem area visual feature unit set, and the fruit area visual feature unit set with each stage core visual feature unit in the correlation chain; according to the preliminary matching result, screening out a core visual feature unit group with a matching degree exceeding a preset matching degree threshold from each area visual feature unit set, determining the pest and disease type and development stage corresponding to each group of core visual feature units, and forming a candidate combination list; The multi-dimensional feature interaction module of the AI vision model is called to construct an interaction correlation model between the visual feature units in each region based on the pest and disease information in the candidate combination list, and the interaction correlation model is used to analyze the collaborative matching relationship between the visual feature units in different regions. The interaction matching parameters of each regional visual feature unit and the corresponding core visual feature unit group are calculated through the interaction correlation model, and the interaction matching parameters include the intra-regional feature matching degree, the cross-regional feature collaboration degree, and the stage feature fit degree. The interaction matching parameters of each region are integrated to form a feature interaction result corresponding to the to-be-identified plant visual data set. Based on the distribution of the interaction matching parameters in the feature interaction result, the intra-regional feature matching degree, the cross-regional feature collaboration degree, and the stage feature fit degree are normalized to the same numerical interval, respectively; the candidate combination with the highest normalized intra-regional feature matching degree, the optimal normalized cross-regional feature collaboration degree, and the best normalized stage feature fit degree is selected from the candidate combination list to determine the pest and disease type and development stage of the to-be-identified plant. A preliminary pest and disease identification result including the determined pest and disease type identifier, development stage identifier, and corresponding interaction matching parameters is generated. 6.The AI vision model-based pest and disease identification method according to claim 5, characterized in that, The multi-dimensional feature interaction module of the AI vision model is called to construct an interaction correlation model between the visual feature units in each region based on the pest and disease information in the candidate combination list, and the interaction correlation model is used to analyze the collaborative matching relationship between the visual feature units in different regions. A candidate combination is selected from the candidate combination list, and the pest and disease type information and development stage information corresponding to the candidate combination are extracted. According to the pest and disease type information and development stage information, the leaf core visual feature unit, the stem core visual feature unit, and the fruit core visual feature unit of the pest and disease at the corresponding stage are retrieved from the pest and disease feature dynamic correlation chain. The correlation weights between the leaf core visual feature unit and the stem core visual feature unit, the correlation weights between the leaf core visual feature unit and the fruit core visual feature unit, and the correlation weights between the stem core visual feature unit and the fruit core visual feature unit are determined, and the correlation weights are set based on the cross-regional lesion propagation rule of the pest and disease at the corresponding stage. The input layer of the interaction correlation model is constructed, and the leaf region visual feature unit set, the stem region visual feature unit set, and the fruit region visual feature unit set of the to-be-identified plant are taken as input vectors, respectively. The hidden layer of the interaction correlation model is constructed, and three feature processing sub-modules are set, corresponding to the leaf region, the stem region, and the fruit region, respectively. Each feature processing sub-module performs feature mapping processing on the input regional feature unit set to obtain a regional feature mapping vector. A cross-region interaction module is added in the hidden layer, and the regional feature mapping vectors of the three regions are weighted and fused based on the preset correlation weights to calculate the feature collaboration values of the leaf region and the stem region, the feature collaboration values of the leaf region and the fruit region, and the feature collaboration values of the stem region and the fruit region. The output layer of the interactive correlation model is constructed, and the intra-regional feature matching degree and the cross-regional feature synergy value are taken as output parameters, wherein the intra-regional feature matching degree refers to the matching degree of the regional feature mapping vector of each region and the corresponding core visual feature unit, and the cross-regional feature synergy value refers to the average of the three cross-regional synergy values; Based on historical interactive case data, the hidden layer parameters and cross-regional correlation weights of the interactive correlation model are adjusted, so that the intra-regional feature matching degree and the cross-regional feature synergy value output by the model can accurately reflect the synergistic matching relationship between the visual feature units of different regions and the core visual feature unit; According to the above steps, an interactive correlation model corresponding to each candidate combination in the candidate combination list is constructed to form a multi-model parallel processing framework. 7.The AI vision model-based pest and disease identification method according to claim 5, characterized in that, Based on the distribution of the interactive matching parameters in the feature interaction result, the candidate combination with the highest intra-regional feature matching degree, the optimal cross-regional feature synergy degree, and the best stage feature matching degree is selected from the candidate combination list to determine the type and development stage of the plant to be identified, which includes: Extract the interactive matching parameters in the feature interaction result, and according to each candidate combination in the candidate combination list, sort the corresponding intra-regional feature matching degree, cross-regional feature synergy degree, and stage feature matching degree, wherein the intra-regional feature matching degree refers to the average of the matching degrees of the leaf region, stem region, and fruit region, the cross-regional feature synergy degree refers to the average of the three cross-regional synergy values, and the stage feature matching degree refers to the overall matching degree with the corresponding stage core feature in the correlation chain; The intra-regional feature matching degree, cross-regional feature synergy degree, and stage feature matching degree of each candidate combination are normalized to the same numerical interval; weight coefficients are set for the normalized intra-regional feature matching degree, normalized cross-regional feature synergy degree, and normalized stage feature matching degree, and the weight coefficients are set according to the priority of pest and disease identification, wherein the weight coefficient of the normalized intra-regional feature matching degree is the highest; based on the weight coefficients, the three normalized parameters of each candidate combination are weighted and calculated to obtain the comprehensive matching score of each candidate combination; The candidate combinations in the candidate combination list are sorted in descending order of the comprehensive matching score to obtain a sorted candidate combination list; The top K candidate combinations with the highest comprehensive matching scores in the sorted candidate combination list are selected as target candidate combinations; The common disease conditions of the target candidate combination corresponding to the type of pest and disease on the plant variety to be identified are analyzed, and if the difference between the disease frequency of any target candidate combination corresponding to the type of pest and disease on the plant variety and the disease frequency of the other two target candidate combinations corresponding to the type of pest and disease on the plant variety exceeds the preset frequency difference range, the target candidate combination with the higher disease frequency is preferred; If the disease frequency difference of the target candidate combination corresponding to the type of pest and disease on the plant variety is within the preset frequency difference range, the stage feature matching degree of each target candidate combination is further analyzed, and the target candidate combination with the highest stage feature matching degree is selected. If the difference of the stage feature fitting degrees of each target candidate combination is within the preset fitting degree difference range, the cross-region feature coordination degrees of each target candidate combination are analyzed, and the target candidate combination with the highest cross-region feature coordination degree is selected; The type and development stage of the disease and pest of the selected target candidate combination are determined as the type and development stage of the disease and pest of the plant to be identified. 8.The AI vision model-based pest and disease identification method of claim 1, wherein, The parameter adjustment rule based on the historical disease and pest identification feedback data is used to perform multiple rounds of iterative optimization processing on the interaction matching parameters in the feature interaction result, to obtain the optimized interaction matching parameters and the corrected disease and pest identification result, which includes: Obtain historical disease and pest identification feedback data, which includes a set of historical visual data of plants to be identified, historical feature interaction results, historical preliminary disease and pest identification results, and historical corrected disease and pest identification results; Extract the historical interaction matching parameters corresponding to the historical feature interaction results and the target interaction matching parameters corresponding to the historical corrected disease and pest identification results from the historical disease and pest identification feedback data; Calculate the parameter difference between the historical interaction matching parameters and the target interaction matching parameters in each historical case, and classify the parameter difference according to the feature attributes of the set of historical visual data of plants to be identified, including plant variety attributes, collection season attributes, and lesion region attributes; normalize the parameter difference under each type of feature attribute, analyze the trend of the parameter difference under each type of feature attribute, determine the parameter adjustment direction and parameter adjustment amplitude corresponding to different feature attribute combinations, and form a mapping relationship table; Extract the plant variety attributes, collection season attributes, and lesion region attributes of the set of visual data of plants to be identified, and find the corresponding parameter adjustment direction and parameter adjustment amplitude in the mapping relationship table; According to the found parameter adjustment direction and parameter adjustment amplitude, the intra-region feature matching degree, the cross-region feature coordination degree, and the stage feature fitting degree in the feature interaction result are adjusted for the first time, to obtain the first adjusted interaction matching parameters; Based on the first adjusted interaction matching parameters, reselect the candidate combinations to determine the type and development stage of the disease and pest of the plant to be identified, and obtain the first adjusted disease and pest identification result; Extract the historical corrected disease and pest identification results of the same type and stage as the first adjusted disease and pest identification result from the historical feedback data, and obtain the corresponding target interaction matching parameter range; Determine whether the first adjusted interaction matching parameters are within the target interaction matching parameter range, if not, further adjust the parameter adjustment amplitude according to the mapping relationship table; Based on the adjusted parameter adjustment amplitude, the first adjusted interaction matching parameters are adjusted for the second time to obtain the second adjusted interaction matching parameters and the second adjusted disease and pest identification result; Repeat the above steps until the adjusted interaction matching parameters are within the target interaction matching parameter range, and the interaction matching parameters at this time are used as the optimized interaction matching parameters, and the corresponding disease and pest identification result is used as the corrected disease and pest identification result. 9.The AI vision model-based pest and disease identification method according to claim 8, characterized in that, The analysis of the change trend of the parameter difference value under each type of characteristic attribute determines the parameter adjustment direction and the parameter adjustment amplitude corresponding to different characteristic attribute combinations, and forms a mapping relationship table, including: Divide the characteristic attributes of the historical to-be-identified plant visual data set into multiple variety categories according to plant variety attributes, multiple season categories according to collection season attributes, and multiple lesion area categories according to lesion area attributes; For each combination of variety categories, season categories, and lesion area categories, all historical cases under the combination are screened out to form a characteristic attribute combination case set; Calculate the parameter difference value mean of the historical interaction matching parameters and the target interaction matching parameters in each characteristic attribute combination case set, and the parameter difference value mean includes the intra-regional feature matching degree difference value mean, the cross-regional feature synergy degree difference value mean, and the stage feature fit degree difference value mean; Analyze the positive and negative nature of the parameter difference value mean. If the intra-regional feature matching degree difference value mean is positive, it is determined that the parameter adjustment direction of the intra-regional feature matching degree under the characteristic attribute combination is to be reduced. If it is negative, it is determined that the parameter adjustment direction of the intra-regional feature matching degree under the characteristic attribute combination is to be increased. The parameter adjustment directions of the cross-regional feature synergy degree and the stage feature fit degree are determined according to the same logic. Calculate the standard deviation of the parameter difference value in each characteristic attribute combination case set, and determine the parameter adjustment amplitude level according to the size of the standard deviation. The larger the standard deviation, the higher the parameter adjustment amplitude level, and the higher the parameter adjustment amplitude level, the larger the corresponding parameter adjustment amplitude. Assign the intra-regional feature matching degree parameter adjustment direction, the cross-regional feature synergy degree parameter adjustment direction, and the stage feature fit degree parameter adjustment direction corresponding to each characteristic attribute combination, as well as the corresponding parameter adjustment amplitude. Organize the characteristic attribute combinations and the corresponding parameter adjustment directions and parameter adjustment amplitudes into a table form to form a mapping relationship table, wherein the characteristic attribute combinations are composed of variety categories, season categories, and lesion area categories. Supplement and improve the data in the mapping relationship table. If the number of historical cases corresponding to any characteristic attribute combination is less than the preset case number threshold, refer to the parameter adjustment direction and parameter adjustment amplitude of similar characteristic attribute combinations, and fine-tune according to the particularity of the characteristic attribute combination to make the mapping relationship table cover all possible characteristic attribute combinations. 10.A pest and disease identification system based on an AI vision model, characterized in that, It includes: A processor; A machine-readable storage medium for storing machine-executable instructions of the processor; Wherein, the processor is configured to execute the machine-executable instructions to perform the plant disease and pest identification method based on the AI vision model in any one of claims 1 to 9.

Images