A method and system for diagnosing nutrient stress in fruit trees

Abstract

The present invention discloses a method and system for diagnosing nutrient stress in fruit trees, which relate to the field of image processing. The method comprises acquiring images of key fertilization periods during the annual growth period of an orchard, namely budding, flowering, the initial / final stages of fruit expansion, and after harvest, to form a time-series RGB and hyperspectral image set including a ground-based unmanned aerial vehicle (UAV) three-dimensional platform; extracting corresponding features of the images on each platform, thereby acquiring phenotypic information of the fruit trees based on the three-dimensional time-series features and calculating the fertilizer requirement under the expected yield; and analyzing the deviation of the fruit tree nutrients at that stage from the most recent growth conditions through the fruit tree phenotypic information reflected in the fruit tree phenotypic information at different periods, thereby achieving reverse diagnosis of nutrient stress in the fruit trees.

Description

Technical Field

[0001] The present invention relates to the field of image processing, and in particular to a method and system for diagnosing nutrient stress of fruit trees. Background Art

[0002] Traditional diagnostic methods for nutrient stress in plants rely primarily on visual observation and sampling and testing. Observation is primarily based on specific visible symptoms of nutrient stress in plants. This method requires extensive experience, is relatively slow to successfully determine, and can lead to significant errors when multiple stresses occur simultaneously. Sampling and testing involves laboratory testing of nutrient content and concentration in samples from various organs of fruit trees within an orchard. While this method is accurate, it is cumbersome and laborious, making it difficult to achieve accurate diagnosis on a large scale.

[0003] In recent years, the availability of instruments and equipment to assist in measuring nutrient stress in vegetation has increased significantly. For orchards, soil nutrient sensor equipment is the most widely used, while methods for measuring nutrients in aboveground parts such as leaves and canopies primarily rely on spectral-based nutrient stress diagnosis. Nutrient stress in vegetation can cause changes in leaf color, thickness, and even morphology, which in turn influence changes in corresponding spectral characteristics. Therefore, collecting spectral data from leaves or canopies can effectively assess nutrient stress. However, current research is mostly conducted at a time point, often considering nutrients in leaves, roots, and soil separately. However, fruit trees are perennial plants, and nutrient transfer and accumulation is a long-term process. To achieve efficient nutrient stress diagnosis and thus accurately guide variable-rate fertilization, we need to comprehensively consider changes in various organs throughout the fruit tree's life cycle to make the most accurate assessment.

[0004] Currently, modern information technologies such as remote sensing networks, the Internet of Things, cloud computing, big data, and artificial intelligence have been deeply applied to all aspects of agricultural production. However, the informatization of fruit production still lacks a multi-scale, coordinated observation system for fruit trees and the orchard growth environment. This results in a lack of three-dimensional information support for fruit tree nutrient stress and even production management. Orchard stress diagnosis and decision-making are mainly based on experience, resulting in insufficient timeliness of orchard nutrient stress diagnosis and unstable decision-making reliability. Therefore, it is urgent to establish a three-dimensional dynamic observation system for fruit tree phenotypic information and, based on this, to construct a spatiotemporal dynamic diagnosis method system for orchard nutrient stress. This system can achieve comprehensive multi-factor analysis while providing timely and effective information on orchard stress, improve China's orchard management level, and ultimately enhance the quality and yield of Chinese fruits. Summary of the Invention

[0005] The purpose of the present invention is to provide a method and system for diagnosing nutrient stress of fruit trees, which can realize reverse diagnosis of nutrient stress of fruit trees.

[0006] To achieve the above object, the present invention provides the following solutions:

[0007] A method for diagnosing nutrient stress in fruit trees, comprising:

[0008] Acquire images of individual fruit trees and orchard canopy during the key fertilization period of the orchard's annual growth period, and obtain a time-series RGB image set and a hyperspectral image set;

[0009] Using a ground platform, RGB images of individual fruit trees during the budding and flowering stages are acquired. A target detection network, integrating a convolutional neural network with an attention mechanism module, is used to determine the number of flower buds and leaf buds on a single fruit tree during the budding stage, as well as the number of flowers during the flowering stage.

[0010] The leaf-to-fruit ratio and flower-to-fruit ratio are obtained based on the number of flower buds and leaf buds of a single fruit tree. The fruit yield of a single fruit tree is estimated based on the leaf-to-fruit ratio and flower-to-fruit ratio. At the same time, the single-element fertilizer requirement of a single fruit tree under the expected yield is determined, and then the standard single-element fertilizer dosage for each fruit tree at each stage is determined.

[0011] Using drone RGB and drone hyperspectral images of budding, flowering, the early and late stages of fruit expansion, and post-harvest throughout the orchard's annual growth period, a neural network is used to determine fruit tree phenotypic information based on multi-feature fusion. The multi-features include: structured semantic features, temporal semantic features, and cross-modal high-level semantic features. The fruit tree phenotypic information includes: the location of individual fruit trees and the crown width of individual fruit trees.

[0012] The soil fertilizer content and leaf nutrients were determined based on the fruit tree phenotypic information and drone hyperspectral images taken during the corresponding period of the orchard's annual growth period.

[0013] The diagnosis results of nutrient stress of individual fruit trees during the key fertilization period are determined based on the standard application amount of single-element fertilizers at different stages of individual fruit trees, the single-element fertilizer content in the soil and the nutrients in the leaves; the key fertilization periods include: budding, flowering, the early / late stages of fruit expansion and after harvest.

[0014] Optionally, the acquisition of images of individual fruit trees and orchard canopy images during a key fertilization period during the annual growth period of the orchard to obtain a time-series RGB image set and a hyperspectral image set specifically includes:

[0015] Integrate high-definition digital cameras and hyperspectral cameras on ground and UAV platforms to build a three-dimensional ground-UAV platform;

[0016] A ground-UAV three-dimensional platform was used to obtain images of individual fruit trees and orchard canopies during the key fertilization period of the orchard's annual growth period, and a time series RGB and hyperspectral image collection was obtained.

[0017] Optionally, the leaf-to-fruit ratio and the flower-to-fruit ratio are obtained based on the number of flower buds and the number of leaf buds corresponding to the number of fruit trees, and the fruit yield of the individual fruit trees is estimated based on the leaf-to-fruit ratio and the flower-to-fruit ratio. At the same time, the fertilizer requirement of the individual fruit trees under the expected yield is determined, and then the standard fertilizer amount of the individual fruit trees at each stage is determined, specifically including:

[0018] Based on the expected yield of single-element fertilizer requirements for a single fruit tree, the six fertilization periods (budding period, flowering period, fruit expansion period, before harvest, after harvest, and before freezing) were divided according to the proportion of each single fertilizer, and the division results were obtained.

[0019] The division results will be used as the standard amount of single-element fertilizer for each period.

[0020] Optionally, determining the soil fertilizer content and leaf nutrients based on the fruit tree phenotypic information and the drone hyperspectral images of the corresponding period during the annual growth period of the orchard specifically includes:

[0021] Establish a buffer zone with the corresponding fruit tree location point as the center and the outer boundary of the crown range as the boundary;

[0022] Image binarization was performed on the UAV hyperspectral images of the corresponding period during the annual growth period of the orchard;

[0023] Eliminate the crown range in the binarized image and extract the soil image information within the corresponding range of the individual fruit trees;

[0024] Based on the soil image information within the range corresponding to a single fruit tree, the effective spectral information of soil single fertilizer is determined by using the bivariate band combination method.

[0025] According to the effective spectral information of soil elemental fertilizer, the content of soil elemental fertilizer in the buffer zone of a single fruit tree was obtained by nonlinear inversion.

[0026] Optionally, determining the soil fertilizer content and leaf nutrients based on the fruit tree phenotypic information and the drone hyperspectral images of the corresponding period during the annual growth period of the orchard specifically includes:

[0027] The crown width of a single fruit tree is used as a benchmark, and the drone hyperspectral image within the crown width is used as the basic image for fruit tree nutrient analysis in the corresponding period.

[0028] Determine the nutrient content of the crown width of a single fruit tree during the critical fertilization period based on the basic image;

[0029] The nutrient content corresponding to the crown width of a single fruit tree during the critical fertilization period is used as the leaf nutrient of the single fruit tree in the current period.

[0030] Optionally, determining the diagnosis result of nutrient stress of a single fruit tree during the critical fertilization period based on the standard fertilizer application amount of single fertilizer at each period of the single fruit tree, the single fertilizer content in the soil, and the nutrients in the leaves specifically includes:

[0031] Based on the standard amount of single fertilizer applied to individual fruit trees at different stages, and the soil single fertilizer content and leaf nutrients at the corresponding stages, the diagnostic results of nutrient stress on individual fruit trees during the critical fertilization period were determined.

[0032] A fruit tree nutrient stress diagnosis system, comprising:

[0033] The image acquisition module is used to obtain images of individual fruit trees and orchard canopy layers during the key fertilization period of the orchard's annual growth period, and obtain a time-series RGB image set and a hyperspectral image set;

[0034] The target detection module is used to obtain RGB images of individual fruit trees during the budding and flowering stages using a ground platform; and to determine the number of flower buds and leaf buds on a single fruit tree during the budding stage, and the number of flowers during the flowering stage, respectively, using a target detection network that integrates a convolutional neural network and an attention mechanism module.

[0035] The module for determining the standard amount of single-element fertilizer is used to obtain the leaf-to-fruit ratio and the flower-to-fruit ratio based on the number of flower buds and leaf buds of a single fruit tree. The module also estimates the fruit yield of a single fruit tree based on the leaf-to-fruit ratio and the flower-to-fruit ratio. The module also determines the single-element fertilizer requirement for a single fruit tree under the expected yield, and further determines the standard amount of single-element fertilizer for each fruit tree at different stages.

[0036] The fruit tree phenotypic information determination module is used to determine the fruit tree phenotypic information through a neural network based on multi-feature fusion of drone RGB images and drone hyperspectral images of budding, flowering, the early / late stages of fruit expansion, and after harvest during the orchard's annual growth period. The multi-features include: structured semantic features, temporal semantic features, and cross-modal high-level semantic features. The fruit tree phenotypic information includes: the location points of individual fruit trees and the crown width of individual fruit trees.

[0037] A module for determining soil elemental fertilizer content and leaf nutrients, which is used to determine soil elemental fertilizer content and leaf nutrients based on fruit tree phenotypic information and drone hyperspectral images taken during the corresponding period of the orchard's annual growth period;

[0038] The stress diagnosis module is used to determine the nutrient stress diagnosis results of individual fruit trees during the key fertilization period based on the standard fertilizer application rate of single-element fertilizers in each period of individual fruit trees, the single-element fertilizer content in the soil and the nutrients in the leaves; the key fertilization period includes: budding, flowering, the early / late stage of fruit expansion and after harvest.

[0039] A storage medium stores computer program instructions thereon, which implement the method when the computer program instructions are executed by a processor.

[0040] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0041] The present invention provides a method and system for diagnosing nutrient stress in fruit trees. The system establishes a stereoscopic image set in a time series, which solves the problems of inaccurate recognition caused by spatial resolution and spatial scale range when currently relying solely on single-scale images for nutrient stress diagnosis. The system uses a target detection network that integrates a convolutional neural network (CNN) and a Transformer attention mechanism module to detect flower buds, leaf buds, and flower quantity of fruit trees, while taking into account the local and overall characteristics of the buds and flowers of individual fruit trees. This method is more accurate than current remote sensing-based estimation and detection results obtained using a single target detection network. A multi-sensor information fusion method is used to locate individual fruit trees and extract crown width information. This method can effectively integrate the advantageous information of multi-sensor data at the data level, and is faster and more accurate than relying solely on a single image for information extraction. The system uses stereoscopic images with time series features to diagnose nutrient stress in fruit trees. Compared with relying solely on time point data for diagnosis, it can integrate reliable information about the annual growth cycle of fruit trees, and can provide a scientific basis for timely regulation and precise variable fertilization in actual production. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0043] Figure 1 A schematic flow chart of a method for diagnosing nutrient stress in fruit trees provided by the present invention;

[0044] Figure 2 This is a schematic diagram of the overall process of a fruit tree nutrient stress diagnosis method provided by the present invention. DETAILED DESCRIPTION

[0045] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0046] The purpose of the present invention is to provide a method and system for diagnosing nutrient stress in fruit trees, which can analyze the deviation of the nutrient content of fruit trees at this stage from the most recent growth conditions through the phenotypic information of fruit trees at different stages, and realize reverse diagnosis of nutrient stress in fruit trees.

[0047] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0048] like Figure 1 and Figure 2 As shown, the present invention provides a method for diagnosing nutrient stress in fruit trees, comprising:

[0049] S101, acquiring images of individual fruit trees and orchard canopy layers during a key fertilization period during the annual growth period of the orchard, and obtaining a time-series RGB image set and a hyperspectral image set;

[0050] S101 specifically includes:

[0051] Integrate high-definition digital cameras and hyperspectral cameras on ground and UAV platforms to build a three-dimensional ground-UAV platform;

[0052] A ground-UAV three-dimensional platform was used to obtain images of individual fruit trees and orchard canopies during the key fertilization period of the orchard's annual growth period, and a time series RGB and hyperspectral image collection was obtained.

[0053] S102, using a ground platform to obtain RGB images of a single fruit tree during the budding and flowering stages; and using a target detection network to determine the number of flower buds (A) and leaf buds (B) of a single fruit tree during the budding stage, and the corresponding number of flowers (C) during the flowering stage; the target detection network is a target detection network that integrates a convolutional neural network and an attention mechanism module;

[0054] S103, obtaining a leaf-to-fruit ratio and a flower-to-fruit ratio based on the number of flower buds and leaf buds per fruit tree, estimating the fruit yield of the fruit tree based on the leaf-to-fruit ratio and the flower-to-fruit ratio, and simultaneously determining the single-element fertilizer requirement for the fruit tree at the expected yield, thereby determining the standard single-element fertilizer application rate for the fruit tree at various stages;

[0055] S103 specifically includes:

[0056] Based on the expected yield of single-element fertilizer requirements for a single fruit tree, the six fertilization periods (budding period, flowering period, fruit expansion period, before harvest, after harvest, and before freezing) were divided according to the proportion of each single fertilizer, and the division results were obtained.

[0057] The division results will be used as the standard amount of single-element fertilizer for each period.

[0058] S104, determining fruit tree phenotypic information through multi-feature fusion using a neural network based on drone RGB images and drone hyperspectral images of budding, flowering, early / late fruit expansion, and post-harvest throughout the orchard's annual growth period; the multi-features include: structured semantic features, temporal semantic features, and cross-modal high-level semantic features; the fruit tree phenotypic information includes: individual tree location points (D) and individual tree crown width (H);

[0059] S104 specifically includes:

[0060] By utilizing drone RGB and hyperspectral images of budding, flowering, the early / late stages of fruit expansion, and after harvest during the annual growth period of the orchard, and through neural network adaptive learning, adaptive graph convolutional network selection and learning, and homogeneous / heterogeneous ensemble learning, we can extract fruit tree phenotypic information such as the location point (D) and crown width (H) of individual fruit trees under the fusion mode of structured semantics, temporal semantics, and cross-modal high-level semantics.

[0061] S105, determining the soil fertilizer content and leaf nutrients based on the fruit tree phenotypic information and the drone hyperspectral images of the corresponding period during the annual growth period of the orchard;

[0062] S105 specifically includes:

[0063] Establish a buffer zone with the corresponding fruit tree positioning point (D) as the center and the outer boundary setting range (2m) of the crown width (H) as the boundary. The buffer zone radius is H+2;

[0064] Image binarization was performed on the UAV hyperspectral images of the corresponding period during the annual growth period of the orchard;

[0065] Eliminate the crown range in the binarized image and extract the soil image information within the corresponding range of the individual fruit trees;

[0066] Based on the soil image information within the range corresponding to a single fruit tree, the effective spectral information of soil single fertilizer is determined by using the bivariate band combination method.

[0067] According to the effective spectral information of soil elemental fertilizer, the content of soil elemental fertilizer in the buffer zone of a single fruit tree was obtained by nonlinear inversion.

[0068] The crown width of a single fruit tree is used as a benchmark, and the drone hyperspectral image within the crown width is used as the basic image for fruit tree nutrient analysis in the corresponding period.

[0069] Determine the nutrient content of the crown width of a single fruit tree during the critical fertilization period based on the basic image;

[0070] The nutrient content corresponding to the crown width of a single fruit tree during the critical fertilization period is used as the leaf nutrient of the single fruit tree in the current period.

[0071] S106, determining the diagnosis result of nutrient stress of the individual fruit tree in the key fertilization period according to the standard fertilizer amount of the single-element fertilizer in each period of the individual fruit tree, the single-element fertilizer content in the soil and the leaf nutrients; the key fertilization period includes: budding, flowering, the early / late stage of fruit expansion and after harvest.

[0072] S106 specifically includes:

[0073] Taking the standard amount of single-element fertilizer for individual fruit trees at different periods as the benchmark, and the soil single-element fertilizer content and leaf nutrients in the corresponding period as a reference, the three are subtracted to determine the nutrient stress diagnosis results of individual fruit trees during the critical fertilization period.

[0074] If the nutrient stress diagnosis result of a single fruit tree during the critical fertilization period is greater than 0, it means a lack of corresponding single-element fertilizer, and if it is less than 0, it means excessive fertilization of the corresponding single-element fertilizer, thus realizing nutrient stress diagnosis.

[0075] Corresponding to the above method, the present invention further provides a fruit tree nutrient stress diagnosis system, comprising:

[0076] The image acquisition module is used to obtain images of individual fruit trees and orchard canopy layers during the key fertilization period of the orchard's annual growth period, and obtain a time-series RGB image set and a hyperspectral image set;

[0077] The target detection module is used to obtain RGB images of individual fruit trees during the budding and flowering stages using a ground platform; and to determine the number of flower buds and leaf buds on a single fruit tree during the budding stage, and the number of flowers during the flowering stage, respectively, using a target detection network that integrates a convolutional neural network and an attention mechanism module.

[0078] The module for determining the standard amount of single-element fertilizer is used to obtain the leaf-to-fruit ratio and the flower-to-fruit ratio based on the number of flower buds and leaf buds of a single fruit tree. The module also estimates the fruit yield of a single fruit tree based on the leaf-to-fruit ratio and the flower-to-fruit ratio. The module also determines the single-element fertilizer requirement for a single fruit tree under the expected yield, and further determines the standard amount of single-element fertilizer for each fruit tree at different stages.

[0079] The fruit tree phenotypic information determination module is used to determine the fruit tree phenotypic information through a neural network based on multi-feature fusion of drone RGB images and drone hyperspectral images of budding, flowering, the early / late stages of fruit expansion, and after harvest during the orchard's annual growth period. The multi-features include: structured semantic features, temporal semantic features, and cross-modal high-level semantic features. The fruit tree phenotypic information includes: the location points of individual fruit trees and the crown width of individual fruit trees.

[0080] A module for determining soil elemental fertilizer content and leaf nutrients, which is used to determine soil elemental fertilizer content and leaf nutrients based on fruit tree phenotypic information and drone hyperspectral images taken during the corresponding period of the orchard's annual growth period;

[0081] The stress diagnosis module is used to determine the nutrient stress diagnosis results of individual fruit trees during the key fertilization period based on the standard fertilizer application rate of single-element fertilizers in each period of individual fruit trees, the single-element fertilizer content in the soil and the nutrients in the leaves; the key fertilization period includes: budding, flowering, the early / late stage of fruit expansion and after harvest.

[0082] Based on the above description, the technical solution of the present invention, or the portion that contributes to the prior art, or the portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes instructions for causing a computer device (such as a personal computer, server, or network device) to execute all or part of the steps of the various embodiments of the method of the present invention. The aforementioned computer storage medium includes various media capable of storing program code, such as a USB flash drive, a mobile hard drive, a read-only memory, a random access memory, a magnetic disk, or an optical disk.

[0083] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method description.

[0084] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only intended to help understand the method and core concept of the present invention. At the same time, those skilled in the art will find that the specific implementation methods and application scopes may vary based on the concept of the present invention. In summary, the contents of this specification should not be construed as limiting the present invention.

Claims

1. A method for diagnosing nutrient stress in fruit trees, characterized in that: include: Acquire images of individual fruit trees and orchard canopy during the key fertilization period of the orchard's annual growth period, and obtain a time-series RGB image set and a hyperspectral image set; Using a ground platform, RGB images of individual fruit trees during the budding and flowering stages are acquired. A target detection network, integrating a convolutional neural network with an attention mechanism module, is used to determine the number of flower buds and leaf buds on a single fruit tree during the budding stage, as well as the number of flowers during the flowering stage. The leaf-to-fruit ratio and flower-to-fruit ratio are obtained based on the number of flower buds and leaf buds of a single fruit tree. The fruit yield of a single fruit tree is estimated based on the leaf-to-fruit ratio and flower-to-fruit ratio. At the same time, the single-element fertilizer requirement of a single fruit tree under the expected yield is determined, and then the standard single-element fertilizer dosage for each fruit tree at each stage is determined. Using drone RGB and hyperspectral images of budding, flowering, the early and late stages of fruit expansion, and after harvest during the annual growth period of the orchard, we determined the fruit tree phenotypic information through multi-feature fusion using a neural network. The multiple features include: structural semantic features, temporal semantic features and cross-modal high-level semantic features; the fruit tree phenotypic information includes: the location point of a single fruit tree and the crown width of a single fruit tree; The soil fertilizer content and leaf nutrients were determined based on the fruit tree phenotypic information and drone hyperspectral images taken during the corresponding period of the orchard's annual growth period. The diagnosis results of nutrient stress of individual fruit trees during the key fertilization period are determined based on the standard application amount of single-element fertilizers at different stages of individual fruit trees, the single-element fertilizer content in the soil and the nutrients in the leaves; the key fertilization periods include: budding, flowering, the early / late stages of fruit expansion and after harvest.

2. A method for diagnosing nutrient stress in fruit trees according to claim 1, characterized in that: The method of obtaining images of individual fruit trees and orchard canopy layers during the key fertilization period of the orchard's annual growth period to obtain a time series of RGB image sets and hyperspectral image sets specifically includes: Integrate high-definition digital cameras and hyperspectral cameras on ground and UAV platforms to build a three-dimensional ground-UAV platform; A ground-UAV three-dimensional platform was used to obtain images of individual fruit trees and orchard canopies during the key fertilization period of the orchard's annual growth period, and a time series RGB and hyperspectral image collection was obtained.

3. A method for diagnosing nutrient stress in fruit trees according to claim 1, characterized in that: The method comprises the following steps: obtaining the leaf-to-fruit ratio and the flower-to-fruit ratio based on the number of flower buds and the number of leaf buds of a single fruit tree; estimating the fruit yield of a single fruit tree based on the leaf-to-fruit ratio and the flower-to-fruit ratio; and determining the single-element fertilizer requirement of a single fruit tree under the expected yield, thereby determining the standard single-element fertilizer application rate for each fruit tree at each stage. Based on the expected yield of single-element fertilizer requirements for a single fruit tree, the six fertilization periods (budding period, flowering period, fruit expansion period, before harvest, after harvest, and before freezing) were divided according to the proportion of each single fertilizer, and the division results were obtained. The division results will be used as the standard amount of single-element fertilizer for each period.

4. A method for diagnosing nutrient stress in fruit trees according to claim 1, characterized in that: The determination of soil elemental fertilizer content and leaf nutrients based on fruit tree phenotypic information and drone hyperspectral images taken during the corresponding period of the orchard's annual growth period specifically includes: Establish a buffer zone with the corresponding fruit tree location point as the center and the outer boundary of the crown range as the boundary; Image binarization was performed on the UAV hyperspectral images of the corresponding period during the annual growth period of the orchard; Eliminate the crown range in the binarized image and extract the soil image information within the corresponding range of the individual fruit trees; Based on the soil image information within the range corresponding to a single fruit tree, the effective spectral information of soil single fertilizer is determined by using the bivariate band combination method. According to the effective spectral information of soil elemental fertilizer, the content of soil elemental fertilizer in the buffer zone of a single fruit tree was obtained by nonlinear inversion.

5. A method for diagnosing nutrient stress in fruit trees according to claim 1, characterized in that: The determination of soil elemental fertilizer content and leaf nutrients based on fruit tree phenotypic information and drone hyperspectral images taken during the corresponding period of the orchard's annual growth period specifically includes: The crown width of a single fruit tree is used as a benchmark, and the drone hyperspectral image within the crown width is used as the basic image for fruit tree nutrient analysis in the corresponding period. Determine the nutrient content of the crown width of a single fruit tree during the critical fertilization period based on the basic image; The nutrient content corresponding to the crown width of a single fruit tree during the critical fertilization period is used as the leaf nutrient of the single fruit tree in the current period.

6. A method for diagnosing nutrient stress in fruit trees according to claim 1, characterized in that: The method of determining the nutrient stress diagnosis result of a single fruit tree during the critical fertilization period based on the standard fertilizer application amount of the single fertilizer at each period of the single fruit tree, the single fertilizer content in the soil, and the leaf nutrients specifically includes: Based on the standard amount of single fertilizer applied to individual fruit trees at different stages, and the soil single fertilizer content and leaf nutrients at the corresponding stages, the diagnostic results of nutrient stress on individual fruit trees during the critical fertilization period were determined.

7. A fruit tree nutrient stress diagnosis system, characterized in that: include: The image acquisition module is used to obtain images of individual fruit trees and orchard canopy layers during the key fertilization period of the orchard's annual growth period, and obtain a time-series RGB image set and a hyperspectral image set; The target detection module is used to obtain RGB images of individual fruit trees during the budding and flowering stages using a ground platform; and to determine the number of flower buds and leaf buds on a single fruit tree during the budding stage, and the number of flowers during the flowering stage, respectively, using a target detection network that integrates a convolutional neural network and an attention mechanism module. The module for determining the standard amount of single-element fertilizer is used to obtain the leaf-to-fruit ratio and the flower-to-fruit ratio based on the number of flower buds and leaf buds of a single fruit tree. The module also estimates the fruit yield of a single fruit tree based on the leaf-to-fruit ratio and the flower-to-fruit ratio. The module also determines the single-element fertilizer requirement for a single fruit tree under the expected yield, and further determines the standard amount of single-element fertilizer for each fruit tree at different stages. The fruit tree phenotypic information determination module is used to determine the fruit tree phenotypic information through multi-feature fusion using a neural network based on drone RGB and drone hyperspectral images of budding, flowering, the early and late stages of fruit expansion, and after harvest during the annual growth period of the orchard; The multiple features include: structural semantic features, temporal semantic features and cross-modal high-level semantic features; the fruit tree phenotypic information includes: the location point of a single fruit tree and the crown width of a single fruit tree; A module for determining soil elemental fertilizer content and leaf nutrients, which is used to determine soil elemental fertilizer content and leaf nutrients based on fruit tree phenotypic information and drone hyperspectral images taken during the corresponding period of the orchard's annual growth period; The stress diagnosis module is used to determine the nutrient stress diagnosis results of individual fruit trees during the key fertilization period based on the standard fertilizer application rate of single-element fertilizers in each period of individual fruit trees, the single-element fertilizer content in the soil and the nutrients in the leaves; the key fertilization period includes: budding, flowering, the early / late stage of fruit expansion and after harvest.

8. A storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images