Plant growth management method and system based on image recognition

By using image recognition technology and multimodal neural network model in plant planting areas, identifying and counting disease conditions and building a knowledge base, the problem of incomplete assessment of disease in the planting area in the existing technology is solved, and more accurate disease is achieved Detection and management strategies.

CN119068346AInactive Publication Date: 2024-12-03JIANG XI LEDUN PHOTOELECTRIC TECH CO LTD

Patent Information

Application Number
CN202411570346.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-11-06
Publication Date
2024-12-03
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

Existing image recognition technologies rely on image quality and disease area characteristics in plant disease detection, making it difficult to provide a comprehensive assessment of the entire planting area, limiting the refinement and accuracy of identification accuracy and management strategies.

Method used

By obtaining image data and growth environment data of plant planting areas, using image abnormality recognition model and plant disease position detection algorithm, combining multimodal neural network model for disease identification and distribution statistics, building a knowledge base for plant growth management measures to provide accurate Disease warning and management strategies.

Benefits of technology

It improves the accuracy of plant disease identification and the ability to quantify disease conditions in the planted area, and improves the accuracy and effectiveness of management strategies.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119068346A_ABST
    Figure CN119068346A_ABST
Patent Text Reader

Abstract

The invention relates to the field of plant growth management, in particular to a plant growth management method and system based on image recognition. The method comprises the steps that a planting area of a target plant is acquired and divided, and plant image data and growth environment data of each sub-area are acquired; removing abnormal images through an image abnormity identification model, and obtaining plant disease state data by using a plant disease position detection algorithm and a disease identification model; plant disease index distribution data are obtained through a statistical analysis method, and plant disease early warning data are calculated. Meanwhile, a plant growth management measure knowledge base is constructed, and plant growth management measures are extracted and summarized through a large model retrieval enhancement method in combination with disease data and environmental data. Finally, the implementation effect of the management measures is evaluated, the plant growth management measure knowledge base is updated according to the evaluation result, and intelligent management of plant growth is achieved. Through the image recognition technology, the accuracy and efficiency of plant disease detection and management are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of plant growth management, and in particular to a plant growth management method and system based on image recognition. Background Art

[0002] With the continuous progress of agricultural production, timely detection of plant diseases has become a core link in ensuring food security and improving agricultural production efficiency. At present, plant growth management methods and systems using image recognition technology have received great attention in agricultural practice. Existing research shows that deep learning technology has achieved remarkable results in the field of plant disease identification and detection.

[0003] Nevertheless, current image recognition technology is highly dependent on image quality and the significance of diseased area features, which limits further improvement in recognition accuracy. Existing image recognition technology is mainly aimed at identifying diseases of a small number of plants in a local area, and lacks the ability to comprehensively evaluate the disease status of the entire planting area, making it difficult to provide refined and precise management strategies. There is an urgent need for a method that can accurately predict diseases and provide refined and precise management strategies. Summary of the invention

[0004] (1) Technical issues to be resolved The purpose of the present invention is to provide a plant growth management method and system based on image recognition to solve the problem of being unable to efficiently and accurately detect plant diseases and provide accurate management measures.

[0005] (2) Technical solution To achieve the above object, on the one hand, the present invention provides a plant growth management method based on image recognition, the method comprising: S1. Obtain the planting area of ​​the target plant, and divide the planting area according to a preset area division method to obtain Planting sub-areas; obtain the The plant image data in the planting area is obtained by obtaining the first plant image data, and the Plant growth environment data within each planting area; the plant growth environment data includes temperature, humidity, and light intensity; S2, identifying the first plant image data through an image abnormality recognition model, removing abnormal images therein, and obtaining second plant image data; the abnormal images include: abnormal occlusion, abnormal exposure and blurred images; processing the second plant image data through a plant disease position detection algorithm to obtain third plant image data, and identifying the third plant image data through a plant disease recognition model to obtain plant disease status data; S3, obtaining plant disease status regional distribution data by a data statistical analysis method, obtaining plant disease index distribution data by a disease grade weighted average method, and obtaining plant disease early warning data by a threshold analysis function; S4, constructing a knowledge base of plant growth management measures, searching and extracting the plant disease regional distribution data, the plant growth environment data and the plant disease early warning data through a large model retrieval enhancement method, and obtaining plant growth management measures; implementing the plant growth management measures to monitor the implementation effect and obtain the evaluation effect; S5. Analyze and summarize the evaluation results by using a preset statistical analysis method to obtain plant growth management improvement measures, and use the plant growth management improvement measures to update the plant growth management measures knowledge base by using a preset method.

[0006] Furthermore, the method of processing the second plant image data through a plant disease position detection algorithm to obtain third plant image data, and identifying the third plant image data through a plant disease recognition model to obtain plant disease status data includes: The plant disease location detection algorithm is obtained by fine-tuning the target detection algorithm; the second plant image data is scaled to obtain the target detection image data, and the scaling ratio is recorded as ; The target detection image data is used to obtain the position data of the plant disease in the image through the target detection algorithm and recorded as the first plant disease position coordinate ( ); Calculate the first plant disease coordinate data through a coordinate scaling formula to obtain the disease position coordinates corresponding to the second plant image data and record them as the second plant disease position coordinates ( ); The coordinate scaling formula is: ; Cutting the second plant image data according to the second plant disease location coordinates to obtain third plant image data; The plant disease recognition model adopts a multimodal neural network model; the third plant image data is constructed through a prompt word construction method to obtain a plant disease recognition guidance prompt; the plant disease recognition guidance prompt is inferred through a multimodal neural network model to obtain plant disease status data; the plant disease status data includes: plant disease type and plant disease grade.

[0007] Furthermore, the method of obtaining plant disease status regional distribution data by using a data statistical analysis method, obtaining plant disease index distribution data by using a disease grade weighted average method, and obtaining plant disease early warning data by using a threshold analysis function to calculate the disease index distribution data comprises: The plant disease status data are classified and counted according to the plant disease type and plant disease level to obtain plant disease status regional distribution data; the plant disease status regional distribution data includes: the number of plants with various types of diseases and various disease levels in each region, set as ,in Indicates the disease type number, Indicates the disease grade number; the total number of plants in each area is set to ; Preset the number of disease types to , preset the number of disease levels to , set the disease level to , through the disease grade weighted average formula: ; The plant disease index in a sub-region can be calculated. The above calculation is performed on the data of different sub-regions to obtain the plant disease index distribution data. ; Set the warning level and record it as (in ; 0 means no disease, the higher the disease level, the more serious the disease situation), and set the thresholds for each level, recorded as (in ; respectively represent reaching The plant disease early warning level data of the sub-region and the plant disease early warning level data of the whole region can be calculated by the following functions, and the plant disease distribution index is used to obtain the plant disease early warning data through the threshold analysis function. The threshold analysis function is: .

[0008] Furthermore, the method of constructing a plant growth management measure knowledge base, retrieving and extracting and summarizing the plant disease regional distribution data, the plant growth environment data and the plant disease early warning data through a large model retrieval enhancement method, and obtaining plant growth management measures includes:

[0009] Acquire professional knowledge of plant growth management, extract the professional knowledge of plant growth management through a large language model to obtain plant growth management extracted knowledge; obtain expert sorting and summary knowledge through the Delphi method; slice the plant growth management extracted knowledge and the expert sorting and summary knowledge according to a preset method to obtain a plant growth management measures knowledge slice set; process the plant growth management measures knowledge slices, the plant growth management extracted knowledge and the expert sorting and summary knowledge through a context fusion slicing method, and use a BGE text vectorization model for vectorization processing to obtain a plant growth management measures knowledge vector set; store the plant growth management measures knowledge slice set and the plant growth management measures vector set through a structured database to obtain a plant growth management measures knowledge base; The plant growth environment data and the plant disease warning data are constructed into a first prompt for guiding plant growth management measures through a preset prompt word construction method; the first prompt for guiding plant growth management measures is searched against a plant growth management knowledge base through a vector retrieval algorithm to obtain matching plant growth management knowledge; the matching plant growth management knowledge and the first prompt for guiding plant growth management measures are constructed into a prompt through a preset prompt word construction method to obtain a second prompt for guiding plant growth management measures; the second prompt for guiding plant growth management measures is generated through a large language model to generate an inference result to obtain plant growth management measures.

[0010] Furthermore, the method of slicing the plant growth management measures knowledge, the plant growth management extracted knowledge and the expert sorted and summarized knowledge, processing them by a context fusion slicing method, and vectorizing them using a BGE text vectorization model to obtain a plant growth management measures knowledge vector set includes: The plant growth management extracted knowledge or the expert sorted and summarized knowledge is sliced ​​with the corresponding plant growth management measures knowledge, and a guide construction fusion context generation prompt is constructed through a preset prompt word construction method; the guide construction fusion context generation prompt is generated through large language model reasoning to obtain a slice fusion context text; The slice fusion context text and the plant growth management measure knowledge slice are spliced ​​in a preset manner to obtain a plant growth management measure knowledge retrieval text set, and the plant growth management measure knowledge retrieval text set is vectorized through the BGE text vectorization model to obtain a plant growth management measure knowledge vector set.

[0011] Based on the same inventive concept, on the other hand, the present invention also provides a plant growth management system based on image recognition, the system comprising: The data collection module is used to obtain the planting area of ​​the target plant and divide the planting area according to a preset area division method. Planting sub-areas; obtain the The plant image data in the planting area is obtained by obtaining the first plant image data, and the Plant growth environment data within each planting area; the plant growth environment data includes temperature, humidity, and light intensity; A plant disease recognition module is connected to the data collection module and is used to identify the first plant image data through an image abnormality recognition model, remove abnormal images therein, and obtain second plant image data; the abnormal images include: abnormal occlusion, abnormal exposure and blurred images; the second plant image data is processed through a plant disease position detection algorithm to obtain third plant image data, and the third plant image data is identified through a plant disease recognition model to obtain plant disease status data; A plant disease analysis and early warning module is connected to the plant disease identification module, and is used to obtain plant disease state regional distribution data by a data statistical analysis method from the plant disease state data, obtain plant disease index distribution data by a disease grade weighted average method from the plant disease state regional distribution data, and obtain plant disease early warning data by a threshold analysis function from the disease index distribution data; The strategy generation and evaluation module is connected to the plant disease analysis and early warning module, and is used to construct a knowledge base of plant growth management measures, retrieve and extract the plant disease regional distribution data, the plant growth environment data and the plant disease early warning data from the plant growth management measures knowledge base through a large model retrieval enhancement method to obtain plant growth management measures; implement the plant growth management measures to monitor the implementation effect and obtain the evaluation effect; analyze and summarize the evaluation effect through a preset statistical analysis method to obtain plant growth management improvement measures, and update the plant growth management improvement measures to the plant growth management measures knowledge base through a preset method.

[0012] Furthermore, the plant disease identification module includes: The image data processing module is used to obtain a plant disease position detection algorithm by fine-tuning the target detection algorithm; obtain the target detection image data by scaling the second plant image data, and the scaling ratio is recorded as R; obtain the position data of the plant disease in the image by the target detection algorithm and record it as the first plant disease position coordinate ( ); Calculate the first plant disease coordinate data through a coordinate scaling formula to obtain the disease position coordinates corresponding to the second plant image data and record them as the second plant disease position coordinates ( ); The coordinate scaling formula is: ; Cutting the second plant image data according to the second plant disease location coordinates to obtain third plant image data; The plant disease recognition model adopts a multimodal neural network model; the third plant image data is constructed through a prompt word construction method to obtain a plant disease recognition guidance prompt; the plant disease recognition guidance prompt is inferred through a multimodal neural network model to obtain plant disease status data; the plant disease status data includes: plant disease type and plant disease grade.

[0013] Furthermore, the plant disease analysis and early warning module includes: The plant disease distribution calculation module is used to classify and count the plant disease status data according to the plant disease type and plant disease level to obtain plant disease status regional distribution data; the plant disease status regional distribution data includes: the number of plants with various types of diseases and various disease levels in each region, set as ,in Indicates the disease type number, Indicates the disease grade number; the total number of plants in each area is set to ; Preset the number of disease types to , preset the number of disease levels to , set the disease level to , through the disease grade weighted average formula: ; The plant disease index in a sub-region can be calculated. The above calculation is performed on the data of different sub-regions to obtain the plant disease index distribution data. ; The plant disease early warning calculation module is connected to the plant disease distribution calculation module to set the early warning level and record it as (in ; 0 means no disease, the higher the disease level, the more serious the disease situation), and set the thresholds for each level, recorded as (in ; respectively represent reaching The plant disease early warning level data of the sub-region and the plant disease early warning level data of the whole region can be calculated by the following functions, and the plant disease distribution index is used to obtain the plant disease early warning data through the threshold analysis function. The threshold analysis function is: .

[0014] Furthermore, the strategy generation and evaluation module includes: The plant management knowledge base module is used to obtain plant growth management expertise, extract the plant growth management expertise through a large language model to obtain plant growth management extracted knowledge; obtain expert combing and summarizing knowledge through the Delphi method; slice the plant growth management extracted knowledge and the expert combing and summarizing knowledge according to a preset method to obtain a plant growth management measures knowledge slice set; process the plant growth management measures knowledge slices, the plant growth management extracted knowledge and the expert combing and summarizing knowledge through a context fusion slicing method, and use a BGE text vectorization model for vectorization processing to obtain a plant growth management measures knowledge vector set; store the plant growth management measures knowledge slice set and the plant growth management measures vector set through a structured database to obtain a plant growth management measures knowledge base; The large model retrieval enhancement and management strategy generation module is connected to the plant management knowledge base module, and is used to construct a first guiding plant growth management measure generation prompt by using a preset prompt word construction method for the plant growth environment data and the plant disease warning data; retrieve the plant growth management knowledge base through a vector retrieval algorithm for the first guiding plant growth management measure generation prompt to obtain matching plant growth management knowledge; construct a prompt by using a preset prompt word construction method for the matching plant growth management knowledge and the first guiding plant growth management measure generation prompt to obtain a second guiding plant growth management measure generation prompt; generate an inference result for the second guiding plant growth management measure generation prompt through a large language model to obtain a plant growth management measure.

[0015] Furthermore, the plant management knowledge base module includes: The context knowledge fusion module is used to slice the plant growth management extracted knowledge or the expert summarized knowledge with the corresponding plant growth management measures knowledge, and construct a guide to construct a fusion context generation prompt through a preset prompt word construction method; the guide to construct a fusion context generation prompt is generated through a large language model reasoning to obtain a slice fusion context text; The fusion knowledge vectorization module is connected to the context knowledge fusion module, and is used to merge the slices with the context text and the plant growth management measures knowledge slices in a preset manner to obtain a plant growth management measures knowledge retrieval text set, and vectorize the plant growth management measures knowledge retrieval text set through the BGE text vectorization model to obtain a plant growth management measures knowledge vector set.

[0016] (3) Beneficial effects Compared with the prior art, the present invention has the following beneficial effects: 1. Abnormal images are screened out through the image anomaly recognition model, and the plant disease location detection algorithm is used to detect and process the diseased area to enhance the overall weight of the image, thereby enhancing the accuracy of disease recognition in the image.

[0017] 2. Through disease distribution statistics and disease warning calculations, the overall disease situation in the planting area is accurately quantified; a vector knowledge base is constructed through the context fusion slicing method to improve the accuracy of vector retrieval, thereby improving the accuracy of management strategy generation. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a flowchart of a plant growth management method based on image recognition according to Embodiment 1 of the present invention; Figure 2 This is a schematic diagram of the module composition of a plant growth management system based on image recognition according to Example 2 of the present invention. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. 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 creative work are within the scope of protection of the present invention.

[0020] Before giving examples, it is necessary to explain the application scenarios of the present invention. The present invention is used to solve the problem of efficiently and accurately detecting plant diseases and providing accurate management measures.

[0021] Example 1: Figure 1 As shown, this embodiment provides a plant growth management method based on image recognition, the method comprising: S1. Obtain the planting area of ​​the target plant, and divide the planting area according to a preset area division method to obtain Planting sub-areas; obtain the The plant image data in the planting area is obtained by obtaining the first plant image data, and the The plant growth environment data in each planting sub-area; the plant growth environment data includes temperature, humidity, and light intensity; the plant growth environment is different for different planting areas, and the disease conditions will also be different accordingly, so the sub-area division method is adopted to monitor and manage plants in different areas separately. This embodiment takes the monitoring and management of tomato planting diseases as an example. Common diseases in tomato planting include: cotton blight, bacterial canker, early blight, leaf mold, bacterial wilt, damping-off, etc. These diseases mainly occur in the stems, leaves or fruits of tomato plants, so when acquiring their image data, the stems, leaves and fruits of the plants need to be captured as the main body.

[0022] S2, identify the first plant image data through an image anomaly recognition model, remove abnormal images therein, and obtain second plant image data; the abnormal images include: abnormal occlusion, abnormal exposure and blurred images; process the second plant image data through a plant disease position detection algorithm to obtain third plant image data, and identify the third plant image data through a plant disease recognition model to obtain plant disease status data; take tomato disease recognition as an example: first, perform abnormal recognition on the tomato plant image, and remove images with abnormal occlusion, abnormal exposure or blur; process the image after the abnormality is removed through a disease position detection algorithm to obtain an image of the tomato diseased area, that is, the third plant image data; note here: there is a possibility that the plant is not suffering from the disease, then the tomato diseased area obtained by the disease position detection algorithm at this time is the area with the highest probability of the disease; through the plant disease recognition model, according to the characteristics of the tomato stem, leaf or fruit lesions in the image, determine which diseases the plant has, and obtain the disease level of the tomato plant yellowing disease by the proportion of the stem, leaf or fruit with the lesion characteristics in the image; S3, obtaining plant disease status regional distribution data by a data statistical analysis method, calculating the plant disease status regional distribution data by a disease grade weighted average method to obtain plant disease index distribution data, and calculating the disease index distribution data by a threshold analysis function to obtain plant disease early warning data; S4. Construct a knowledge base of plant growth management measures, retrieve and extract the plant disease regional distribution data, the plant growth environment data and the plant disease early warning data through a large model retrieval enhancement method, and obtain plant growth management measures; implement the plant growth management measures to monitor the implementation effect and obtain the evaluation effect. In this embodiment, the types of tomato plant disease management measures mainly include: watering, fertilizing, pesticide application, light regulation, temperature regulation, and humidity regulation, and different types and degrees of measures will be taken for plants in different growth environments and different disease levels. For example, for leaf mold, if the plant is in the early stage of the disease and the severity is relatively mild (level 1~2 warning), you can use 10% fluorosilicone emulsion, 10% benzyl chlorpyrifos water-dispersible granules and other agents for spraying every 7-10 days; if the plant is more seriously ill (level 3~4 warning), increase the concentration of the agent, use 15% fluorosilicone emulsion, 15% benzyl chlorpyrifos water-dispersible granules and other agents for spraying, and increase the spraying frequency accordingly, spraying every 5-7 days. At the same time, it is also necessary to avoid high temperature and high humidity environments, and implement temperature and humidity adjustment measures for the plant growth environment according to actual needs; S5. Analyze and summarize the evaluation results by using a preset statistical analysis method to obtain plant growth management improvement measures, and use the plant growth management improvement measures to update the plant growth management measures knowledge base by using a preset method.

[0023] Furthermore, the method of processing the second plant image data through a plant disease position detection algorithm to obtain third plant image data, and identifying the third plant image data through a plant disease recognition model to obtain plant disease status data includes: The plant disease position detection algorithm is obtained by fine-tuning the target detection algorithm; the second plant image data is scaled to obtain the target detection image data, and the scaling ratio is recorded as R; the target detection image data is used to obtain the position data of the plant disease in the image through the target detection algorithm and recorded as the first plant disease position coordinates ( ); Calculate the first plant disease coordinate data through a coordinate scaling formula to obtain the disease position coordinates corresponding to the second plant image data and record them as the second plant disease position coordinates ( ); The coordinate scaling formula is: ; The second plant image data is cropped according to the second plant disease position coordinates to obtain the third plant image data; Common target detection algorithms include: YOLO, SSD, DenseBox and other deep learning-based algorithms. In this embodiment, the YOLO algorithm is used as the target detection algorithm. Compared with other target detection algorithms, its main advantage is that it maintains high accuracy while having higher detection efficiency. By fine-tuning the YOLO algorithm using labeled image data, a model dedicated to plant disease position detection can be obtained. The purpose of cropping the second plant image data according to the second plant disease position coordinates is to increase the weight of the diseased area in the image so as to more accurately identify the disease situation. Taking the tomato plant disease in this embodiment as an example, assuming that due to the shooting angle and distance, the diseased part of the plant in the original image only accounts for a small part (such as: about 2%), then the plant disease recognition model may ignore this part due to the small weight of the overall image during detection, resulting in inaccurate recognition; however, by detecting the disease position and cropping the original image according to the disease position coordinates obtained by detection, the proportion of the diseased part of the obtained image is relatively larger, and the plant disease recognition model is more likely to accurately identify the disease and the degree of the disease.

[0024] The plant disease recognition model adopts a multimodal neural network model; the third plant image data is constructed through a prompt word construction method to obtain a plant disease recognition guidance prompt; the plant disease recognition guidance prompt is inferred through a multimodal neural network model to obtain plant disease status data; the plant disease status data includes: plant disease type, plant disease level. The prompt word construction is mainly through a preset template, and the plant information and pictures are embedded in the template to obtain a graphic description of the expected result of the guidance model output. For example, in this embodiment, it is hoped that the multimodal neural network model can be guided to identify the disease type and corresponding disease level of the tomato plant, then the preset template can be: "You are an expert in the field of agriculture and can accurately identify plant diseases through images. The following is an image of a #plant name (tomato)# plant. Please identify the type of disease the plant has and the disease level of each disease based on the image. The disease level is determined by the proportion of the diseased area to the diseased part of the plant. A proportion of 0% is set as level 0, a proportion of (0%, 5%] is set as level 1, a proportion of (5%, 10%] is set as level 1, a proportion of (10%, 20%] is set as level 2, a proportion of (20%, 30%] is set as level 3, and a proportion of more than 30% is set as level 4: #plant image#" The content between “#” can be replaced with corresponding plant information according to actual dynamics.

[0025] Furthermore, the method of obtaining plant disease status regional distribution data by using a data statistical analysis method, obtaining plant disease index distribution data by using a disease grade weighted average method, and obtaining plant disease early warning data by using a threshold analysis function to calculate the disease index distribution data comprises: The plant disease status data are classified and counted according to the plant disease type and plant disease level to obtain plant disease status regional distribution data; the plant disease status regional distribution data includes: the number of plants with various types of diseases and various disease levels in each region, set as ,in Indicates the disease type number, Indicates the disease grade number; the total number of plants in each area is set to ; Preset the number of disease types to , preset the number of disease levels to , set the disease level to , through the disease grade weighted average formula: ; The plant disease index in a sub-region can be calculated. The above calculation is performed on the data of different sub-regions to obtain the plant disease index distribution data. ; Set the warning level and record it as (in ; 0 means no disease, the higher the disease level, the more serious the disease situation), and set the thresholds for each level, recorded as (in ; respectively represent reaching The plant disease early warning level data of the sub-region and the plant disease early warning level data of the whole region can be calculated by the following functions, and the plant disease distribution index is used to obtain the plant disease early warning data through the threshold analysis function. The threshold analysis function is: .

[0026] Furthermore, the method of constructing a plant growth management measure knowledge base, retrieving and extracting and summarizing the plant disease regional distribution data, the plant growth environment data and the plant disease early warning data through a large model retrieval enhancement method, and obtaining plant growth management measures includes: Acquire professional knowledge of plant growth management, extract the professional knowledge of plant growth management through a large language model to obtain plant growth management extracted knowledge; obtain expert sorting and summary knowledge through the Delphi method; slice the plant growth management extracted knowledge and the expert sorting and summary knowledge according to a preset method to obtain a plant growth management measures knowledge slice set; process the plant growth management measures knowledge slices, the plant growth management extracted knowledge and the expert sorting and summary knowledge through a context fusion slicing method, and use a BGE text vectorization model for vectorization processing to obtain a plant growth management measures knowledge vector set; store the plant growth management measures knowledge slice set and the plant growth management measures vector set through a structured database to obtain a plant growth management measures knowledge base; The plant growth environment data and the plant disease warning data are constructed into a first prompt for guiding plant growth management measures through a preset prompt word construction method; the first prompt for guiding plant growth management measures is searched against a plant growth management knowledge base through a vector retrieval algorithm to obtain matching plant growth management knowledge; the matching plant growth management knowledge and the first prompt for guiding plant growth management measures are constructed into a prompt through a preset prompt word construction method to obtain a second prompt for guiding plant growth management measures; the second prompt for guiding plant growth management measures is generated through a large language model to generate an inference result to obtain plant growth management measures.

[0027] Furthermore, the method of slicing the plant growth management measures knowledge, the plant growth management extracted knowledge and the expert sorted and summarized knowledge, processing them by a context fusion slicing method, and vectorizing them using a BGE text vectorization model to obtain a plant growth management measures knowledge vector set includes: The plant growth management extracted knowledge or the expert sorted and summarized knowledge is sliced ​​with the corresponding plant growth management measures knowledge, and a guide construction fusion context generation prompt is constructed through a preset prompt word construction method; the guide construction fusion context generation prompt is generated through large language model reasoning to obtain a slice fusion context text; The slice fusion context text and the plant growth management measure knowledge slice are spliced ​​in a preset manner to obtain a plant growth management measure knowledge retrieval text set, and the plant growth management measure knowledge retrieval text set is vectorized through the BGE text vectorization model to obtain a plant growth management measure knowledge vector set.

[0028] Embodiment 2: Based on the same inventive concept, Figure 2 As shown, this embodiment also provides a plant growth management system based on image recognition, the system comprising: The data collection module is used to obtain the planting area of ​​the target plant and divide the planting area according to a preset area division method. Planting sub-areas; obtain the The plant image data in the planting area is obtained by obtaining the first plant image data, and the Plant growth environment data within each planting area; the plant growth environment data includes temperature, humidity, and light intensity; A plant disease recognition module is connected to the data collection module and is used to identify the first plant image data through an image abnormality recognition model, remove abnormal images therein, and obtain second plant image data; the abnormal images include: abnormal occlusion, abnormal exposure and blurred images; the second plant image data is processed through a plant disease position detection algorithm to obtain third plant image data, and the third plant image data is identified through a plant disease recognition model to obtain plant disease status data; A plant disease analysis and early warning module is connected to the plant disease identification module, and is used to obtain plant disease state regional distribution data by a data statistical analysis method from the plant disease state data, obtain plant disease index distribution data by a disease grade weighted average method from the plant disease state regional distribution data, and obtain plant disease early warning data by a threshold analysis function from the disease index distribution data; The strategy generation and evaluation module is connected to the plant disease analysis and early warning module, and is used to construct a knowledge base of plant growth management measures, retrieve and extract the plant disease regional distribution data, the plant growth environment data and the plant disease early warning data from the plant growth management measures knowledge base through a large model retrieval enhancement method to obtain plant growth management measures; implement the plant growth management measures to monitor the implementation effect and obtain the evaluation effect; analyze and summarize the evaluation effect through a preset statistical analysis method to obtain plant growth management improvement measures, and update the plant growth management improvement measures to the plant growth management measures knowledge base through a preset method.

[0029] Furthermore, the plant disease identification module includes: The plant disease location detection algorithm is obtained by fine-tuning the target detection algorithm; the second plant image data is scaled to obtain the target detection image data, and the scaling ratio is recorded as ; The target detection image data is used to obtain the position data of the plant disease in the image through the target detection algorithm and recorded as the first plant disease position coordinate ( ); Calculate the first plant disease coordinate data through a coordinate scaling formula to obtain the disease position coordinates corresponding to the second plant image data and record them as the second plant disease position coordinates ( ); The coordinate scaling formula is: ; Cutting the second plant image data according to the second plant disease location coordinates to obtain third plant image data; The plant disease recognition model adopts a multimodal neural network model; the third plant image data is constructed through a prompt word construction method to obtain a plant disease recognition guidance prompt; the plant disease recognition guidance prompt is inferred through a multimodal neural network model to obtain plant disease status data; the plant disease status data includes: plant disease type and plant disease grade.

[0030] Furthermore, the plant disease analysis and early warning module includes: The plant disease distribution calculation module is used to classify and count the plant disease status data according to the plant disease type and plant disease level to obtain plant disease status regional distribution data; the plant disease status regional distribution data includes: the number of plants with various types of diseases and various disease levels in each region, set as ,in Indicates the disease type number, Indicates the disease grade number; the total number of plants in each area is set to ; Preset the number of disease types to , preset the number of disease levels to , set the disease level to , through the disease grade weighted average formula: ; The plant disease index in a sub-region can be calculated. The above calculation is performed on the data of different sub-regions to obtain the plant disease index distribution data. ; The plant disease early warning calculation module is connected to the plant disease distribution calculation module to set the early warning level and record it as (in ; 0 means no disease, the higher the disease level, the more serious the disease situation), and set the thresholds for each level, recorded as (in ; respectively represent reaching The plant disease early warning level data of the sub-region and the plant disease early warning level data of the whole region can be calculated by the following functions, and the plant disease distribution index is used to obtain the plant disease early warning data through the threshold analysis function. The threshold analysis function is: .

[0031] Furthermore, the strategy generation and evaluation module includes: The plant management knowledge base module is used to obtain plant growth management expertise, extract the plant growth management expertise through a large language model to obtain plant growth management extracted knowledge; obtain expert combing and summarizing knowledge through the Delphi method; slice the plant growth management extracted knowledge and the expert combing and summarizing knowledge according to a preset method to obtain a plant growth management measures knowledge slice set; process the plant growth management measures knowledge slices, the plant growth management extracted knowledge and the expert combing and summarizing knowledge through a context fusion slicing method, and use a BGE text vectorization model for vectorization processing to obtain a plant growth management measures knowledge vector set; store the plant growth management measures knowledge slice set and the plant growth management measures vector set through a structured database to obtain a plant growth management measures knowledge base; The large model retrieval enhancement and management strategy generation module is connected to the plant management knowledge base module, and is used to construct a first guiding plant growth management measure generation prompt by using a preset prompt word construction method for the plant growth environment data and the plant disease warning data; retrieve the plant growth management knowledge base through a vector retrieval algorithm for the first guiding plant growth management measure generation prompt to obtain matching plant growth management knowledge; construct a prompt by using a preset prompt word construction method for the matching plant growth management knowledge and the first guiding plant growth management measure generation prompt to obtain a second guiding plant growth management measure generation prompt; generate an inference result for the second guiding plant growth management measure generation prompt through a large language model to obtain a plant growth management measure.

[0032] Furthermore, the plant management knowledge base module includes: The context knowledge fusion module is used to slice the plant growth management extracted knowledge or the expert summarized knowledge with the corresponding plant growth management measures knowledge, and construct a guide to construct a fusion context generation prompt through a preset prompt word construction method; the guide to construct a fusion context generation prompt is generated through a large language model reasoning to obtain a slice fusion context text; The fusion knowledge vectorization module is connected to the context knowledge fusion module, and is used to merge the slices with the context text and the plant growth management measures knowledge slices in a preset manner to obtain a plant growth management measures knowledge retrieval text set, and vectorize the plant growth management measures knowledge retrieval text set through the BGE text vectorization model to obtain a plant growth management measures knowledge vector set.

[0033] It should be noted that, regarding the system in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0034] Finally, it should be noted that: Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A plant growth management method based on image recognition, characterized in that: The method comprises: Obtain the planting area of ​​the target plant, and divide the planting area according to a preset area division method to obtain Planting sub-areas; obtain the The plant image data in the planting area is obtained by obtaining the first plant image data, and the Plant growth environment data within each planting area; the plant growth environment data includes temperature, humidity, and light intensity; The first plant image data is identified by an image abnormality recognition model, and abnormal images are removed to obtain second plant image data; the abnormal images include: abnormal occlusion, abnormal exposure and blurred images; the second plant image data is processed by a plant disease position detection algorithm to obtain third plant image data, and the third plant image data is identified by a plant disease recognition model to obtain plant disease status data; The plant disease status data are subjected to a data statistical analysis method to obtain plant disease status regional distribution data, the plant disease status regional distribution data are subjected to a disease grade weighted average method to obtain plant disease index distribution data, and the disease index distribution data are subjected to a threshold analysis function to obtain plant disease early warning data; Constructing a knowledge base of plant growth management measures, searching and extracting the plant disease regional distribution data, the plant growth environment data and the plant disease early warning data through a large model retrieval enhancement method, and obtaining plant growth management measures; implementing the plant growth management measures to monitor the implementation effect and obtain the evaluation effect; The evaluation effect is analyzed and summarized by a preset statistical analysis method to obtain plant growth management improvement measures, and the plant growth management improvement measures are used to update the plant growth management measures knowledge base by a preset method.

2. A plant growth management method based on image recognition according to claim 1, characterized in that: The method of processing the second plant image data by a plant disease position detection algorithm to obtain third plant image data, and identifying the third plant image data by a plant disease recognition model to obtain plant disease status data includes: The plant disease location detection algorithm is obtained by fine-tuning the target detection algorithm; the second plant image data is scaled to obtain the target detection image data, and the scaling ratio is recorded as ; The target detection image data is used to obtain the position data of the plant disease in the image through the target detection algorithm and recorded as the first plant disease position coordinate ( ); Calculate the first plant disease coordinate data through a coordinate scaling formula to obtain the disease position coordinates corresponding to the second plant image data and record them as the second plant disease position coordinates ( ); The coordinate scaling formula is: ; Cutting the second plant image data according to the second plant disease location coordinates to obtain third plant image data; The plant disease recognition model adopts a multimodal neural network model; the third plant image data is constructed through a prompt word construction method to obtain a plant disease recognition guidance prompt; the plant disease recognition guidance prompt is inferred through a multimodal neural network model to obtain plant disease status data; the plant disease status data includes: plant disease type and plant disease grade.

3. A plant growth management method based on image recognition according to claim 1, characterized in that: The method of obtaining plant disease status regional distribution data by using a data statistical analysis method, obtaining plant disease index distribution data by using a disease grade weighted average method, and obtaining plant disease early warning data by using a threshold analysis function for the disease index distribution data comprises: The plant disease status data are classified and counted according to the plant disease type and plant disease level to obtain plant disease status regional distribution data; the plant disease status regional distribution data includes: the number of plants with various types of diseases and various disease levels in each region, set as ,in Indicates the disease type number, Indicates the disease grade number; the total number of plants in each area is set to ; Preset the number of disease types to , preset the number of disease levels to , set the disease level to , through the disease grade weighted average formula: ; The plant disease index in a sub-region can be calculated. The above calculation is performed on the data of different sub-regions to obtain the plant disease index distribution data. ; Set the warning level and record it as (in ; 0 means no disease, the higher the disease level, the more serious the disease situation), and set the thresholds for each level, recorded as (in ; respectively represent reaching The threshold value of abnormal level) can be used to calculate the sub-region plant disease warning level data and the whole region plant disease warning level data respectively through the threshold analysis function, and the plant disease distribution index is used to obtain the plant disease warning data through the threshold analysis function. The threshold analysis function is: 。 4. A plant growth management method based on image recognition according to claim 1, characterized in that: The method of constructing a plant growth management measure knowledge base, retrieving and extracting and summarizing the plant disease regional distribution data, the plant growth environment data and the plant disease early warning data through a large model retrieval enhancement method, and obtaining plant growth management measures includes: Acquire professional knowledge of plant growth management, extract the professional knowledge of plant growth management through a large language model to obtain plant growth management extracted knowledge; obtain expert sorting and summary knowledge through the Delphi method; slice the plant growth management extracted knowledge and the expert sorting and summary knowledge according to a preset method to obtain a plant growth management measures knowledge slice set; process the plant growth management measures knowledge slices, the plant growth management extracted knowledge and the expert sorting and summary knowledge through a context fusion slicing method, and use a BGE text vectorization model for vectorization processing to obtain a plant growth management measures knowledge vector set; store the plant growth management measures knowledge slice set and the plant growth management measures vector set through a structured database to obtain a plant growth management measures knowledge base; The plant growth environment data and the plant disease warning data are constructed into a first prompt for guiding plant growth management measures through a preset prompt word construction method; the first prompt for guiding plant growth management measures is searched against a plant growth management knowledge base through a vector retrieval algorithm to obtain matching plant growth management knowledge; the matching plant growth management knowledge and the first prompt for guiding plant growth management measures are constructed into a prompt through a preset prompt word construction method to obtain a second prompt for guiding plant growth management measures; the second prompt for guiding plant growth management measures is generated through a large language model to generate an inference result to obtain plant growth management measures.

5. A plant growth management method based on image recognition according to claim 4, characterized in that: The method of slicing the plant growth management measures knowledge, the plant growth management extracted knowledge and the expert sorted and summarized knowledge, processing them by a context fusion slicing method, and vectorizing them using a BGE text vectorization model to obtain a plant growth management measures knowledge vector set includes: The plant growth management extracted knowledge or the expert sorted and summarized knowledge is sliced ​​with the corresponding plant growth management measures knowledge, and a guide construction fusion context generation prompt is constructed through a preset prompt word construction method; the guide construction fusion context generation prompt is generated through large language model reasoning to obtain a slice fusion context text; The slice fusion context text and the plant growth management measure knowledge slice are spliced ​​in a predetermined manner to obtain a plant growth management measure knowledge retrieval text set, and the plant growth management measure knowledge retrieval text set is vectorized through the BGE text vectorization model to obtain a plant growth management measure knowledge vector set.

6. A plant growth management system based on image recognition, characterized in that: The system comprises: The data collection module is used to obtain the planting area of ​​the target plant and divide the planting area according to a preset area division method. Planting sub-areas; obtain the The plant image data in the planting area is obtained by obtaining the first plant image data, and the Plant growth environment data within each planting area; the plant growth environment data includes temperature, humidity, and light intensity; A plant disease recognition module is connected to the data collection module and is used to identify the first plant image data through an image abnormality recognition model, remove abnormal images therein, and obtain second plant image data; the abnormal images include: abnormal occlusion, abnormal exposure and blurred images; the second plant image data is processed through a plant disease position detection algorithm to obtain third plant image data, and the third plant image data is identified through a plant disease recognition model to obtain plant disease status data; A plant disease analysis and early warning module is connected to the plant disease identification module, and is used to obtain plant disease state regional distribution data by a data statistical analysis method from the plant disease state data, obtain plant disease index distribution data by a disease grade weighted average method from the plant disease state regional distribution data, and obtain plant disease early warning data by a threshold analysis function from the disease index distribution data; The strategy generation and evaluation module is connected to the plant disease analysis and early warning module, and is used to construct a knowledge base of plant growth management measures, retrieve and extract the plant disease regional distribution data, the plant growth environment data and the plant disease early warning data from the plant growth management measures knowledge base through a large model retrieval enhancement method to obtain plant growth management measures; implement the plant growth management measures to monitor the implementation effect and obtain the evaluation effect; analyze and summarize the evaluation effect through a preset statistical analysis method to obtain plant growth management improvement measures, and update the plant growth management improvement measures to the plant growth management measures knowledge base through a preset method.

7. A plant growth management system based on image recognition according to claim 6, characterized in that: The plant disease identification module comprises: The image data processing module is used to obtain a plant disease location detection algorithm by fine-tuning the target detection algorithm; the second plant image data is scaled to obtain the target detection image data, and the scaling ratio is recorded as ; The target detection image data is used to obtain the position data of the plant disease in the image through the target detection algorithm and recorded as the first plant disease position coordinate ( ); Calculate the first plant disease coordinate data through a coordinate scaling formula to obtain the disease position coordinates corresponding to the second plant image data and record them as the second plant disease position coordinates ( ); The coordinate scaling formula is: ; Cutting the second plant image data according to the second plant disease location coordinates to obtain third plant image data; The plant disease condition identification and analysis module is connected to the image data processing module, and is used to construct the plant disease identification guidance prompt through the prompt word construction method of the third plant image data; the plant disease identification guidance prompt is inferred through a multimodal neural network model to obtain plant disease status data; the plant disease status data includes: plant disease type and plant disease grade.

8. A plant growth management system based on image recognition according to claim 6, characterized in that: The plant disease analysis and early warning module includes: The plant disease distribution calculation module is used to classify and count the plant disease status data according to the plant disease type and plant disease level to obtain plant disease status regional distribution data; the plant disease status regional distribution data includes: the number of plants with various types of diseases and various disease levels in each region, set as ,in Indicates the disease type number, Indicates the disease grade number; the total number of plants in each area is set to ; Preset the number of disease types to , preset the number of disease levels to , set the disease level to , through the disease grade weighted average formula: ; The plant disease index in a sub-region can be calculated. The above calculation is performed on the data of different sub-regions to obtain the plant disease index distribution data. ; The plant disease early warning calculation module is connected to the plant disease distribution calculation module to set the early warning level and record it as (in ; 0 means no disease, the higher the disease level, the more serious the disease situation), and set the thresholds for each level, recorded as (in ; respectively represent reaching The plant disease early warning level data of the sub-region and the plant disease early warning level data of the whole region can be calculated by the following functions, and the plant disease distribution index is used to obtain the plant disease early warning data through the threshold analysis function. The threshold analysis function is: 。 9. A plant growth management system based on image recognition according to claim 6, characterized in that: The strategy generation and evaluation module includes: The plant management knowledge base module is used to obtain plant growth management expertise, extract the plant growth management expertise through a large language model to obtain plant growth management extracted knowledge; obtain expert combing and summarizing knowledge through the Delphi method; slice the plant growth management extracted knowledge and the expert combing and summarizing knowledge according to a preset method to obtain a plant growth management measures knowledge slice set; process the plant growth management measures knowledge slices, the plant growth management extracted knowledge and the expert combing and summarizing knowledge through a context fusion slicing method, and use a BGE text vectorization model for vectorization processing to obtain a plant growth management measures knowledge vector set; store the plant growth management measures knowledge slice set and the plant growth management measures vector set through a structured database to obtain a plant growth management measures knowledge base; The large model retrieval enhancement and management strategy generation module is connected to the plant management knowledge base module, and is used to construct a first guiding plant growth management measure generation prompt by using a preset prompt word construction method for the plant growth environment data and the plant disease warning data; retrieve the plant growth management knowledge base through a vector retrieval algorithm for the first guiding plant growth management measure generation prompt to obtain matching plant growth management knowledge; construct a prompt by using a preset prompt word construction method for the matching plant growth management knowledge and the first guiding plant growth management measure generation prompt to obtain a second guiding plant growth management measure generation prompt; generate an inference result for the second guiding plant growth management measure generation prompt through a large language model to obtain a plant growth management measure.

10. A plant growth management system based on image recognition according to claim 9, characterized in that: The plant management knowledge base module includes: The context knowledge fusion module is used to slice the plant growth management extracted knowledge or the expert summarized knowledge with the corresponding plant growth management measures knowledge, and construct a guide to construct a fusion context generation prompt through a preset prompt word construction method; the guide to construct a fusion context generation prompt is generated through a large language model reasoning to obtain a slice fusion context text; The fusion knowledge vectorization module is connected to the context knowledge fusion module, and is used to merge the slices with the context text and the plant growth management measures knowledge slices in a preset manner to obtain a plant growth management measures knowledge retrieval text set, and vectorize the plant growth management measures knowledge retrieval text set through the BGE text vectorization model to obtain a plant growth management measures knowledge vector set.

Citation Information

Patent Citations

  • Lightweight optimization Yolo v4-based tea disease identification method and system

    CN114359727A

  • Intelligent monitoring and control method for tea garden leaf diseases

    CN115830003A

  • Knowledge base construction method, video automatic production method and software product

    CN117952203A

  • Landscaping plant growth monitoring method and system

    CN117994729A

  • Crop disease and pest detection method and system based on multi-modal deep learning

    CN118258769A

Cited By

  • Crop fruit quality detection method and system based on multi-modal large model

    CN119810826A

  • Greenhouse management device and method based on sensor array and large model

    CN121028629A

  • Multi-modal small sample plant disease identification method

    CN121505459A

  • A multi-modal small sample plant disease identification method

    CN121505459B