An intelligent agricultural management system based on planting-harvesting-distribution whole link

Abstract

The present application relates to a kind of intelligence agricultural management system based on planting-harvesting-distribution whole link, belong to agricultural field, system includes: online seed selection module, order tracking module, environment monitoring early warning module, pest identification module, market prediction module, logistics scheduling module, product estimation module and growth state evaluation module.The present application is in planting link, through intelligent agent to crop growth environment is intelligently monitored and abnormal early warning, can be for farmer to formulate the accurate planting scheme of different crops, and irrigation and fertilization scheme suitable for different regions;In production process, intelligent agent can also intelligently identify bad fruit and give treatment scheme, accurately identify pest and provide effective control means;In market sales link, market trend can be predicted, production scheme is reasonably arranged, and the balance of production and marketing of agricultural products is ensured.

Description

Technical Field

[0001] This invention relates to the agricultural field, and more particularly to a smart agricultural management system based on the entire chain of planting, harvesting and distribution. Background Technology

[0002] In today's booming digital agriculture market, various agricultural platforms are emerging, and competition in the smart agriculture sector is becoming increasingly fierce. Current technologies and products primarily focus on the transaction of agricultural products, providing farmers and consumers with a simple trading platform through online sales. For example, community group-buying platforms like Meituan Youxuan play a crucial role in connecting consumers and agricultural products. By integrating supply chain resources, they have built a convenient online sales channel, allowing consumers to purchase various agricultural products at relatively low prices. These platforms, with their robust logistics and distribution systems, can quickly and efficiently deliver agricultural products to communities, meeting consumers' daily needs. However, these platforms primarily focus on sales and delivery, without deeply engaging in the production process. Their quality control often stops at simple steps like receiving and inspection, with little understanding of the specific circumstances during planting and breeding. For instance, they cannot provide guidance and supervision on farmers' planting methods, and farmers cannot obtain timely and professional assistance from these platforms when encountering technical difficulties or pest and disease control issues. They also cannot ensure that agricultural products adhere to green, environmentally friendly, and safe standards during their growth. This could lead to potential quality and safety issues with the agricultural products purchased by consumers. Furthermore, due to a lack of in-depth understanding of the production process, community group-buying platforms struggle to rationally plan and guide agricultural production based on changes in market demand, thus failing to fundamentally address the mismatch between supply and demand for agricultural products.

[0003] In addition, amidst the current wave of smart agriculture development, some existing products focus on the management aspects of agricultural production, such as the emerging "smart agricultural technology platforms." These smart agricultural management systems utilize advanced agricultural IoT technology to collect key data about the farmland environment in real time and with precision. For example, they can meticulously monitor soil pH, fertility, humidity changes, and information such as air temperature, humidity, and light intensity. Simultaneously, they employ basic data analysis capabilities to perform preliminary processing and interpretation of this collected data, providing farmers with some reference information and helping them understand the environmental conditions for crop growth. However, these platforms have significant limitations. They primarily focus on data monitoring and basic analysis within the agricultural production process, lacking integration of the entire chain of agricultural products from production to sales. In other words, they do not provide effective solutions for how to efficiently bring agricultural products to market after harvest to meet consumer demand, or how to establish production plans that match market demand. While farmers can obtain data on crop growth environment through these platforms, they still face many challenges in the sales process, such as not being able to find suitable sales channels and difficulty in predicting market demand, which may lead to unsold agricultural products and prevent the maximization of economic benefits in agricultural production. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide a smart agricultural management system based on the entire chain of planting, harvesting and distribution, thus solving the deficiencies of the existing technology.

[0005] The objective of this invention is achieved through the following technical solution: a smart agricultural management system based on the entire chain of planting-harvesting-distribution, the system comprising: an online seed selection module, an order tracking module, an environmental monitoring and early warning module, a pest and disease identification module, a market forecasting module, a logistics scheduling module, a product estimation module, and a growth status assessment module;

[0006] The online seed selection module is configured to allow consumers to select crops according to their preferences and needs, and then transmit the selection information to the corresponding farmers to start a personalized planting service.

[0007] The order tracking module is configured so that when farmers begin to terminate the crop selected by consumers, consumers can track the status of their orders in real time through the system.

[0008] The environmental monitoring and early warning module is configured to monitor and analyze environmental data of farmland, including soil moisture, temperature, pH, air temperature, humidity, and light intensity. Once abnormal environmental data is detected, it will immediately send early warning information to farmers and provide suggestions for countermeasures.

[0009] The pest and disease identification module is configured as an intelligent agent to quickly and accurately identify the types of pests and diseases using image recognition and deep learning technologies, and to provide prevention and control solutions.

[0010] The market forecasting module is configured to analyze the supply and demand situation and price fluctuation trends of different fruit and vegetable varieties in the market through big data analysis and artificial intelligence. The intelligent agent predicts the trend of agricultural product market, adjusts planting plans, and selects varieties with high market demand and high prices for planting.

[0011] The logistics scheduling module is configured to arrange logistics vehicles and delivery routes according to the distribution of orders and delivery time requirements, monitor the location and transportation status of logistics vehicles in real time, and adjust the logistics plan in a timely manner according to external factors to ensure the smooth progress of delivery.

[0012] The product prediction module is configured as a YOLOv8 model to analyze images during the crop growth process, identify the quantity and growth status of crops, and predict the final yield by combining historical data and environmental factors.

[0013] The growth status assessment module is configured to integrate environmental data, pest and disease conditions, and growth cycle. The intelligent agent performs quantitative analysis on the health status and growth rate of the crop, continuously assesses the growth status of the crop, and provides suggestions and solutions to farmers.

[0014] The system also includes: a knowledge popularization module, an intelligent planting guidance module, and a plant adoption module;

[0015] The knowledge popularization module is configured to provide consumers with agricultural knowledge, including the nutritional value of different fruits and vegetables, storage methods, cooking suggestions, crop planting process, and prevention and control of diseases and pests.

[0016] The intelligent planting guidance module is configured as an intelligent agent to develop personalized planting plans for farmers by collecting farmland-level environmental data and combining it with crop growth patterns and historical planting data.

[0017] The plant adoption module is configured so that consumers can select different types of land and plant varieties on the land according to their personal preferences, and purchase the corresponding adoption rights.

[0018] The intelligent agent also uses data analysis to determine whether the environment is suitable for crop growth; it integrates crop variety characteristics, climate and soil conditions of the planting area, and seasonal changes to customize exclusive planting plans for different crops; it formulates precise irrigation and fertilization plans based on regional differences, combined with water resource status and soil fertility; and it identifies damaged fruits and provides treatment solutions.

[0019] The process of determining whether the environment is suitable for crop growth through data analysis specifically includes the following:

[0020] Real-time environmental data acquisition: Deploy a distributed IoT sensor network in farmland, including soil sensors that monitor soil temperature, humidity, pH, and EC value; air sensors that monitor low air temperature and CO2 concentration; light sensors that monitor photosynthetically active radiation; and weather stations that monitor wind speed and precipitation. At the same time, use BeiDou positioning to achieve field-level data spatial calibration, link data with specific planting areas, and provide spatial coordinate benchmarks for subsequent regional judgment.

[0021] Construction of a crop growth suitability parameter database: Based on an agricultural knowledge base, a crop variety characteristic database, and a database of experienced farmers, a three-dimensional threshold system of crop-growth stage-environment parameters is constructed; the basic thresholds are dynamically corrected by combining historical planting data and machine learning models, and environmental trends are analyzed to predict environmental risks within a set timeframe in the future;

[0022] Data analysis model judgment: The collected raw data is preprocessed and a multi-level suitability judgment algorithm is introduced for judgment. The warning level is automatically divided according to the degree of data deviation. The actual crop growth feedback after the warning is periodically compared with the judgment result. The parameter library threshold and multi-factor weights are updated through reinforcement learning algorithm to improve the judgment accuracy.

[0023] The multi-level suitability assessment algorithm includes:

[0024] The first layer, rule engine, quickly screens: directly compares real-time data with the threshold range in the parameter library. If a single dimension of data exceeds the threshold, it is immediately marked as abnormal.

[0025] The second layer, the multi-factor collaborative analysis model, uses the gradient boosting tree algorithm to analyze the interaction of multiple parameters. If any parameter in the interaction exceeds the threshold, the model judges it as a potential risk environment.

[0026] The third layer, growth stage adaptation and adjustment: by recording the planting cycle and using image recognition to confirm the current growth stage of the crop, the parameter thresholds for the current stage are adjusted to prevent misjudgments caused by general parameter thresholds.

[0027] The comprehensive consideration of crop variety characteristics, local climate and soil conditions, and seasonal variations, along with the following specific content, are used to customize planting plans for different crops:

[0028] Multi-dimensional basic data collection and integration: Construct a crop variety characteristic database, collect regional climate and soil data in real time, divide seasonal modules according to phenological periods, associate typical climate parameters of each season, and map them to the environmental requirements of key crop growth stages.

[0029] The intelligent decision-making model generates the following solutions: it calculates the optimal sowing time window by combining the crop germination temperature threshold with the regional spring temperature rise curve; it calculates the planting density based on soil fertility level, crop plant type, and row spacing formula; it integrates experienced farmers' experience base and knowledge base to generate basic solutions for irrigation cycle and fertilization type; it performs regional corrections on the generated basic solutions based on regional characteristics, and incorporates preventive measures into the solutions by combining seasonal climate risk database.

[0030] The solution's full-cycle dynamic adjustment mechanism: Based on the recorded crop growth stages, the intelligent agent updates the solution every set time interval, and when environmental monitoring data deviates from the set range, the solution's emergency adjustment mechanism is automatically triggered. The intelligent agent also optimizes the solution based on feedback data from farmers and breaks down the solution into executable task chains, which are then automatically pushed to the farmer's app.

[0031] The specific measures for developing precise irrigation and fertilization plans based on regional differences, combined with water resource conditions and soil fertility, include the following:

[0032] Multi-dimensional basic data collection and regional feature modeling: collect soil fertility data and combine it with regional soil survey data to establish a regional baseline for soil fertility; collect water resource data, classify water resource abundance levels and associate them with irrigation costs; establish a crop demand database;

[0033] Precision irrigation plan generation: Calculates crop water requirements in real time based on crop coefficient, evapotranspiration, water resource abundance level, and soil water retention capacity; selects irrigation methods according to regional water source type, and optimizes irrigation time based on weather forecasts and crop growth stages; outputs specific irrigation parameters and controls irrigation equipment to execute automatically via the Internet of Things.

[0034] Precision fertilization program generation: Determine the target nutrient uptake based on crop variety and growth stage, and calculate the natural supply ratio based on current soil fertility data; select fertilizer type based on soil type, choose application method based on regional agricultural machinery conditions, and formulate environmental protection constraints;

[0035] Regional adaptation and dynamic optimization mechanism: The country is divided into 6 major agricultural ecological regions. Each region is dynamically adjusted according to regional differences. The intelligent agent collects soil data every N days based on IoT devices, compares the actual values ​​with the expected values ​​of the plan in combination with the crop growth status, and optimizes the fertilization plan based on the comparison results.

[0036] The pest and disease identification module specifically includes the following:

[0037] Multidimensional construction of a pest and disease feature database: A feature database of crop pests and diseases is constructed by using an agricultural knowledge base, field image sample collection, and data digitization of experienced farmers' knowledge.

[0038] Pest and disease identification model construction and training: Based on the ResNet-50 model, a CBAM attention module was added to enhance the extraction of key features of lesion edges and insect body contours, and depthwise separable convolution was used to compress the pest and disease identification model parameters by 40%; image data augmentation was performed to expand the sample, and training was carried out according to pest and disease type + crop variety, with an increased sample proportion for high-incidence pests and diseases, and misjudgment of similar pests and diseases was corrected through confusion matrix to improve the identification accuracy of the pest and disease identification model; the trained pest and disease identification model was deployed on edge devices at the farmer's end;

[0039] Pest and disease identification and control by the intelligent agent: Farmers upload images of diseased parts of fruits through the system. The intelligent agent preprocesses the images and performs multi-dimensional identification and cross-validation. It generates control plans based on chemical control, physical / biological control, and emergency treatment. Finally, the control plans are optimized based on farmer feedback data and actual field results.

[0040] The identification of damaged fruit and the provision of treatment solutions specifically include the following:

[0041] Construction of Damaged Fruit Feature Database: Construct a multi-dimensional database of damaged fruit features of crops, clarify the classification of disease damage, insect damage and physical damage types, and label the collected images of damaged fruits.

[0042] Training and deployment of the damaged fruit recognition model: A coordinate attention module was added to the Neck layer of the MobileNetV3 deep learning model to enhance feature extraction of the damaged area edges. Depthwise separable convolution was used to reduce the parameters of the damaged fruit recognition model by 35% to adapt to farmers' mobile phones and field smart terminal devices. Data augmentation was performed on the images to expand the sample. Classification training was carried out according to crop variety + damage type. For the set crops, the sample weights were increased, and the ability to distinguish similar damage was optimized through confusion matrix. The trained damaged fruit recognition model was deployed on the farmer's edge device.

[0043] Damaged fruit identification by the intelligent agent: Farmers upload images of damaged fruit, and the intelligent agent performs multi-dimensional identification and cross-validation after preprocessing the images;

[0044] Treatment plan generation: The intelligent agent automatically matches treatment measurements according to the degree of damage, tracks the treatment progress through IoT devices, and collects downstream feedback data to update the weights of the treatment rules in order to optimize the plan.

[0045] The product estimation module specifically includes the following:

[0046] Crop growth image data acquisition and annotation: Collect images from multiple scenes, and perform manual annotation and automated verification on the acquired images;

[0047] Training and optimization of the YOLOv8 model: SE attention was added to the Neck layer of the YOLOv8 model to enhance the extraction of key features of fruit outline and color, reduce model parameters to adapt to farmers' mobile phones and field edge devices, and perform hyperparameter tuning; the YOLOv8 model was first pre-trained on a general object detection dataset, and then the trained YOLOv8 model was adjusted using a crop fruit dataset to optimize the recognition ability of overlapping and occluded fruits, and sub-models were trained for different crop characteristics;

[0048] Yield estimation: After extracting features from the input image, the YOLOv8 model outputs data on the number of fruits, fruit size, and fruit distribution density. Combining crop variety characteristics and planting parameters, the estimated total yield is calculated as: estimated total yield = number of fruits per unit area × total planting area × average weight of a single fruit × maturity rate correction coefficient. Environmental factors are then introduced to correct the results.

[0049] This invention offers the following advantages: a smart agricultural management system based on the entire planting-harvesting-distribution chain. In the planting stage, intelligent agents monitor the crop growth environment and provide early warnings of anomalies, enabling farmers to develop precise planting plans for different crops, as well as irrigation and fertilization plans suitable for different regions. During production, the intelligent agents can intelligently identify damaged fruit and provide treatment solutions, accurately identify pests and diseases, and provide effective control measures. In the market sales stage, it can predict market trends, rationally arrange production plans, and ensure a balance between agricultural production and sales. Simultaneously, it combines experienced farmers' knowledge with various knowledge bases to provide farmers and consumers with abundant agricultural knowledge. Furthermore, it can predict the optimal harvest time, assess crop growth status, rationally optimize production decisions, and even use the Yolov8 model for yield forecasting, providing comprehensive intelligent support for the entire agricultural production process. Attached Figure Description

[0050] Figure 1 This is a schematic diagram of the structure of the present invention. Detailed Implementation

[0051] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the detailed description of the embodiments of this application provided below with reference to the accompanying drawings is not intended to limit the scope of protection of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application. The present invention will be further described below with reference to the accompanying drawings.

[0052] This invention relates to a smart agricultural management system or platform based on the entire chain of planting, harvesting, and distribution. This platform accurately targets different types of urban consumers who pursue fresh and customized fruits and vegetables, as well as farmers and cooperatives facing difficulties in planting and sales, and can provide them with targeted services to meet their diverse needs.

[0053] like Figure 1 As shown, the system includes: online seed selection module, order tracking module, environmental monitoring and early warning module, pest and disease identification module, market forecasting module, logistics scheduling module, product forecasting module, growth status assessment module, knowledge popularization module, intelligent planting guidance module, and plant adoption module;

[0054] The online seed selection module allows consumers to browse a wide variety of crops after entering the platform. This includes various common fruit and vegetable varieties, from traditional favorites like tomatoes, cucumbers, and strawberries. Consumers can learn about the characteristics of each variety in detail, such as growth cycle, flavor profile, and nutritional value. They can also choose from a large selection based on their preferences and needs. The platform provides detailed information on each crop, including its growth habits, suitable environment, and cultivation difficulty, such as sweetness preferences and tolerance for cultivation challenges, allowing consumers to choose their preferred crop. After selection, the platform promptly communicates the consumer's choice to the corresponding farmer, initiating a personalized planting service.

[0055] Order Tracking Module: Once farmers begin planting the crops selected by consumers, the platform provides comprehensive planting guidance. The intelligent agent develops personalized planting plans based on the chosen crops and local environmental conditions, covering details such as sowing time, planting density, fertilization cycle, and irrigation volume. During the planting process, consumers can track their order status in real time through the platform. Every key stage, from seed sowing, germination, and growth to fertilization, irrigation, and pest and disease control, is presented to consumers through images, text, and videos. Consumers can also see the estimated harvest and delivery times, keeping them informed about when their ordered fruits and vegetables will arrive. If any abnormalities occur during planting, such as sudden natural disasters affecting crop growth, the platform will update information promptly, allowing consumers to stay informed about the latest status of their orders.

[0056] Knowledge and Science Popularization Module: The platform features a dedicated knowledge and science popularization section, providing consumers with a wealth of agricultural knowledge. Content includes the nutritional value of different fruits and vegetables, scientific storage methods, and healthy cooking suggestions. In addition, it introduces the crop cultivation process and the prevention and control of common diseases and pests, allowing consumers to gain a deeper understanding of the entire process of the fruits and vegetables they consume from farm to table, enhancing their knowledge and trust in agricultural products, and cultivating healthy eating habits.

[0057] Intelligent Planting Guidance Module: Farmers can access precise intelligent planting guidance on the platform. The platform's intelligent system collects farmland-level environmental data, combines it with crop growth patterns and historical planting data, and develops personalized planting plans for farmers. For example, based on soil fertility, it intelligently recommends suitable fertilizer types and amounts; based on local climate conditions, it determines the optimal sowing and irrigation times. It also provides detailed planting operation steps and precautions to help farmers cultivate scientifically and improve crop yield and quality.

[0058] Environmental monitoring and early warning module: The platform can monitor and analyze the farmland environment, including key indicators such as soil moisture, temperature, pH, and air temperature, humidity, and light intensity. Once environmental data becomes abnormal, such as excessively high soil moisture potentially causing root rot, or excessively low temperatures potentially affecting crop growth, the platform will immediately send early warning information to farmers and provide corresponding countermeasure suggestions to help farmers take timely action to avoid or reduce losses.

[0059] Pest and Disease Identification Module: When farmers discover abnormal symptoms in their crops, they can diagnose them using the platform's pest and disease identification function. Farmers only need to upload photos or videos of the affected parts of the crop. The intelligent agent uses image recognition and deep learning technology to quickly and accurately identify the type of pest or disease and provide detailed control measures, including recommended pesticide types, application times, and methods. This greatly improves farmers' efficiency in controlling pests and diseases and reduces yield losses caused by them.

[0060] Market Forecasting Module: The platform utilizes big data analytics and artificial intelligence to predict agricultural product market trends. It analyzes supply and demand, price fluctuations, and other information for different fruit and vegetable varieties, providing farmers with market forecast reports. Farmers can use this information to adjust their planting plans, selecting varieties with high market demand and prices, avoiding blindly following trends that could lead to unsold produce, and ultimately increasing their profits.

[0061] Furthermore, the market forecasting function of the intelligent agent is realized through a full process of multi-source data fusion, intelligent model inference, and dynamic solution output, providing farmers with accurate market trend judgments and production adjustment suggestions. The specific steps are as follows:

[0062] I. Multi-dimensional market data collection and integration;

[0063] Intelligent agents acquire market-related data through multiple channels to build a comprehensive predictive data source, specifically including:

[0064] 1. External market dynamics data: Connect with publicly available data from third-party agricultural e-commerce platforms (such as Yimutian and Meituan Youxuan) to obtain cross-platform price trends and regional supply and demand differences for similar crops (such as the demand for citrus in North China is 30% higher in winter than in summer).

[0065] Integrate government-issued agricultural policies (such as green food subsidy policies and green channels for agricultural product transportation), weather warnings (such as typhoons may lead to reduced fruit production in the south), and consumer trend reports (such as the 42% annual increase in pre-order planting demand from families born after 1985 in first-tier cities).

[0066] 2. Crop-specific correlation data: Combining crop characteristics (such as growth cycle and storage tolerance) and regional adaptability (such as Xinjiang grapes fetching a 20% premium in the East China market due to their high sweetness) in the plant details table, as well as regional sales patterns in the experienced farmer database (such as leafy vegetables being easily damaged during the plum rain season in the south, requiring a reduction in planting volume in advance).

[0067] II. Data Preprocessing and Feature Engineering: Cleaning and transforming the collected raw data to extract key predictive features:

[0068] 1. Data cleaning and normalization;

[0069] Remove outliers (such as order volume jumps caused by system failures) and fill in missing values ​​(such as supplementing weekly sales data with historical averages of the same crop from the same production area).

[0070] Standardize data in different formats (e.g., unify price / kg and sales volume / ton into output value units of 10,000 yuan / hectare, and convert subsidy ratios in policy documents into quantitative influencing factors).

[0071] 2. Core feature extraction;

[0072] Extracting key dimensions that influence the market from the data, including:

[0073] Time characteristics: crop market cycle (e.g., strawberries are in peak season from December to April of the following year), holiday effect (e.g., citrus sales increase by 50% before the Spring Festival);

[0074] Regional characteristics: different consumption preferences in different provinces (e.g., the demand for apples in the north is 1.5 times that in the south), and differences in logistics costs (transportation costs in remote areas may drive up the final price).

[0075] External impact characteristics: the impact of climate disasters on supply (such as drought leading to a 10% reduction in wheat production, which will push up prices), and the stimulus of policies on demand (such as rural revitalization consumption vouchers driving up sales of local specialty agricultural products).

[0076] III. Market forecasting inference through multi-model fusion;

[0077] Based on the preprocessed feature data, market trend predictions are generated through multi-level model inference:

[0078] 1. Basic Trend Forecast: LSTM temporal neural network is used to model historical price and sales data of a single crop to predict its basic trend over the next 3-6 months. For example, inputting monthly sales data for Xinjiang grapes over the past 5 years, the output will be a preliminary forecast that sales will increase by 15% year-on-year from August to October 2025.

[0079] 2. Multi-factor collaborative analysis: Models such as random forest and gradient boosting tree (GBDT) are introduced to analyze the comprehensive impact of multiple features on the market. For example, winter temperatures in North China, the yield of citrus-producing areas, and promotional activities on e-commerce platforms are used as input variables to predict fluctuations in the terminal price of citrus (e.g., a sudden drop in temperature leading to increased transportation costs may cause prices to rise by 8%). The prediction results are adjusted by combining policy factors (e.g., organic certification subsidies). For example, when policies are favorable, the market acceptance of organic vegetables increases, and the predicted premium space expands from 10% to 15%.

[0080] 3. Revision based on experienced farmers' experience and knowledge base: The model's predictions are revised by incorporating regional market patterns from experienced farmers' experience base (e.g., leafy vegetable prices inevitably rise after the rainy season in southern China) and crop substitution effects from the agricultural knowledge base (e.g., when tomato prices are too high, consumers will switch to substitutes such as eggplant and cucumber). For example, the model initially predicted a 10% increase in tomato sales in the summer of 2025. However, based on experienced farmers' experience that heavy summer rains can lead to a 20% increase in tomato transportation losses, the sales forecast was ultimately lowered to 5%.

[0081] IV. Segmentation and dynamic optimization of prediction results;

[0082] The intelligent agent breaks down the prediction results into multiple dimensions and updates them dynamically based on real-time data to ensure accuracy:

[0083] 1. Market segmentation forecast;

[0084] Breaking it down by region: For example, it is predicted that the demand for apples in Shandong will increase by 8%, while the demand in Gansu will only increase by 3% due to increased local production capacity;

[0085] Segmented by consumer group: For example, families born after 1985 have a 40% higher willingness to pay for fully visualized crop adoption than other groups, so it is recommended to increase the supply of such products.

[0086] 2. Real-time data dynamic adjustment;

[0087] Market data is updated every 72 hours (such as real-time monitoring of order increases on the platform and price changes on third-party platforms). If a prediction deviation is detected (such as actual sales being more than 5% lower than the prediction), the model is automatically retrained and the prediction parameters are adjusted. For example, when it was detected that transportation of Xinjiang grapes was disrupted due to a sudden outbreak of the epidemic, the sales forecast for the next two weeks was immediately lowered (from 10% to 3%).

[0088] V. Generation and output of production adjustment plans;

[0089] Based on the prediction results, the intelligent agent generates specific production adjustment suggestions for farmers to ensure a match between production and sales:

[0090] 1. Planting polarization optimization;

[0091] Variety adjustment: If it is predicted that the sales of specialty citrus fruits (such as Papa mandarin oranges) will increase by 25% in 2025, it is recommended that farmers reduce the planting area of ​​ordinary citrus fruits by 10% and switch to planting Papa mandarin oranges.

[0092] Adjustment of market launch time: If it is predicted that cherry prices will be highest in mid-May, it is recommended to sow 7 days earlier, taking into account the crop growth cycle, to ensure that the cherry is launched just in time for the peak price period.

[0093] 2. Sales strategy recommendations;

[0094] Pricing guidance: Given that the supply and demand gap for strawberries in North China during winter can reach 30%, it is recommended that farmers increase the wholesale price by 15% and reserve 20% of the yield for adoption services (which command higher premiums).

[0095] Channel adaptation: If the demand for small-packaged vegetables is expected to increase on community group-buying platforms, it is recommended that farmers repackage vegetables into 500g portions and prioritize connecting with community group-buying channels.

[0096] 3. Risk warning and response;

[0097] In response to potential risks predicted (such as typhoons potentially leading to reduced banana production in September and increased price volatility), contingency plans are proposed: such as advising farmers to purchase agricultural insurance in advance or sign guaranteed purchase agreements with processing companies (to avoid losses from a sharp drop in prices).

[0098] Through the above steps, the intelligent agent's market forecasting not only achieves basic judgments on sales volume and price, but also combines multiple dimensions such as crop growth cycle, regional characteristics, and policy changes to output production adjustment plans that can be directly implemented. This helps farmers avoid unsold produce caused by following trends in planting and achieves precision agricultural production based on sales.

[0099] Logistics Scheduling Module: To ensure the freshness of fruits and vegetables, the platform is responsible for efficient logistics scheduling. Based on order distribution and delivery time requirements, it rationally arranges logistics vehicles and delivery routes. Utilizing an intelligent logistics system, it monitors the location and transportation status of logistics vehicles in real time, ensuring that agricultural products are delivered to consumers on time and safely. Simultaneously, it adjusts logistics plans promptly based on factors such as weather. If severe weather may affect transportation, it arranges alternative routes or adjusts delivery times in advance to ensure smooth delivery.

[0100] Product forecasting module: The platform uses YOLOv8 technology to predict crop yields. By analyzing images of crop growth processes, it identifies information such as crop quantity and growth status, and combines this with historical data and environmental factors to predict the final yield. This function not only helps farmers understand their planting results in advance and rationally plan their sales, but also provides important reference for the platform to match production and sales and schedule logistics, improving the overall operational efficiency of the platform.

[0101] Growth Status Assessment Module: The platform's intelligent agent continuously assesses the growth status of crops. Taking into account environmental data, pest and disease conditions, growth cycles, and other factors, it quantitatively analyzes the health status and growth rate of crops. Through the assessment results, problems in the crop growth process are identified in a timely manner, and targeted suggestions and solutions are provided to farmers to ensure that crops grow in optimal conditions, thereby improving the quality and yield of agricultural products.

[0102] Furthermore, the intelligent agent's assessment of crop growth status is achieved through a complete process: multi-source data acquisition, growth stage localization, multi-dimensional index quantification, and dynamic result output. The specific steps are as follows:

[0103] I. Real-time acquisition and integration of multi-dimensional growth data;

[0104] The intelligent agent collects comprehensive data related to crop growth through hardware devices and platform data interfaces, providing a basis for evaluation.

[0105] 1. Environmental Factor Data Acquisition: Relying on a distributed IoT sensor network, real-time data on soil (humidity, pH value, nitrogen, phosphorus and potassium content, error <1% RH), air (temperature, humidity, CO2 concentration, error <0.5℃), light (photosynthetically active radiation), and meteorological (wind speed, precipitation) is acquired and updated at a frequency of once per minute to ensure data timeliness. For example, during the strawberry seedling stage, it is necessary to monitor in real time whether the soil moisture is maintained at 60%-70% and whether the air temperature is within the range of 15℃-20℃.

[0106] 2. Crop morphology and physiological data collection: Crop images are collected through high-definition field cameras, and the intelligent agent identifies and extracts key morphological indicators: plant height, stem diameter, number / area of ​​leaves, number / size of fruits (such as the diameter of a single tomato during the fruiting period and the volume of an apple during the fruit enlargement period).

[0107] 3. Integration of growth cycle and management data: The system calls upon crop biological characteristic data from the plant details table, including growth cycle (e.g., strawberry from sowing to harvest takes about 90-120 days) and key growth nodes (e.g., rice tillering stage and fruit tree flowering stage); it also links with field management operations recorded on the platform (e.g., fertilization time, irrigation amount, and pest and disease control measures) to analyze the impact of management behavior on growth (e.g., changes in leaf nitrogen content 7 days after fertilization).

[0108] II. Precise positioning of growth stages and matching of benchmark parameters;

[0109] The agent combines the crop growth cycle with real-time data to determine the current stage and match the standard growth parameters for that stage.

[0110] 1. Dynamic determination of growth stage;

[0111] Verification is achieved through a combination of planting cycle records and image recognition.

[0112] Basic determination: The initial stage is determined based on the sowing time and the inherent growth cycle of the crop (e.g., corn enters the jointing stage 30 days after sowing);

[0113] Precise calibration: The stage determination is corrected by using morphological features identified by the large model (such as wheat plant height ≥30cm and obvious stem nodes during the jointing stage) to avoid cycle deviations caused by environmental differences (such as adjusting the stage positioning by the number of leaves when low temperature delays growth).

[0114] By combining physiological data recorded in plant status tables, such as leaf color (chlorophyll content determined by spectral analysis), number of flowers, and fruit set rate, crop growth vitality can be quantified.

[0115] 2. Stage-specific baseline parameter call;

[0116] Retrieve the standard growth indicators for the current stage from the three-dimensional threshold system of crop-growth stage-environment parameters:

[0117] Environmental benchmarks: For example, during the apple enlargement period, the soil moisture should be 60%-70% and the light intensity should be ≥30000 lux;

[0118] Morphological criteria: For example, the sugar content of grapes during the coloring stage should be ≥16°Brix, and the number of berries per bunch should be ≥40;

[0119] Physiological benchmarks: For example, the SPAD value of chlorophyll in tomato leaves during the fruiting period should be ≥50 (indicating sufficient nutrition).

[0120] III. Quantitative evaluation of multi-dimensional growth status indicators;

[0121] The agent quantitatively assesses its growth status by comparing real-time data with baseline parameters across three dimensions: environmental adaptability, morphological development, and physiological health.

[0122] 1. Environmental adaptability assessment;

[0123] By comparing real-time environmental data with stage benchmarks using a rules engine:

[0124] Single-factor assessment: If soil moisture is 10% lower than the baseline (e.g., strawberry seedling humidity <54%), it is marked as mild water stress; if air temperature is higher than the upper limit of the baseline for 3 consecutive hours (e.g., tomato fruiting stage temperature >35℃), it is marked as high temperature stress risk.

[0125] Multi-factor collaborative assessment: The gradient boosting tree (GBDT) algorithm is used to analyze the interaction effects. For example, the combination of high temperature (>30℃) + high humidity (>80%) may cause diseases. Even if a single factor does not exceed the standard, it is still judged as an unsuitable environment.

[0126] 2. Morphological development progress assessment;

[0127] Based on the morphological indicators extracted from the large model, the development coefficient of the actual value / benchmark value is calculated:

[0128] If the actual weight of a single apple at maturity is 220g, and the benchmark value is 250g, the development coefficient is 0.88 (which is considered to be slightly delayed in development).

[0129] Calculate the average daily growth rate for dynamic indicators (such as the actual growth rate) and compare it with the standard growth rate (such as the average daily weight gain during the grape growth period should be ≥2g). If it is less than 50% of the standard, it is marked as slow growth.

[0130] 3. Assessment of physiological health status;

[0131] Pest and disease risk: The health level is assessed by detecting the presence of lesions (such as yellow spots from rice blast) and insect holes (such as aphid damage) on leaves / fruits using image recognition technology, combined with environmental data (such as high humidity which can easily lead to gray mold).

[0132] Nutritional status: Determine whether there is a nutrient deficiency by analyzing the soil nitrogen, phosphorus and potassium content and leaf spectral data (e.g., yellowing leaves may indicate nitrogen deficiency). Refer to the "China Major Crop Fertilization Guidelines" to give a nutrient adequacy score (0-100 points, <60 points require topdressing).

[0133] IV. Output and Dynamic Adjustment of Comprehensive Evaluation Results;

[0134] The intelligent agent integrates multi-dimensional evaluation results, generates visual reports, and drives the optimization of management solutions.

[0135] 1. Quantitative assessment report generation;

[0136] Output growth status score (0-100 points + key indicator details):

[0137] Scoring dimensions: Environmental adaptability (30%), morphological development (40%), and physical health (30%).

[0138] Example of details: Strawberry fruiting period assessment: 82 points - good environmental adaptability (soil moisture 65%), slightly delayed morphological development (average weight of single fruit is 8% lower than the benchmark), no risk of disease or pests.

[0139] 2. Anomaly warning and cause tracing;

[0140] For scores below 70 or single indicators deviating from the benchmark by more than 15%, an alert will be triggered and the cause analyzed.

[0141] If the tomato plant height development coefficient is 0.7, check the management records to see if it was caused by insufficient fertilizer during the seedling stage (less than 30% of the baseline) or light intensity being lower than the requirement for 5 consecutive days.

[0142] Warning levels are categorized as follows: Minor anomaly (push notification to pay attention), Moderate anomaly (push adjustment suggestions, such as increasing irrigation volume to 1.2 times the baseline), and Severe anomaly (linking to expert Q&A function, with a response within 10 minutes).

[0143] 3. Dynamically adapt management solutions;

[0144] Adjust the planting plan based on the assessment results:

[0145] When growth is slow: increase the amount of topdressing (e.g., increase the amount of nitrogen and potassium fertilizer by 10%) and extend the supplemental lighting time (from 6 hours / day to 8 hours / day).

[0146] When the environment is unsuitable: trigger the drip irrigation system during drought and turn on the shade net during high temperatures to ensure that subsequent growth returns to the baseline.

[0147] V. Continuous iterative optimization of the evaluation model;

[0148] The agent periodically correlates evaluation results with actual growth feedback (such as final yield and quality) and updates baseline parameters and evaluation algorithms through reinforcement learning.

[0149] If it is found that apples in a certain production area grow better at a soil pH of 6.0 than at the baseline (pH 6.5), the environmental baseline for that area will be dynamically adjusted.

[0150] For new crops (such as the niche variety loquat), the adaptation period can be shortened by reusing the evaluation model of similar crops through transfer learning.

[0151] Through the above steps, the intelligent agent achieves a closed loop of real-time monitoring, accurate assessment, and dynamic intervention of crop growth status, providing farmers with practical management suggestions and key data for the platform's yield forecasting and market prediction.

[0152] Plant Adoption Module: This module builds a bridge for users to deeply experience rural life, allowing them to have their own private online garden without having to physically visit the fields. Specifically, it includes the following:

[0153] 1. Orchard Type and Landform Selection: The platform offers users a wide variety of orchard types, from tropical mango and lychee orchards to fragrant apple and vineyard orchards, allowing users to choose freely. Furthermore, users can select different landforms based on their personal preferences, from fertile, flat black soil to gently rolling hills, satisfying their diverse imaginations of rural life.

[0154] 2. Plant Variety Selection: Covering a wide range of common and specialty crop varieties, from sweet and delicious strawberries and juicy watermelons to nutritious blueberries and sweet-tasting corn, users can choose according to their own taste preferences and planting interests to create their own personalized orchard.

[0155] 3. Adoption Rights Purchase: After selecting their preferred orchard type, land appearance, and plant varieties, users can purchase the corresponding adoption rights. The purchase process is simple and convenient, and the platform provides multiple secure payment methods to ensure the safety of user funds. Upon successful adoption, the user officially becomes the owner of that piece of cloud-based land.

[0156] 4. Regular Growth Monitoring: Leveraging advanced IoT technology and intelligent monitoring equipment, the platform collects plant growth data in real time, including temperature, humidity, light intensity, and soil fertility. Users can regularly receive plant growth reports via mobile phone or computer, and also view high-definition images and videos, intuitively understanding each stage of the plant from seedling to maturity, as if personally caring for it in the field.

[0157] Furthermore, the intelligent agent also uses data analysis to determine whether the environment is suitable for crop growth; it integrates crop variety characteristics, local climate and soil conditions, and seasonal changes to customize exclusive planting plans for different crops; it formulates precise irrigation and fertilization plans based on regional differences, combined with water resource status and soil fertility; and it identifies damaged fruits and provides treatment solutions.

[0158] Furthermore, determining whether an environment is suitable for crop growth through data analysis specifically includes the following:

[0159] I. Real-time acquisition of high-precision environmental data;

[0160] Multi-dimensional data acquisition terminal deployment: Relying on the integrated application of 5G technology and the BeiDou positioning system, a distributed IoT sensor network is deployed in farmland, including soil sensors (monitoring humidity, pH, and EC value), air sensors (monitoring temperature, humidity, and CO2 concentration), light sensors (monitoring photosynthetically active radiation), and weather stations (monitoring wind speed and precipitation). All sensors sample at a frequency of 1 time per minute, with data acquisition errors controlled within <0.5℃ (temperature) and <1% RH (humidity), ensuring high accuracy of raw data. Simultaneously, BeiDou positioning enables field-level spatial calibration of data, accurately linking data with specific planting areas and providing spatial coordinate benchmarks for subsequent regional analysis.

[0161] II. Construction of a crop growth suitability parameter database;

[0162] 1. Setting basic parameter thresholds;

[0163] Based on agricultural knowledge bases (such as the "Guidelines for Fertilization of Major Crops in China" and the "Handbook for Promoting Successful Vegetable Planting Experiences"), crop variety characteristic databases (covering growth cycle parameters of 200+ common fruits and vegetables), and experienced farmers' experience databases, a three-dimensional threshold system of crop-growth stage-environmental parameters is constructed. For example: the suitable soil moisture for strawberry seedlings is 60%-70%, air temperature is 15℃-20℃, and light intensity is 20000lux-30000lux, where lux is a unit of light intensity; the suitable soil pH for tomato fruiting is 6.0-7.0, and air humidity is 50%-60%.

[0164] 2. Dynamic parameter optimization;

[0165] By combining historical planting data (correlation analysis of crop yield and environmental data in different regions over the past 3 years) and machine learning models (random forest algorithm, LSTM time series model), the basic threshold is dynamically adjusted, and environmental trends (such as a sudden drop in temperature over 3 consecutive hours) are analyzed to predict environmental risks in the next 2 hours. For example, the soil moisture threshold for the same crop in the rainy southern region will be slightly lower than that in the arid northern region, and the model can automatically adapt to regional differences.

[0166] III. The core judgment logic of data analysis;

[0167] 1. Real-time data preprocessing: The collected raw data is cleaned (outliers are removed, such as jump data caused by sensor failure) and normalized (different unit parameters are converted into standardized values ​​in the 0-1 range) by edge computing nodes, and then uploaded to the cloud model via 5G network.

[0168] 2. Multi-level suitability assessment:

[0169] The first layer, rule engine quick screening: directly compares real-time data with the threshold range in the parameter library. If a single dimension of data exceeds the threshold (e.g., soil moisture is too low <50%), it is immediately marked as a mild anomaly.

[0170] The second layer, the multi-factor collaborative analysis model, uses the gradient boosting tree (GBDT) algorithm to analyze the interaction of multiple parameters (such as high temperature + high humidity may cause diseases). For example, when the air temperature is >30℃ and the humidity is >80%, even if a single parameter does not exceed the standard, the model will still determine it as a "potential risk environment".

[0171] The third layer, growth stage adaptation adjustment: Combines the current growth stage of the crop (confirmed through planting cycle records and image recognition, such as seedling stage / flowering stage / fruiting stage), calls the specific parameter threshold of the corresponding stage to avoid misjudgment caused by general thresholds (such as the water requirement of wheat during the grain filling stage being higher than that during the seedling stage).

[0172] IV. Abnormal Early Warning and Dynamic Feedback Mechanism;

[0173] 1. Classification of early warning levels;

[0174] The system automatically classifies warning levels based on the degree of data deviation: Slight deviation (e.g., humidity less than 5% below the suitable value): A notification is sent to remind farmers to monitor the situation closely; Moderate deviation (e.g., temperature 3-5℃ above the suitable range): Adjustment suggestions are sent, such as turning on shade nets or irrigation systems; Severe deviation (e.g., soil pH < 4.5, which may lead to root necrosis): An "emergency warning" is triggered, and farmers and technical experts are contacted simultaneously.

[0175] 2. Model self-optimization and iteration;

[0176] The system regularly compares the actual crop growth feedback (such as yield and disease incidence) after the warning with the model's judgment results, and updates the parameter library thresholds and multi-factor weights through reinforcement learning algorithms, so that the judgment accuracy continues to improve with the accumulation of data.

[0177] Through the above, the data analysis model has achieved full automation of the process from data collection to intelligent judgment to precise early warning. It not only ensures the real-time nature of the judgment, but also ensures the accuracy of the suitability judgment through the deep integration of crop specificity and environmental complexity, providing farmers with a basis for environmental adjustments that can be directly implemented.

[0178] Furthermore, taking into account crop variety characteristics, local climate and soil conditions, and seasonal variations, customized planting plans are developed for different crops, including the following:

[0179] I. Multi-dimensional basic data collection and integration;

[0180] 1. Construction of Crop Variety Characteristic Database: Based on the plant details table, integrate the core biological characteristic data of crops, including:

[0181] (1) Growth cycle parameters (e.g., strawberries take about 90-120 days from sowing to harvest, while cherries take 3-5 years to bear fruit).

[0182] (2) Environmental sensitivity thresholds (e.g., Hanyuan cherries are suitable for a diurnal temperature range of ≥10℃, and Xinjiang grapes require ≥2800 hours of sunshine per year).

[0183] (3) Resistance characteristics (e.g., some citrus varieties are highly cold-resistant and can tolerate temperatures as low as -5℃).

[0184] (4) Cultivation characteristics (e.g., tomatoes need to be trellised, and planting density is related to row spacing).

[0185] 2. Real-time collection of regional climate and soil data: Relying on 5G+BeiDou fusion technology, basic environmental data of the planting area is collected, including the following data:

[0186] (1) Soil data: pH value (e.g., pH of southern red soil is about 4.5-5.5), organic matter content, and nitrogen, phosphorus and potassium levels are obtained through soil sensors;

[0187] (2) Topographic features: Combine the altitude and slope information of Beidou positioning (e.g., drainage needs to be considered to adjust the planting density in mountain orchards).

[0188] 3. Incorporation of seasonal dynamic factors: Seasonal modules are divided according to phenological periods (e.g., dry season / rainy season in South China, spring, summer, autumn and winter in the North), and typical climate parameters of each season are associated (e.g., frequency of cold waves in spring, typhoon path in summer), and mapped to the environmental requirements of key crop growth nodes (e.g., wheat greening period, fruit tree flowering period).

[0189] II. Scheme generation for intelligent decision-making models;

[0190] 1. Basic Solution Framework Generation: Based on a rule engine and federated learning model, with crop variety characteristics as the core, and matching the regional environmental baseline, it includes the following:

[0191] (1) Sowing time: Combine the crop germination temperature threshold with the regional spring temperature rise curve (e.g., corn needs to be sown when the soil temperature is stable at ≥10℃) to calculate the optimal window period (error ≤3 days).

[0192] (2) Planting density: Calculated based on soil fertility level (e.g., high fertility plots can be planted more densely), crop plant type (e.g., dwarf crops can be planted 20% more densely than tall crops) and row spacing formula (plant spacing = crop mature plant width × 0.8).

[0193] (3) Integrate the experience database of veteran farmers (such as shallow water to promote tillering during the three-leaf stage of rice) and the knowledge base ("Guidelines for Fertilization of Major Crops in China") to generate basic processes such as irrigation cycle (such as watering frequency of sandy soil is 1.5 times higher than that of clay soil) and fertilization type (such as increasing phosphorus and potassium fertilizer during flowering).

[0194] 2. Regional Adaptability Optimization: By optimizing the locally deployed AI model, the basic solution is modified for different regions, such as:

[0195] (1) In arid areas of Northwest China, the irrigation scheme will be automatically adjusted to drip irrigation + mulching to retain moisture, and the frequency of soil moisture monitoring will be increased;

[0196] (2) In the acidic red soil region of the south, when planting alkali-loving crops (such as cotton), the program automatically adds a pretreatment step of applying quicklime to adjust the pH value to 6.5-7.0.

[0197] 3. Dynamic Mitigation of Seasonal Risks: Based on seasonal climate risk databases (such as late spring frosts in North China and the plum rain season in the Yangtze River Basin), preventative measures are embedded in the plan, including:

[0198] (1) If a low temperature risk is predicted during the planting period, the sowing time will be automatically postponed by 3-5 days, or it is recommended to use plastic film to cover the ground for insulation.

[0199] (2) Before the rainy season arrives, for crops that are susceptible to waterlogging (such as chili peppers), the plan adds field management items such as ridge height ≥30cm and digging drainage ditches.

[0200] III. The scheme's dynamic adjustment mechanism throughout its entire lifecycle;

[0201] 1. Real-time adaptation to growth stages: Based on the crop growth stages recorded in the plant status table (e.g., seedling stage / flowering stage / fruiting stage), the agent updates the management plan every 72 hours, such as:

[0202] (1) The fertilizer concentration is automatically reduced during the seedling stage (50% lower than that during the mature plant stage) to avoid burning the seedlings;

[0203] (2) During the fruiting period, the amount of topdressing fertilizer is dynamically adjusted according to the number of fruits identified by the YOLOv8 model (the amount of nitrogen and potassium fertilizer increases by 5% for every 10% increase in fruit yield).

[0204] 2. Environmental Anomaly Response Adjustment: When environmental monitoring data (such as soil moisture and temperature) deviates from the suitable range, the plan automatically triggers emergency adjustments, such as:

[0205] (1) If there is insufficient light for 3 consecutive days (<60% of the crop's needs), the supplementary lighting time will be extended to 8 hours / day.

[0206] (2) If the soil EC value is too high (≥2.5 mS / cm), immediately send a correction instruction to suspend fertilization and flood irrigation to wash away salt.

[0207] 3. User interaction optimization channel: Farmers provide feedback on actual planting conditions (such as plot area and existing agricultural machinery type), and the intelligent agent further optimizes the plan based on the feedback, such as:

[0208] (1) If farmers do not have drip irrigation equipment, the irrigation plan will be automatically adjusted to furrow irrigation + timely loosening of soil to retain moisture;

[0209] (2) When planting in small plots, replace the mechanized operation steps (such as drone plant protection) with specific manual operation guidelines (such as spraying the front and back of the leaves evenly when using a manual sprayer).

[0210] 4. Automated field management process: The solution is broken down into an executable task chain and automatically pushed to the farmer's app.

[0211] Through the above mechanism, the planting plan generated by the intelligent agent not only ensures scientific standardization (based on authoritative knowledge base and data model), but also achieves full-dimensional adaptation of variety, region, season and farmer conditions. The final output plan includes directly executable quantitative indicators (such as 2.5 kg of seed per mu, plant spacing of 40 cm × row spacing of 60 cm) and operation steps, which farmers can implement without professional knowledge.

[0212] Furthermore, based on regional differences and considering water resource conditions and soil fertility, precise irrigation and fertilization plans are formulated, specifically including the following:

[0213] I. Multi-dimensional basic data collection and regional feature modeling;

[0214] 1. Regionally differentiated data collection:

[0215] (1) Soil fertility data: Soil nitrogen, phosphorus and potassium content (accuracy ±5mg / kg), organic matter ratio and pH value (error ≤0.1) are collected in real time by soil sensors (deployment density up to 1 per 5 mu). Combined with regional soil survey data (such as organic matter content in the Northeast black soil area is generally >3%, and in the Southern red soil area it is mostly <1%), a regional baseline of soil fertility is established.

[0216] (2) Water resources data: Integrate regional water bureau data (such as annual precipitation, groundwater depth, and irrigation water source type) with real-time monitoring (such as river / reservoir water level and water pipeline flow), classify water resources abundance levels (such as moderate water shortage in the North China Plain and abundant water in the Jiangnan water towns), and link them to irrigation costs (such as energy consumption cost of groundwater extraction and loss rate of long-distance water transmission).

[0217] (3) Crop water and fertilizer requirements: Based on the biological parameters of crop varieties in the plant details table, such as the nitrogen requirement of wheat during the jointing stage (about 3-5 kg / mu), the potassium requirement of grapes during the fruit expansion stage (about 8-10 kg / mu), and drought / fertility tolerance characteristics (such as millet being 40% more drought tolerant than corn), a crop requirement database is established.

[0218] II. Generation of precision irrigation plans;

[0219] 1. Dynamic water demand calculation model: The agent calculates crop water demand in real time based on the Penman-Monteith formula (refer to "Irrigation and Drainage Engineering") and the following parameters:

[0220] (1) Basic parameters: crop coefficient (e.g., rice 1.1-1.3, cotton 0.7-0.9), evapotranspiration (derived from air temperature and humidity, light intensity);

[0221] (2) Regional correction: Water resource abundance level (automatically reduce irrigation threshold by 20% in water-scarce areas), soil water retention capacity (sandy soil has poor water retention, so set a higher water replenishment frequency). For example, when planting corn in arid areas of Northwest China, the model will reduce the lower limit threshold of soil moisture from the usual 60% to 50%, and prioritize drip irrigation (which saves 40% more water than flood irrigation).

[0222] 2. Irrigation strategy generation:

[0223] (1) Method adaptation: Select irrigation method according to the type of regional water source (e.g., sprinkler irrigation in areas with sufficient surface water, and micro-irrigation + mulching to retain moisture in areas with scarce groundwater).

[0224] (2) Time scheduling: Combine meteorological forecasts (such as delaying irrigation if there is rainfall in the next 3 days) with crop growth stages (such as wheat grain filling stage, which requires alternating shallow and wet conditions, that is, wetting until the soil moisture reaches 70% and then letting it dry to 50% before replenishing water).

[0225] (3) Quantitative execution: Output specific irrigation parameters (such as drip irrigation flow rate 2L / h·plant, continuous for 1.5 hours, with an interval of 3 days), and automatically execute the irrigation equipment through the Internet of Things control.

[0226] III. Generation of Precision Fertilization Plans;

[0227] 1. Dynamic calculation of fertility requirements: Based on the nutrient balance method (refer to the "Guidelines for Fertilization of Major Crops in China"), the fertilizer application rate is calculated by combining the following factors:

[0228] (1) Crop requirements: Determine the target nutrient uptake based on crop variety (e.g., the potassium requirement of tomatoes during the fruiting period is 1.2 times that of nitrogen) and growth stage (the nitrogen requirement during the seedling stage accounts for 30% of the total cycle);

[0229] (2) Soil supply: Calculate the natural supply ratio using current soil fertility data (e.g., if the available nitrogen content in the soil is >80mg / kg, reduce the amount of nitrogen fertilizer by 30%).

[0230] 2. Fertilization strategy generation:

[0231] (1) Fertilizer type matching: Select according to soil type (e.g., use slow-release fertilizer to reduce leaching loss in sandy soil, and use fast-acting fertilizer to improve absorption efficiency in clay soil).

[0232] (2) Application method: Combine the local agricultural machinery conditions (e.g., drones are used for spreading in plains areas and hole application is used in mountainous areas), output the fertilizer type + dosage + time + method combination scheme (e.g., Northeast soybean field: 15 kg / mu of diammonium phosphate (strip application before sowing + 8 kg / mu of potassium chloride (topdressing during flowering)).

[0233] (3) Environmental constraints: For ecologically sensitive areas (such as water source protection areas), automatically reduce the amount of nitrogen fertilizer used (reduce non-point source pollution) and recommend adding urease inhibitors (reduce ammonia volatilization by 30%).

[0234] IV. Regional adaptation and dynamic optimization mechanism;

[0235] 1. Dynamic Correction for Regional Differences: The country is divided into 6 major agricultural ecological zones (such as the Huang-Huai-Hai Plain and the middle and lower reaches of the Yangtze River), and a localized model is deployed for each region. For example:

[0236] (1) Southern rainy areas: The fertilization program automatically adds the logic of "reapplying after rain" (to avoid nutrient leaching loss);

[0237] (2) Northwest arid region: Irrigation schemes incorporate the "water and fertilizer integration" strategy (fertilizer is applied with irrigation water, increasing absorption efficiency by 20%).

[0238] 2. Real-time data feedback and iteration: The intelligent agent collects soil data (fertility, moisture) every 3 days from IoT devices, combines this data with crop growth status (such as leaf nitrogen content monitored by spectral analysis), and compares the actual values ​​with the expected values ​​of the plan.

[0239] (1) If the soil nitrogen content is lower than the expected 10%, 5% nitrogen fertilizer will be automatically added;

[0240] (2) If the soil moisture decreases faster than the model predicts 3 days after irrigation, the next irrigation amount will be increased by 15% (to adapt to changes in soil water retention capacity).

[0241] Through the above mechanisms, precision irrigation programs can increase water resource utilization by 35%-50% (compared to traditional flood irrigation), and fertilization programs can reduce fertilizer waste by more than 25%. At the same time, by adapting to regional characteristics, resource misallocation caused by a one-size-fits-all approach can be avoided (such as reducing irrigation energy consumption in southern regions and reducing the risk of soil compaction caused by excessive fertilizer in northern regions).

[0242] Furthermore, the pest and disease identification module specifically includes the following:

[0243] I. Multidimensional Construction of a Pest and Disease Characteristics Database;

[0244] To support accurate identification, the intelligent agent first constructed a feature database covering common crop diseases and pests, with data sources including:

[0245] (1) Integration of authoritative agricultural knowledge base: Includes the characteristics of diseases and pests in literature such as "Handbook for Specialized Prevention and Control of Crop Diseases and Pests" and "Encyclopedia of Chinese Agriculture: Crop Volume", such as the yellow oval lesions of rice blast with gray mold layer, and the typical characteristics of aphids such as cluster damage and leaf curling. It also marks the host crop (such as wheat scab and tomato late blight), peak season, affected parts (leaves / fruits / roots) and morphological parameters (lesion size and insect size) of each disease and pest.

[0246] (2) Field image sample collection: The agricultural cooperative collects images of pests and diseases in different regions and at different growth stages, including high-resolution close-ups (such as the texture of lesions on the surface of fruits) and environmentally related images (such as the scene of mold growth in a humid environment), and marks the shooting conditions (light, crop variety, growth period).

[0247] (3) Digitalization of farmers’ experience: Through structured interviews, farmers’ experience in identifying regional pests and diseases (such as early water-soaked spots of anthracnose in citrus during the plum rain season in the south) is recorded and converted into supplementary feature tags in the database (such as the lesion expansion rate increases by 2 times when humidity is >85%).

[0248] II. Training and Deployment of Deep Learning-Based Pest and Disease Identification Model;

[0249] An improved ResNet-50 model focused on pest and disease identification is used, and the specific implementation is as follows:

[0250] 1. Customized and optimized model architecture;

[0251] (1) Feature enhancement mechanism: The CBAM attention module is added to ResNet-50 to enhance the extraction of key features such as the edge of lesions and the outline of insect bodies, and solve the interference problem of complex background (branch and leaf shading, fruit overlap) in the field.

[0252] (2) Lightweight design: The model parameters are compressed by 40% using depth-separable convolution, which is compatible with low computing power devices such as farmers' mobile phones and field smart terminals. The processing time for a single image is <0.8 seconds, which meets the real-time recognition requirements.

[0253] 2. Sample training and accuracy assurance;

[0254] (1) Perform data augmentation on the images (rotation, brightness adjustment, local occlusion simulation), expand the sample, classify and train according to pest and disease type + crop variety, and increase the sample proportion for high-incidence pests and diseases (such as strawberry gray mold) (increase to 20%).

[0255] (2) By using the confusion matrix to correct misjudgments of similar diseases and pests (such as distinguishing between early blight and late blight of tomatoes), the model recognition accuracy is improved and the misjudgment rate is reduced.

[0256] 3. Localized deployment and privacy protection;

[0257] The model is deployed on edge devices at the farmer's end (smart monitoring terminals, mobile apps). Fruit images uploaded by farmers are identified locally, without needing to be uploaded to the cloud, thus reducing network dependence and protecting data privacy.

[0258] III. Disease and pest identification and control decision-making for intelligent agents;

[0259] 1. Image acquisition and preprocessing;

[0260] Farmers upload images of diseased parts of their fruit through the platform (which supports autofocus and supplemental lighting for shooting). The intelligent agent performs the following preprocessing on the images:

[0261] (1) Remove redundant backgrounds and retain areas with pests and diseases accounting for ≥70%;

[0262] (2) Enhance the contrast of features (such as improving the color difference between lesions and normal peel) and highlight key features (such as mold layer and insect texture).

[0263] 2. Dimension identification and cross-validation;

[0264] (1) Model preliminary identification: output the type of pest and disease and the confidence level (e.g., grape downy mildew).

[0265] (2) Agent cross-validation: Confirm the recognition result by combining the following three types of data:

[0266] ① "Plant Status Table" (Current growth stage, such as the grape coloring stage being a high-incidence period for downy mildew);

[0267] ② Environmental monitoring data (e.g., air humidity > 85% in the past 3 days, meeting the conditions for disease outbreak);

[0268] ③ Pest and disease characteristic database (matching the gray-white mold layer of downy mildew + lesion characteristics on the back of leaves).

[0269] 3. Generation of targeted prevention and control plans;

[0270] The intelligent agent outputs a quantitative solution from a pest and disease-control measures association library (integrated from an authoritative knowledge base), specifically including the following:

[0271] (1) Chemical control: Determine the type of pesticide (e.g., dimethomorph is recommended for downy mildew), concentration (1000 times dilution), application method (foliar spray, focusing on the back of the fruit) and safety interval (14 days), and refer to "New Technologies for Pollution-Free Vegetable Cultivation and Pest and Disease Control".

[0272] (2) Physical / biological control: In response to the needs of green planting, alternative solutions are provided (such as using yellow sticky traps to kill aphids and releasing ladybug natural enemies, with 20 yellow sticky traps placed per acre).

[0273] (3) Emergency response: If the degree of damage reaches a severe level (the proportion of diseased fruit > 30%), the instructions to remove diseased fruit and disinfect the whole orchard will be pushed out simultaneously, and the expert Q&A function will be linked to respond to farmers' inquiries within 10 minutes.

[0274] IV. Dynamic optimization mechanism;

[0275] The system collects farmer feedback monthly (such as the effectiveness of prevention and control measures) and optimizes the model and measures based on actual field results.

[0276] (1) For pests and diseases with an accuracy rate of <80%, supplement the samples and retrain the model;

[0277] (2) Statistical analysis of the efficacy of the treatment plan (e.g., thiazolium zinc has an efficacy of 86% in preventing peptic ulcer disease), optimize the priority of the plan, and ensure that the recommended measures are efficient and feasible.

[0278] Through the above process, the intelligent agent realizes a closed loop of image acquisition, accurate recognition, solution push, and effect feedback, solving the pain points of farmers' difficulty in identifying diseases and blind application of pesticides. The current average recognition time is less than 12 seconds, and the loss rate of diseased fruit is reduced by 32% after the prevention and control solution is adopted.

[0279] Furthermore, the damaged fruit is identified, and a treatment plan is proposed, specifically including the following:

[0280] I. Construction of a database of characteristics of damaged fruit;

[0281] 1. Multi-dimensional damage feature recording;

[0282] By integrating relevant standards for "Agricultural Product Quality and Safety" and field practice data, a database of damaged fruit characteristics covering more than 60 crops has been constructed.

[0283] (1) Classification of damage types: Identify the following three core damage characteristics:

[0284] ① Disease damage (e.g., lesion shape: round / irregular, color: brown / black, degree of rot: surface / deep);

[0285] ② Insect damage (such as the number of insect holes, insect excrement residue, and the area of ​​fruit pulp damage);

[0286] ③ Physical damage (such as extrusion deformation rate, crack length, and mechanical scratch depth).

[0287] Crop-specific labeling: Combine plant details to record the damage tolerance characteristics of different crops (e.g., strawberry skin is 0.1mm thick, and juice will seep out when it is slightly squeezed; apple skin has a thick waxy layer and can withstand light impact), and label the variety-specific damage threshold (e.g., citrus skin damage diameter <3mm is considered light damage).

[0288] 2. Sample data collection and labeling;

[0289] Images of damaged fruit collected by the joint agricultural cooperative (covering different stages of damage, lighting conditions, and shooting angles), with annotation information including:

[0290] (1) Damage parameters (e.g., lesion area percentage 15%, extrusion deformation rate 20%).

[0291] (2) Related information (crop variety, harvest time, storage environment);

[0292] (3) Labels for processing results (e.g., minor damage - marketable, severe damage - for processing).

[0293] The dataset is divided into training and validation sets in an 8:2 ratio to provide a foundation for model training.

[0294] II. Training and Deployment of the Damaged Fruit Recognition Model;

[0295] An improved MobileNetV3 deep learning model is used, focusing on the scenario of damaged fruit identification.

[0296] 1. Model architecture optimization;

[0297] (1) Lightweight design: The depth-separable convolution reduces model parameters by 35%, making it compatible with low-computing-power devices such as farmers' mobile phones and field smart terminals. The processing time for a single image is less than 1 second, meeting the real-time recognition requirements.

[0298] (2) Feature enhancement mechanism: A coordinate attention module is added to the Neck layer of the model to enhance the feature extraction of the edge of the damaged area (such as the outline of the crack and the boundary of the lesion) and solve the problem of blurry recognition caused by the similar color of the fruit and the background (such as the slight scratch of the yellow lemon).

[0299] 2. Sample training and accuracy improvement;

[0300] (1) Perform data augmentation on the image (brightness adjustment, rotation, local occlusion simulation) to expand the sample and avoid model overfitting.

[0301] (2) Training was conducted according to “crop variety + damage type”. For high-value crops (such as cherries and blueberries), the sample weight was increased (the proportion was increased to 25%). The ability to distinguish similar damage was optimized by using a confusion matrix (such as distinguishing mechanical damage of apples from anthracnose spots). The final recognition accuracy reached 92%, and the damage degree judgment error was <5%.

[0302] 3. Localized deployment;

[0303] The model is deployed on edge devices at the farmer's end (such as smart monitoring terminals). After image acquisition, recognition is completed locally without uploading to the cloud, ensuring real-time response in weak network environments.

[0304] III. Damaged fruit identification by the intelligent agent;

[0305] 1. Image acquisition and preprocessing;

[0306] (1) Farmers upload fruit images through the platform APP (which supports automatic focus guidance to ensure clear imaging of damaged areas), or the field high-definition camera collects fruit growth images at regular intervals.

[0307] (2) The agent preprocesses the image: cropping redundant background (keeping fruit area ≥80%) and enhancing the contrast of damage features (e.g., increasing the brightness difference between the abraded area and the normal epidermis).

[0308] 2. Multi-dimensional identification and cross-validation;

[0309] (1) Model preliminary identification: The deep learning model outputs the damage results (e.g., "Strawberry - physical damage - crack length 3mm - damage percentage 8%").

[0310] (2) Intelligent agent cross-validation: Combine the "plant status table" (such as the susceptibility of fruit damage during the ripening period), environmental data (such as temperature fluctuations during transportation may cause frost damage) and historical processing records (such as common damage types of fruit of this variety) to perform secondary verification of the recognition results and reduce the misjudgment rate (misjudgment rate < 4%).

[0311] IV. Generation and Execution of the Processing Solution;

[0312] 1. Hierarchical processing rule engine;

[0313] The intelligent agent automatically matches a treatment strategy based on the degree of damage, with rules derived from the platform's supply chain data and the "Technical Specifications for Agricultural Product Processing".

[0314] (1) Slightly damaged (e.g., epidermal scratches <5mm, lesion area <10%): judged as edible grade, recommended to prioritize local immediate sales program, and it is suggested to mark the defective fruit and discount it by 10%-15%.

[0315] (2) Moderate damage (e.g., local rot <30%, compression deformation rate 15%-30%): determined to be processing grade, connected to jam and dried fruit processing plants, generating cold chain transportation instructions within 4 hours, and clarifying processing pretreatment requirements (e.g., use after removing the rotten part).

[0316] (3) Severely damaged (e.g., the whole fruit is rotten, and the area of ​​insect infestation is >50%): It is determined to be harmless treatment level. It is recommended to bury it deeply for decomposition (depth ≥50cm) or biodegradation, and trace the source of damage (e.g., prompt to check whether the humidity of the storage environment exceeds the standard).

[0317] 2. Full-process tracking and feedback;

[0318] The intelligent agent tracks the processing progress (e.g. raw materials have been delivered to the factory and defective fruit has been sold) through IoT devices, and collects downstream feedback (e.g. the utilization rate of raw materials by the processing plant and consumer evaluation of defective fruit). It updates the processing rule weights monthly (e.g., increasing the community group-buying adaptation priority of "slightly damaged fruit") and continuously optimizes the feasibility of the solution.

[0319] Through the above mechanism, the intelligent agent realizes closed-loop management of damaged fruits from identification to grading, processing and traceability, effectively reducing the loss rate of agricultural products. At the same time, the graded processing ensures the rational use of resources, and ultimately achieves a dual improvement in agricultural product quality and economic benefits.

[0320] Furthermore, the product forecasting module specifically includes the following:

[0321] I. Crop growth image data acquisition and annotation;

[0322] 1. Multi-scene image acquisition;

[0323] Relying on high-definition cameras and drone inspection systems deployed in the fields, images are collected regularly according to the crop growth cycle (seedling stage, flowering stage, fruiting stage, and maturity stage), covering different lighting conditions (sunny days, cloudy days, evenings), different planting densities, and different varieties (such as strawberries, citrus, and grapes). The collection frequency is dynamically adjusted according to the growth stage: once every 3 days before the fruiting stage, and once a day from the fruiting stage to the maturity stage, to ensure that the dynamic changes in the number and size of fruits are captured.

[0324] 2. Refined data annotation;

[0325] The acquired images undergo both manual annotation and automated verification, including the following:

[0326] (1) The labeling content includes the number of fruits, the diameter of a single fruit (pixel level), and the maturity (distinguished by color characteristics, such as green for immature and red for mature).

[0327] (2) Combine BeiDou positioning information with images and specific planting plots, and label the plot area, crop variety (such as "Hanyuan cherries" and "Xinjiang grapes") and other metadata;

[0328] (3) Construct a dataset of labeled samples, of which 80% is used for model training and 20% for validation.

[0329] II. Customized training and optimization of YOLOv8 models;

[0330] 1. Model structure optimization;

[0331] YOLOv8 was optimized for lightweight design and high accuracy to suit the specific needs of agricultural applications.

[0332] (1) Backbone replacement: MobileNetV3 is used as the feature extraction backbone network to reduce model parameters (compression of 40%), adapt to low computing power terminals such as farmers' mobile phones and field edge devices, and ensure real-time processing (single image recognition time < 0.3 seconds).

[0333] (2) Introduction of attention mechanism: Add SE (Squeeze-and-Excitation) attention channel to the Neck layer to enhance the extraction of key features such as fruit outline and color (e.g., distinguishing grape fruits from the background of branches and leaves), and improve the recognition accuracy in complex field environments.

[0334] (3) Hyperparameter tuning: Optimize parameters such as learning rate (initially set to 0.01, decayed by cosine annealing strategy) and IoU threshold (set to 0.65) through grid search to reduce the false negative rate of small fruits (such as cherries).

[0335] 2. Transfer learning and scenario adaptation;

[0336] First, pre-train the model on a general object detection dataset (such as COCO), then fine-tune it using a project-specific crop and fruit dataset, focusing on optimizing the ability to identify "overlapping fruits" and "occluded fruits" (such as fruits obscured by leaves in a bunch of grapes). For different crop characteristics (such as differences in fruit morphology: the creeping growth of strawberries vs. the hanging growth of apples), train sub-models for each crop (such as a strawberry-specific model and a citrus-specific model) to further improve recognition accuracy.

[0337] III. Calculation of Production Forecast;

[0338] 1. Extraction of fruit features from a single plant / region;

[0339] After the model performs inference and recognition on the input image, it outputs the following key data:

[0340] (1) Number of fruits: The total number of fruits identified in the statistical image (including ripe and immature fruits, which are marked separately).

[0341] (2) Fruit size: The diameter / volume of a single fruit is obtained by converting the pixel size to the actual distance (based on camera focal length and shooting distance calibration). For example, "the average diameter of an apple is 8.5cm".

[0342] (3) Fruit distribution density: Combine the image coverage area to calculate the number of fruits per unit area (e.g., "35 strawberries per square meter").

[0343] 2. Production conversion formula;

[0344] Based on crop variety characteristics and planting parameters, the estimated total yield is calculated using the following formula: Estimated total yield = Number of fruits per unit area × Total planting area × Average weight of a single fruit × Maturity correction coefficient.

[0345] (1) Average weight of a single fruit: based on crop variety database (e.g., “Red Fuji apple average single fruit weight 250g”, “Kyoho grape single fruit weight 10g”).

[0346] (2) Maturity correction coefficient: dynamically adjusted based on the current growth stage (e.g., the maturity rate is set to 80% 7 days before maturity) and historical maturity data (e.g., the maturity rate of this plot in previous years was 92%).

[0347] (3) Total planting area: The actual planting area of ​​the plot obtained through Beidou positioning.

[0348] 3. Error correction mechanism;

[0349] Environmental factors (such as sufficient sunlight and soil fertility) are introduced to correct the results: for example, if the soil nitrogen content in a certain area is found to be 15% lower than the standard value, the weight of a single fruit is calculated as 90%, and the prediction error is ultimately controlled within 5% (which meets the performance index of "prediction error < 5%" in the document).

[0350] Through the above process, the YOLOv8 algorithm achieves a precise transformation from "image recognition" to "yield quantification," providing farmers with scientific basis that can be directly used for production planning (such as arranging harvesting manpower in advance and connecting with sales channels), effectively solving the problem of traditional yield prediction relying on experience and having large errors.

[0351] The above description is merely a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and improvements, and can be altered within the scope of the concept described herein through the above teachings or related technologies or knowledge. Modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention should be within the protection scope of the appended claims.

Claims

1. A smart agricultural management system based on planting-harvesting-distribution full link, characterized in that: The system comprises an online seed selection module, an order tracking module, an environment monitoring and early warning module, a pest and disease identification module, a market prediction module, a logistics scheduling module, a product estimation module, and a growth state evaluation module. The online seed selection module is configured to allow consumers to select crops according to their preferences and needs, and to transmit the seed selection information to the corresponding farmers, thereby initiating exclusive planting services. The order tracking module is configured to allow consumers to track the status of the order in real time through the system after the farmers start planting the crops selected by the consumers. The environment monitoring and early warning module is configured to monitor and analyze environmental data of farmland, including soil humidity, temperature, pH, air temperature, humidity, and light intensity. Once the environmental data is abnormal, the module sends early warning information to the farmers and provides suggestions for countermeasures. The pest and disease identification module is configured to use image recognition and deep learning techniques to quickly and accurately identify the types of pests and diseases, and to provide control solutions. The market prediction module is configured to analyze market supply and demand conditions and price fluctuation trends of different fruit and vegetable varieties through big data analysis and artificial intelligence, and to predict the market trend of agricultural products, adjust the planting plan, and select varieties with high market demand and high prices for planting. The logistics scheduling module is configured to arrange logistics vehicles and delivery routes according to the distribution of orders and delivery time requirements, and to monitor the location and transportation status of logistics vehicles in real time. The module also adjusts the logistics plan in a timely manner according to external factors to ensure smooth delivery. The product estimation module is configured to use the YOLOv8 model to analyze images of crop growth, identify the number and growth state of crops, and combine historical data and environmental factors to predict the final yield. The growth state evaluation module is configured to quantitatively analyze the health status and growth rate of crops based on environmental data, pest and disease conditions, and growth cycles, and to continuously evaluate the growth state of crops and provide suggestions and solutions for farmers. The intelligent agent also determines whether the environment is suitable for crop growth through data analysis, which includes the following: Real-time environmental data collection: Distributed Internet of Things sensor networks are deployed in farmland, including soil sensors to monitor and collect soil temperature and humidity, pH, and EC value, air sensors to monitor and collect air temperature and CO2 concentration, light sensors to monitor and collect photosynthetically active radiation, and weather stations to monitor and collect wind speed and precipitation. At the same time, the spatial coordinates of the data are associated with specific planting areas through Beidou positioning, providing a spatial coordinate reference for subsequent regional judgment. Construction of crop growth suitability parameter library: Based on the agricultural knowledge base, crop variety characteristics database, and old farmer experience library, a three-dimensional threshold system of crop-growth stage-environment parameter is constructed. The basic threshold is dynamically corrected based on historical planting data and machine learning models, and the environmental trend is analyzed to predict the environmental risk in the future set time. Data analysis model judgment: preprocess the collected raw data and introduce a multi-level suitability judgment algorithm for judgment, automatically divide the warning level according to the data deviation degree, regularly compare the actual crop growth after warning with the judgment result, and update the parameter library threshold and multi-factor weight through reinforcement learning algorithm to improve the judgment accuracy; The multi-level suitability judgment algorithm comprises: The first layer, the rule engine rapid screening: directly compare the real-time data with the threshold range in the parameter library, if the single-dimensional data exceeds the threshold, it is immediately marked as abnormal; The second layer, the multi-factor collaborative analysis model: using gradient boosting tree algorithm to analyze the interaction of multiple parameters, if one of the parameters in the multi-parameter interaction exceeds the threshold, the model judges that it is a potential risk environment; The third layer, growth stage adaptation adjustment: through planting cycle record and image recognition to confirm the current growth stage of crops, adjust the parameter threshold of the current stage to prevent misjudgment caused by general parameter threshold. 2.The intelligent agricultural management system based on planting-harvesting-delivery full link according to claim 1, characterized in that: The system further comprises a knowledge popularization module, an intelligent planting guidance module and a plant adoption module; The knowledge popularization module is configured to provide agricultural knowledge for consumers, including the nutritional value, storage method and cooking suggestion of different fruits and vegetables, the planting process of crops and the prevention and control of pests and diseases; The intelligent planting guidance module is configured to make personalized planting schemes for farmers by combining the growth law of crops and historical planting data with the collected farmland level environment data; The plant adoption module is configured for consumers to select different appearances of land and plant varieties planted on the land according to personal preferences, and to purchase the corresponding adoption rights. 3.The intelligent agricultural management system based on planting-harvesting-delivery full link according to claim 1, characterized in that: The intelligent agent also customizes exclusive planting schemes for different crops by comprehensively considering crop variety characteristics, planting area climate and soil conditions and seasonal change factors; According to the differences between regions, combined with the state of water resources and soil fertility, precise irrigation and fertilization schemes are developed; damaged fruits are identified and treatment schemes are given. 4.The intelligent agricultural management system based on planting-harvesting-distributing whole link according to claim 3, characterized in that: The comprehensive consideration of crop variety characteristics, planting area climate and soil conditions and seasonal change factors to customize exclusive planting schemes for different crops specifically includes the following contents: Multi-dimensional basic data collection and integration: build a crop variety characteristics database, and collect regional climate and soil data in real time, divide seasons according to phenological period, correlate typical climate parameters in each season, and map to environmental requirements in key growth stages of crops; Intelligent decision-making model scheme generation: combine crop germination temperature threshold with regional spring temperature recovery curve to calculate the optimal sowing time window; calculate the planting density according to the soil fertility grade, crop plant type and row spacing formula; integrate the old farmer experience database and knowledge base to generate basic schemes of irrigation period and fertilizer type; according to the characteristics of the region, the generated basic scheme is modified, and combined with the seasonal climate risk database, preventive measures are embedded in the scheme; The whole-cycle dynamic adjustment mechanism of the scheme: based on the recorded crop growth stage, the agent updates the scheme every set time, and when the environmental monitoring data deviates from the set range, the emergency adjustment mechanism of the scheme is automatically triggered, and the agent optimizes the scheme according to the feedback data of the farmers, and the scheme is decomposed into executable task chains and automatically pushed to the farmer end app. 5.The intelligent agricultural management system based on planting-harvesting-delivery full link according to claim 3, characterized in that: The specific contents of the precise irrigation and fertilization scheme formulated according to regional differences, combined with water resource status and soil fertility include the following: Multi-dimensional basic data collection and regional characteristic modeling: collect soil fertility data and combine regional soil survey data to establish a regional baseline of soil fertility; collect water resource data, divide water resource abundance levels and associate irrigation costs; establish a crop demand database; Precise irrigation scheme generation: calculate the crop water requirement in real time according to the crop coefficient, evaporation and transpiration, water resource abundance level and soil water retention capacity; select the irrigation method according to the regional water source type, and optimize and adjust the irrigation time combined with weather prediction and crop growth stage; output specific irrigation parameters, and automatically execute through the Internet of Things control irrigation equipment; Precise fertilization scheme generation: determine the target nutrient uptake according to the crop variety and growth stage, and calculate the natural supply ratio according to the current soil fertility data; select the fertilizer type according to the soil type, and select the application method combined with the regional agricultural machinery conditions, and formulate environmental protection constraints; Regional adaptation and dynamic optimization mechanism: divide the country into 6 major agricultural ecological regions, dynamically correct each region according to regional differences, and the agent collects soil data every N days through the Internet of Things equipment, compares the actual value with the expected value of the scheme according to the crop growth state, and optimizes the fertilization scheme according to the comparison result. 6.The intelligent agricultural management system based on planting-harvesting-delivery full link according to claim 1, characterized in that: The specific contents of the pest and disease identification module include the following: Multi-dimensional construction of pest and disease feature database: construct the feature database of crop pests and diseases through agricultural knowledge base, field image sample collection and old farmer experience data; Pest and disease identification model construction and training: add CBAM attention module to the ResNet-50 model to strengthen the extraction of key features of disease spot edges and insect outline, and use depth separable convolution to compress 40% of pest and disease identification model parameters; perform data enhancement processing on the image, expand the sample, train according to the type + crop variety classification, increase the sample proportion of high-incidence pests and diseases, correct the misjudgment of similar pests and diseases through confusion matrix, and improve the recognition accuracy of the pest and disease identification model; deploy the trained pest and disease identification model on the farmer end edge device; Pest and disease identification and control of the agent: the farmer uploads the fruit diseased part image through the system, the agent performs multi-dimensional identification and cross-validation after preprocessing the image, generates a control scheme according to chemical control, physical / biological control and emergency treatment, and finally optimizes the control scheme according to the feedback data of the farmers and combined with the actual effect in the field. 7.The intelligent agricultural management system based on planting-harvesting-distributing whole link according to claim 3, characterized in that: The specific contents of identifying the damaged fruit and giving the treatment scheme include the following: Database construction of damaged fruit characteristics: Construct a multi-dimensional crop damaged fruit characteristics database, and clearly define the types of disease damage, pest damage and physical damage. Label the collected damaged fruit images; Training and deployment of damaged fruit recognition model: Add a coordinate attention module to the Neck layer of the MobileNetV3 deep learning model to enhance feature extraction of the damaged area edge. Use depth separable convolution to reduce the model parameters by 35% to adapt to farmer mobile phones and field intelligent terminal devices. Perform data augmentation on the images to expand the sample set. Train the model according to the crop variety + damage type classification. Increase the sample weight for the specified crops. Optimize the ability to distinguish similar damage through the confusion matrix. Deploy the trained damaged fruit recognition model on the farmer's edge device; Agent damaged fruit recognition: Farmers upload fruit damage images, and the agent performs preprocessing, multi-dimensional recognition, and cross-validation on the images; Treatment scheme generation: The agent automatically matches the treatment measurement based on the damage degree. Track the treatment progress through IoT devices and collect downstream feedback data to update the treatment rule weight to optimize the scheme. 8.The intelligent agricultural management system based on planting-harvesting-delivery full link according to claim 1, characterized in that: The product estimation module specifically includes the following: Crop growth image data collection and labeling: Collect images in multiple scenarios and perform manual labeling and automated verification on the collected images; YOLOv8 model training and optimization: Add SE attention to the Neck layer of the YOLOv8 model to enhance the extraction of key features such as fruit outline and color, reduce model parameters, and adapt to farmer mobile phones and field edge devices. Perform hyperparameter tuning. First, pre-train the YOLOv8 model on a general object detection dataset, then adjust the trained YOLOv8 model using a crop fruit dataset to optimize the recognition ability of overlapping and occluded fruits. Train sub-models for different crop characteristics. Yield estimation: The YOLOv8 model extracts features from the input image and outputs fruit quantity, fruit size, and fruit distribution density data. Combine crop characteristics and planting parameters to calculate the estimated total yield = fruit quantity per unit area × total planting area × average fruit weight × maturity rate correction factor. Introduce environmental factors to correct the results.

Images