An AI-driven method for precise control of the peach fruit moth based on its growth stage.

Abstract

This invention discloses an AI-driven method for precise control of the oriental fruit moth during its growth stages, belonging to the field of crop pest and disease control technology. Addressing the shortcomings of poor targeted control, extensive pesticide application, and lack of precise decision-making and closed-loop optimization in peach orchards in the Jiangsu and Zhejiang regions, this invention constructs a four-dimensional feature library containing crop, pest, environment, and control methods (each dimension labeled with plant protection-related weights). Through a multimodal AI model fusing visual, sensor, and phenological data, it achieves precise determination of the oriental fruit moth throughout its entire growth stage (accuracy ≥95%). Using weighted collaborative filtering and the NSGA-III multi-objective optimization algorithm, it dynamically matches the optimal combination of microemulsion pesticides, physical control, and auxins, meeting the targets of control efficacy ≥85%, pesticide reduction ≥30%, cost ≤15 yuan / mu, and phytotoxicity rate ≤0.5%. A closed-loop system is formed through a field efficacy feedback iterative model and the feature library. This method improves peach yield and quality, reduces pesticide residues and costs, is suitable for the climate of Jiangsu and Zhejiang, and has broad application prospects.

Description

Technical Field

[0001] This invention relates to the field of crop pest and disease control technology, specifically to an AI-driven precise control method for the peach fruit moth in peach orchards in the Jiangsu and Zhejiang regions, and particularly to a synergistic matching technology of AI with microemulsion pesticides, physical control, and auxins. Background Technology

[0002] The peach fruit borer (Carposina sasakii Matsumura) is the primary fruit-boring pest in core peach-producing areas of Jiangsu and Zhejiang provinces, including Wuxi, Suzhou, and Fenghua. Its larvae bore into the peach fruit, causing rot and fruit drop. In severe cases, the infestation rate can reach 15-20%, resulting in significant economic losses for growers. Current control methods for the peach fruit borer have the following technical shortcomings: 1) Poor targeting: Traditional control methods often use general-purpose pesticides, failing to target the specific larvae (1st, 2nd, 3rd instars) of the peach fruit borer. 1) The physiological characteristics and activity habits of different growth stages, such as larvae, pupae, adult mating period, and adult oviposition period, are not adapted to the control methods, resulting in the ineffectiveness of pesticides after larvae bore into the fruit and the over-reliance on high-concentration pesticides for adult control; 2) The extensive application of pesticides: the formulation ratio, application concentration, and application method of high-efficiency formulations such as microemulsion pesticides are not dynamically matched with the growth stage of pests and diseases and the field environment, resulting in problems such as excessive pesticide application leading to excessive residues or insufficient concentration leading to poor control efficacy, and the advantages of microemulsion pesticides in reducing dosage and increasing efficacy are not fully utilized; 3) Lack of precise decision support: relying on manual experience to judge the pest situation and the timing of control makes it difficult to accurately grasp the control window period of each growth stage, and there is no quantitative basis for the combination ratio of physical control, chemical control, and growth hormone-assisted control, resulting in poor synergistic effect; 4) Lack of closed-loop optimization mechanism: after control, no linkage between control efficacy feedback and program adjustment is established, resulting in subsequent control still relying on initial experience and being unable to adapt to the changes in pest situation under complex climatic conditions such as plum rain and summer drought in Jiangsu and Zhejiang. Existing AI-based pest and disease control technologies mostly focus on pest and disease identification, without deeply integrating the specific pest's growth period characteristics, regional crop planting characteristics, and pesticide formulation properties. This makes it difficult to achieve precise matching of "growth period - control methods - environmental conditions," and thus cannot meet the needs of refined control of the peach fruit moth in peach orchards in the Jiangsu and Zhejiang regions. Summary of the Invention

[0003] Purpose of the invention To address the shortcomings of existing technologies, this invention provides an AI-driven method for precise control of the peach fruit moth at different growth stages. By constructing a multi-dimensional feature library and AI decision-making model specific to peach orchards in Jiangsu and Zhejiang provinces, it achieves dynamic and precise matching of different growth stages of the peach fruit moth with microemulsion pesticides, physical control methods, and growth hormones, thereby maximizing control efficacy, maximizing pesticide reduction, and minimizing costs. Core technology solutions The core of this invention lies in the "AI closed-loop matching system driven by the quantification of plant protection knowledge," which specifically includes the following steps: Step 1: Construct a four-dimensional feature library of the growth stages of the peach fruit moth. Using core peach orchards in Wuxi, Suzhou, and Fenghua in the Jiangsu and Zhejiang regions as the research area, data on the entire growth period of the peach fruit moth, peach tree phenology, control method characteristics, and field environment were collected to construct a structured and labeled four-dimensional feature library. Each dimension of the feature library is labeled with plant protection-related weights. Crop dimensions: Peach varieties (Wuxi peach, Fenghua peach, etc.), phenological stages (young fruit stage, fruit expansion stage, etc.), growth vigor level (vigorous growth / moderate growth / weak growth), peach diameter (1-2cm / 2-4cm / 4cm and above), new shoot length, with a weight range of 10-30; Pest and disease dimensions: The growth stages of the peach fruit moth (1st instar larvae, 2nd instar larvae, 3rd instar larvae, pupa, adult mating period, adult oviposition period), population density (heads / branch, heads / acre), damage type (fruit borer, soil overwintering), and pesticide resistance levels in the Jiangsu and Zhejiang regions, with a weighting range of 40-60. For each growth stage, morphological characteristics are clearly marked (1st instar larvae: white body, 2-3 mm long; 2nd instar larvae: pale yellow body, 4-7 mm long; 3rd instar larvae: yellowish-white body, 10-15 mm long; pupa: yellowish-brown, 7-10 mm long; adult: grayish-brown, wingspan 13-18 mm), activity habits, and control window periods. Environmental dimensions: field temperature (°C), humidity (%), sunshine duration (h / d), soil moisture (%), and rainfall probability, with a weighting range of 15-25; Prevention and control measures: 1) Chemical control: Microemulsion pesticides (systemic, contact, fumigant), oil phase ratio (10%-20%), emulsifier type (polyoxyethylene ether, sodium dodecylbenzene sulfonate, etc.), active ingredient concentration, minimum effective dose, crop safety threshold and other parameters; 2) Physical control: sex pheromone traps (trapping efficiency), light traps (wavelength 365-400nm, operating time 19-23 o'clock), manual removal (applicable to fruit-boring larvae); 3) Auxins: Brassinolide (concentration gradient 0.01-0.05 mg / L), gibberellin (applicable phenological stage is young fruit stage), indicate the efficacy, cost and suitable growth period of each method, with a weight range of 30-50. The data sources for the feature database include: field trial data from the inventor's previous research and development of microemulsion pesticides, measured insect pest data from peach orchards in Jiangsu and Zhejiang provinces over three growing seasons, and special data on the control of peach fruit moth from the Zhejiang Academy of Agricultural Sciences. Step 2: Precise determination of the reproductive period of the peach fruit moth based on multimodal fusion A multimodal AI perception model integrating visual recognition, sensors, and phenological data was built to achieve high-precision determination of the growth period of the peach fruit moth. Data acquisition module: 1) Visual data acquisition: Drones equipped with high-definition cameras (1080P) capture images of peaches and tender shoots. 10-15 fixed visual monitoring stations are set up in the field to cover different areas of the peach orchard. 2) Insect count: The pheromone trap has a built-in infrared counting module to count the number of adult insects captured in real time; 3) Environmental data collection: Deploy sensors for temperature, humidity, light intensity, and soil moisture, with a data collection frequency of 1 hour per time; 4) Phenological data collection: Manually record the length of new peach shoots and the diameter of peach fruits twice a week; AI Judgment Model: 1) Construct a sub-model for identifying the reproductive stage based on CNN convolutional neural network. Input morphological images of the peach fruit moth at different reproductive stages. 80% of the training samples of the model are measured images from peach orchards in Jiangsu and Zhejiang. The model parameters are fine-tuned through transfer learning. 2) An auxiliary judgment sub-model was constructed by integrating environmental and phenological data, and threshold rules were set: when the peach fruit diameter is 1-2cm, the probability of adult mating is ≥85%; when the temperature is 22-28℃ and the humidity is 60-75%, the larval hatching rate is ≥90%. 3) Multimodal data fusion: The weighted voting method is used to fuse the output results of two sub-models. The accuracy rate of growth period determination is ≥95%, and the complete determination result of "crop - peach fruit moth - specific growth period - insect population density - environmental parameters" is output. Step 3: AI-based precise matching of prevention and control measures based on multi-objective optimization An AI matching model based on "weighted collaborative filtering + multi-objective optimization" is constructed to achieve dynamic adaptation between judgment results and prevention and control measures. Step 1: Weighted Collaborative Filtering Initial Screening Based on the plant protection association weights in the four-dimensional feature database, the matching score between the judgment result and each control method is calculated. The calculation formula is as follows: Matching score = Σ (Crop dimension weight × Crop fit coefficient + Pest and disease dimension weight × Growth period fit coefficient + Environment dimension weight × Environment fit coefficient) Control methods with a matching score ≥80 are selected to form a candidate set; for example, the candidate set for first instar larvae is "systemic microemulsion pesticide + low concentration of brassinolide"; Step 2: Multi-objective optimization calculation Four optimization goals were set: ① Control efficacy ≥ 85%; ② Reduction of microemulsion pesticide usage by ≥ 30% compared to traditional methods; ③ Control cost ≤ 15 yuan / mu; ④ Peach fruit pesticide damage rate ≤ 0.5%; The non-dominated sorting genetic algorithm (NSGA-III) is used to optimize the candidate set and output the optimal prevention and control scheme, specifically including: 1) Chemical control parameters: formulation of microemulsion pesticides (oil phase ratio, emulsifier type), application concentration, dosage per acre, and application method (low-volume spraying by drone or manual spot spraying). 2) Physical control parameters: density of pheromone traps, timing of light trap activation, and timing of manual removal; 3) Auxin parameters: type of auxin, concentration, and spraying time; 4) Synergistic ratio: The combination ratio and application interval of chemical control, physical control and auxin control. Step 4: Implementation of prevention and control measures and optimization of closed-loop feedback Implementation of the plan: Large-scale peach orchards: The prevention and control parameters output by AI are directly connected to intelligent spraying equipment (drones, self-propelled sprayers). The equipment automatically adjusts the spray pressure, droplet diameter, and travel speed to achieve unmanned and precise spraying. For small and medium-sized farmers: Professional parameters are converted into easy-to-understand operation guides through mobile apps (such as "XX microemulsion 1:2000 diluted with water, spray on peaches between 9-11 am, spray 50 catties of water per mu"), and video tutorials on application are also provided. Feedback optimization: Control efficacy monitoring: 7-15 days after application, AI visual monitoring and manual sampling surveys are used to count the insect population reduction rate and peach fruit damage rate to obtain actual control efficacy data; Data feedback: Input data such as actual control efficacy, pesticide residue, and crop growth status into the four-dimensional feature library and update the plant protection correlation weights of each dimension; Model iteration: Fine-tune the parameters of the AI ​​judgment model and matching model based on newly entered data to improve the accuracy of subsequent prevention and control plans, forming a complete closed loop of "monitoring-judgment-matching-execution-feedback-optimization". Key innovations: A four-dimensional feature library of the peach fruit moth, exclusive to peach orchards in Jiangsu and Zhejiang, was constructed. For the first time, plant protection experience was quantified into feature weights, solving the problem of poor targeting of general feature libraries and realizing targeted association between "growth period and control methods". A multimodal fusion AI judgment model is adopted, which integrates visual, environmental and phenological data to overcome the limitations of single image recognition and achieve high-precision and robust judgment of the growth period of the peach fruit moth, accurately locking in the prevention and control window. An innovative dynamic matching algorithm for multi-objective optimization is proposed, which dynamically binds the formulation parameters and application methods of microemulsion pesticides with growth period and environmental conditions, while optimizing four major objectives: efficacy, pesticide reduction, cost, and safety, to achieve quantitative synergy of chemical, physical, and auxin-based control. Establish a closed-loop feedback mechanism to continuously iterate the model and feature library through field measurement data, adapt to the changes in insect pests under the complex climate of Jiangsu and Zhejiang regions, and ensure the stability of long-term control effects. Beneficial effects Significantly improved control efficacy: The control efficacy against the peach fruit moth was ≥90% at all growth stages, and the damage rate of peaches decreased from 15-20% with traditional control methods to below 3%, greatly improving peach yield and quality; Significant reduction in pesticide use and increased efficiency: Microemulsion pesticides reduce the amount used by 30-55% compared to traditional methods, reducing pesticide residues and environmental pollution, while fully leveraging the high efficiency of microemulsion pesticides to reduce planting costs; Easy to operate and highly adaptable: It is compatible with intelligent spraying equipment in large-scale peach orchards, and meets the needs of small and medium-sized farmers through popular guides. No professional AI knowledge is required to operate it. Adapting to regional climate characteristics: Through a closed-loop optimization mechanism, it dynamically adapts to the changes in insect infestation under complex climates such as plum rain and summer drought in the Jiangsu and Zhejiang regions, solving the problem of the traditional "one-size-fits-all" approach to prevention and control. The technology is replicable and scalable: Based on the framework of this solution, it can be extended to other common crop pests and diseases in the Jiangsu and Zhejiang regions, such as citrus leafminer and peach borer, and has broad application prospects. Attached Figure Description Figure 1 A technical flowchart for AI-driven precision control of the peach fruit moth based on its growth stage. Specific Implementation The implementation process of the present invention will be further described in detail with reference to the accompanying drawings and embodiments. For example... Figure 1 As shown: (1) Construct a four-dimensional feature library of the fruit moth of the peach, which includes crop dimension, pest and disease dimension, environmental dimension, and control method dimension, and label the plant protection association weight of each dimension; (2) Through multimodal data collection and AI judgment model, accurately identify peach tree varieties, peach fruit moth growth period and population density, and field environmental parameters; (3) Using a weighted collaborative filtering + multi-objective optimization algorithm, the judgment result is accurately matched with the control measures in the feature library, and a collaborative control scheme including microemulsion pesticide formulation, physical control parameters and auxin parameters is output. (4) After implementing the prevention and control plan, obtain the prevention and control effect data through field monitoring, feed it back to the feature library and iterate the AI ​​model to form a closed loop. Example 1: Taking the control of peach fruit moth in a Wuxi peach orchard (10 mu in area, 3-year-old peach trees, Wuxi Yangshan peach variety) as an example, the peach fruit moth is controlled at the first instar larvae + young fruit stage (peach diameter 1-2cm). 1) Data collection: Images of peaches were taken by drones. Adult insects were not captured by pheromone traps. Sensor data showed a temperature of 25℃ and a humidity of 65%. Peach diameter was recorded as 1.5cm and new shoot length as 8cm. 2) Determination of reproductive stage: After integrating the above data, the AI ​​multimodal model determined it to be "Wuxi peach - first instar larva of the peach fruit moth", with an accuracy rate of 97% and a population density of 8 larvae / branch; 3) Feature database retrieval: The plant protection label "1st instar larvae have thin body walls and do not penetrate deep into the fruit, suitable for systemic microemulsion pesticides + low concentration of brassinolide, with a control window of 3-5 days" was retrieved from the four-dimensional feature database; 4) Matching calculation: The candidate set "systemic microemulsion + brassinolide 0.01 mg / L" was obtained through weighted collaborative filtering, and the final solution was determined after NSGA-III multi-objective optimization; 5) Solution output: 2000 times dilution of systemic microemulsion pesticide (15% oil phase, emulsifier type is polyoxyethylene ether), 50L per mu, using drone low-volume spraying (droplet diameter 40μm), combined with 0.01mg / L brassinolide for foliar spraying, the timing of application is 9-10 am on sunny days; 6) Feedback optimization: Monitoring 10 days after application showed that the insect population reduction rate was 94%, the peach fruit damage rate was 2.1%, the pesticide dosage was reduced by 42%, and no phytotoxicity occurred. This set of data was entered into the four-dimensional feature library to increase the association weight between the first instar larvae and the microemulsion formulation to 92. Example 2: Mating period of adult peach fruit moth + young peach fruit stage (peach diameter 2-3cm) 1) Data collection: Sex pheromone traps captured 62 adult insects per mu per day. Drones captured images of adult insects mating. Sensor data showed a temperature of 26℃ and humidity of 60%. Manual recording showed that the peach fruit diameter was 2.5cm and the new shoot length was 12cm. 2) Determination of reproductive period: After integrating the above data, the AI ​​multimodal model determined it to be "the mating period of adult peach fruit moths in Wuxi peaches", with an accuracy rate of 96%; 3) Feature database retrieval: The plant protection label "Adult mating period is the pre-oviposition window period, prioritize physical control + low concentration contact microemulsion to reduce phytotoxicity to young fruit" was retrieved from the four-dimensional feature database; 4) Matching calculation: The candidate set of "sexual attractant + frequency-vibration insecticidal lamp + contact microemulsion" was obtained through weighted collaborative filtering, and the matching ratio was determined after NSGA-Ⅲ multi-objective optimization; 5) Solution output: Physical control (35 pheromone traps per acre, frequency-vibration insecticidal lamps turned on from 7-11 pm every night) + 2500 times diluted contact-type microemulsion pesticide, 25L per acre, using ultra-low volume spraying, reducing pesticide usage by 55%; 6) Feedback optimization: Monitoring 7 days after application showed that the number of adult insects trapped decreased by 92% and the subsequent larval hatching rate dropped to 8%. The weight of the synergistic scheme was updated to 95 and entered into the four-dimensional feature library for model iteration. This invention has been validated in 12 demonstration sites in the Jiangsu and Zhejiang regions, with a cumulative application area of ​​3,260 mu (approximately 217 hectares). The average yield increase was 860 kg per mu (approximately 57.5 kg per hectare), the rate of high-quality fruit improved to 92.3%, and farmer training satisfaction reached 98.7%. It provides implementable, quantifiable, and sustainable technical support for the green transformation of the peach industry in the middle and lower reaches of the Yangtze River. Data is shown in the table below: test point Area (mu) reproductive period Prevention efficacy (%) Reduction (%) Cost (yuan / mu) Victim rate (%) Pesticide damage rate (%). Wuxi Yangshan 50 1st instar larvae 94.7 42 13.2 2.2 0 Shaoxing 80 Adults mating 92.1 55 11.5 2.8 0 Suzhou Xishan 120 Mixing occurs 90.3 38 14.8 3.5 0.2 Fenghua Xikou 200 1st instar larvae 93.5 48 12.9 2.1 0 average - - 90.3 43 13.2 2.8 0.3 Conclusion: This invention demonstrates excellent stability and adaptability under complex climatic conditions, and all core indicators are significantly superior to existing technologies. The above is the complete technical solution of the present invention. Those skilled in the art can implement all the technical features of the present invention based on the above description. The scope of protection of the present invention is defined by the appended claims. Any equivalent substitutions or improvements that do not depart from the essence of the present invention fall within the scope of protection of the present invention.

Claims

1. An AI-driven method for precise control of the peach fruit moth according to its growth stage, characterized in that, Includes the following steps: (1) Construct a four-dimensional feature library of the fruit moth of the peach, which includes crop dimension, pest and disease dimension, environmental dimension, and control method dimension, and label the plant protection association weight of each dimension; (2) Through multimodal data collection and AI judgment model, accurately identify peach tree varieties, peach fruit moth growth period and population density, and field environmental parameters; (3) Using a weighted collaborative filtering + multi-objective optimization algorithm, the judgment result is accurately matched with the control measures in the feature library, and a collaborative control scheme including microemulsion pesticide formulation, physical control parameters and auxin parameters is output. (4) After implementing the prevention and control plan, obtain the prevention and control effect data through field monitoring, feed it back to the feature library and iterate the AI ​​model to form a closed loop.

2. The method according to claim 1, characterized in that, The four-dimensional feature library includes the morphological characteristics, activity habits, and control window markings of the first, second, and third instar larvae, pupae, and adult mating and oviposition periods of the peach fruit moth.

3. The method according to claim 1, characterized in that, The AI ​​judgment model adopts a multimodal fusion architecture of CNN convolutional neural network and auxiliary judgment sub-model, with an accuracy rate of ≥95% in determining the reproductive period.

4. The method according to claim 1, characterized in that, The multi-objective optimization algorithm has the following optimization objectives: control effect ≥85%, reduction of microemulsion pesticide dosage ≥30%, control cost ≤15 yuan / mu, and pesticide damage rate of peach fruit ≤0.5%.

5. The method according to claim 1, characterized in that, The control methods include systemic / contact / fumigation microemulsion pesticides, pheromone attractants, light traps, manual removal, brassinolide, and gibberellin, and the combination ratio of each method is dynamically calculated and determined by AI.

6. The method according to claim 1, characterized in that, The multimodal data acquisition includes visual acquisition, insect infestation counting, environmental acquisition, and phenological acquisition. Visual acquisition is achieved through a high-definition camera on a drone and a fixed visual monitoring station in the field, while insect infestation counting is achieved through an infrared counting module built into a pheromone trap.

7. The method according to claim 1, characterized in that, The matching score calculation formula for the weighted collaborative filtering is: matching score = Σ (crop dimension weight × crop fit coefficient + pest and disease dimension weight × growth period fit coefficient + environment dimension weight × environment fit coefficient), and prevention and control measures with a matching score ≥ 80 are selected to form a candidate set.

8. The method according to claim 1, characterized in that, The multi-objective optimization algorithm is the non-dominated sorting genetic algorithm (NSGA-Ⅲ).

Images