Crop disease and pest monitoring method based on Internet of Things

Abstract

The invention relates to the field of crop disease and insect pest monitoring, in particular to a crop disease and insect pest monitoring method based on the Internet of Things, which comprises the following steps: inputting an obtained crop image into an edge node to obtain probabilities of different insect pests, and constructing the probabilities of different types of insect pests of the same crop image into a result vector; the distance between any two crop images is calculated, all the crop images are clustered according to the distance to obtain a plurality of clustering clusters, the transmission priority of each clustering cluster is calculated, and the crop images in each clustering cluster are transmitted according to the transmission priority. According to the technical scheme, the precision and efficiency of crop disease and pest monitoring results can be improved.

Description

Technical Field

[0001] The present invention relates to the field of crop disease and insect pest monitoring. More specifically, the present invention relates to a crop disease and insect pest monitoring method based on the Internet of Things. Background Art

[0002] Agricultural production faces numerous challenges, particularly in pest and disease control. Crop pests and diseases not only affect crop yield and quality, but can also lead to severe economic losses and negatively impact the ecological environment. According to the World Health Organization, crop pests and diseases cause hundreds of billions of dollars in economic losses annually. To address this challenge, precision agriculture has emerged as a key direction for agricultural development. Precision agriculture relies on advanced information technology, particularly the Internet of Things, to better monitor, analyze, and prevent crop pests and diseases.

[0003] The application of IoT technology in agriculture is gradually transforming traditional agricultural production methods. Using sensors, data acquisition devices, and communication networks, the IoT collects and transmits a wide range of information from agricultural production processes in real time, enabling precise monitoring and management of the crop growing environment. Farmland environmental data, such as temperature, humidity, soil nutrients, weather changes, and crop growth conditions, can be captured in real time through IoT technology. This data provides crucial insights into crop pest and disease infestation and the need for fertilization, irrigation, or other agricultural management measures.

[0004] Existing technologies mainly rely on traditional visual and manual inspection methods to monitor crop pests. This is not only time-consuming and labor-intensive, but also often affects the accuracy and timeliness of pest forecasts due to the lag in information acquisition. At the same time, existing monitoring technologies have inaccurate judgments on the types and degree of damage of pests, and are unable to effectively distinguish the specific impacts of different pests on crops. Summary of the Invention

[0005] To address the technical issues of low accuracy and timeliness in pest forecasting, the present invention provides an IoT-based crop pest monitoring method. The method includes: inputting captured crop images into edge nodes to obtain the probabilities of different pests, constructing the probabilities of different pest types for the same crop image into a result vector; calculating the distance between any two crop images, clustering all crop images based on the distance to obtain a number of clusters, calculating the transmission priority of each cluster, and transmitting the crop images within each cluster based on the transmission priority.

[0006] Using a clustering algorithm, the system groups similar crop images together, centrally processing images with similar pest types and transmission risks, improving transmission efficiency and accuracy. By assigning transmission priorities to each cluster, the system prioritizes images of high-risk pests based on their severity, transmission characteristics, and outbreak risk, ensuring timely response to the most pressing pest issues.

[0007] Preferably, the edge node includes a lightweight pest and disease identification model.

[0008] Preferably, the distance satisfies the relationship: , Represents a crop image and crop image The distance between Represents a crop image The dimension of the result vector is The probability value of Represents a crop image The dimension of the result vector is The probability value of .

[0009] It can accurately identify similar crop images and cluster them into the same category, ensuring that images within the same category have high similarity in terms of pest type and damage severity. This similarity-based clustering approach not only improves the efficiency of image transmission but also ensures that the system can more intelligently prioritize images of potentially high-risk pests.

[0010] Preferably, the distance satisfies the relationship: , Represents a crop image and crop image The distance between Represents a crop image The dimension of the result vector is The probability value of Represents a crop image The dimension of the result vector is The probability value of represents the vector function, Indicates the total number of dimensions in the result vector.

[0011] The introduction of the vector function enables the calculation process to effectively capture the significant differences between different dimensions, thereby clustering similar crop images into the same category and ignoring irrelevant or highly different information.

[0012] Preferably, the value of the vector function is: In response to the probability values of the result vectors of the two crop images being equal in the same dimension, the value of the vector function of the dimension is 0; in response to the probability values of the result vectors of the two crop images being unequal in the same dimension, the value of the vector function of the dimension is 1.

[0013] Preferably, the clustering method is K-means clustering.

[0014] Preferably, the transmission priority satisfies the relationship: , Represents clusters The transmission priority of Represents the cluster after normalization The number of inner crop images, Represents clusters The dimensions of the result vector in The probability value of Indicates the total number of dimensions in the result vector.

[0015] This system not only considers the number of crop images within a cluster but also integrates the probability of each crop image across various pest dimensions. This allows for intelligent sorting of clusters based on the potential severity and outbreak risk of pests. This prioritization mechanism allows the system to prioritize clusters likely to contain high-risk pest information, ensuring rapid response and treatment of critical areas and high-risk pests.

[0016] Preferably, the transmission priority satisfies the relationship: the hazard levels of different types of pests are preset; , Represents clusters The transmission priority of Represents the cluster after normalization The number of inner crop images, Represents clusters The dimensions of the result vector in The probability value of represents the total number of dimensions in the result vector, Represents clusters degree of harm.

[0017] By presetting the damage levels of different pest types, the system automatically identifies and prioritizes the transmission of images related to pests with the highest potential for damage. This not only improves the intelligence level of crop pest monitoring but also ensures that the most urgent pests with the greatest potential to affect crop growth are addressed first, given limited transmission resources.

[0018] Beneficial effects of the present invention: By processing captured crop images, calculating pest probabilities, and constructing result vectors, the present invention can rapidly identify different types of pests and cluster crop images by calculating inter-image distances. This clustering approach prioritizes the transmission of the most important image data based on pest spread, thereby reducing network transmission burdens and ensuring timely and accurate transmission. Furthermore, the calculation of transmission priority takes into account the number and severity of crop images within a cluster, facilitating dynamic adjustment of data transmission strategies based on actual crop pest and disease conditions, optimizing resource allocation, and improving the response speed and resource utilization efficiency of crop pest and disease monitoring, thereby enhancing the accuracy and efficiency of pest and disease monitoring results. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is a flow chart of a method for monitoring crop diseases and insect pests based on the Internet of Things according to an embodiment of the present invention. DETAILED DESCRIPTION

[0020] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, but not all of the embodiments.

[0021] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0022] Reference Figure 1 The method for monitoring crop diseases and insect pests based on the Internet of Things includes steps S1 and S2, which are specifically as follows: S1: Input the acquired crop image into the edge node to obtain the probabilities of different pests, and construct the probabilities of different types of pests in the same crop image into a result vector.

[0023] In one embodiment, when deploying the pest and disease identification model between edge nodes and servers, reasonable allocation can be made based on differences in computing resources and task requirements.

[0024] The lightweight pest and disease recognition model on edge nodes, with its low computational workload, is suitable for environments with limited computing resources. It can perform preliminary screening of collected image data in real time, quickly identifying potential pest and disease issues, and transmit the screening results to the central server. This approach not only reduces data transmission latency but also effectively reduces the burden on the server.

[0025] On the other hand, pest and disease identification models deployed on servers typically have deeper network structures and more model parameters, enabling higher-precision identification, analysis of more complex pest types and severity, and more accurate predictions. However, due to the high computational complexity and relatively slow processing speed of these models, which require powerful computing resources, they are deployed on central servers to centrally process data from multiple edge nodes.

[0026] Through this "edge computing + cloud computing" collaborative model, we can fully utilize the advantages of each to achieve efficient identification and precise control of pests and diseases, and improve the level of intelligence in agricultural production.

[0027] By inputting the captured crop images into the edge nodes, the probabilities of different pests can be obtained, and the probabilities of all pests can be constructed as a result vector.

[0028] S2: Calculate the distance between any two crop images, cluster all crop images according to the distance to obtain several clusters, calculate the transmission priority of each cluster, and transmit the crop images in each cluster according to the transmission priority.

[0029] It should be noted that because the lightweight models deployed on edge nodes have lower recognition accuracy than models on servers, edge node recognition results may include misjudgments or incorrect identifications. However, edge nodes don't simply recognize a single image; they process multiple crop images within the same area. If pests are present in a particular area, they typically appear in multiple images, not just one or a few. Based on this characteristic, edge nodes perform cluster analysis on the pest recognition results for all input images.

[0030] Clustering can identify pest species with higher frequency, and these high-frequency identifications are generally more likely to be correct. This process effectively reduces the probability of misjudgment. During transmission, edge nodes prioritize images that indicate severe pest conditions and are likely to be correctly identified, sending them to the server. Upon receiving this data, the server utilizes its higher-performance recognition models for more precise analysis, thereby providing more accurate pest identification and early warning. This approach, through rational resource allocation and data optimization, fully leverages the real-time performance of edge nodes and the computing power of servers to achieve efficient and accurate pest monitoring and early warning.

[0031] In one embodiment, the distance between any two crop images is calculated, and the distance satisfies the relationship: , Represents a crop image and crop image The distance between Represents a crop image The dimension of the result vector is The probability value of Represents a crop image The dimension of the result vector is The probability value of .

[0032] The distance between any two crop images is traversed and clustered according to the K-means clustering method to obtain several clusters, where one cluster corresponds to one type of pest and disease.

[0033] After clustering is completed, the number of crop images contained in each cluster is obtained.

[0034] The number of crop images within a cluster reflects the similarity of the crop pest identification results. A larger number indicates that similar pest characteristics appear in multiple images within the region, indicating a more consistent pest situation across crops. Therefore, these images should be prioritized for transmission to the server for further, accurate detection.

[0035] Cluster analysis can identify the distribution of pests in different regions. Clusters with a larger number of images are more likely to represent areas with concentrated pests and diseases, requiring priority treatment. Furthermore, the identification results of each cluster not only reflect the distribution of pests and diseases, but also reveal the severity of the pests. If the identification results of a cluster indicate that the area is severely affected by pests and diseases, or that the type of pest is particularly unique, these images should be uploaded to the server first for analysis and confirmation using a more accurate model. This mechanism of prioritized transmission and intelligent screening ensures the efficiency and accuracy of the pest monitoring process, reduces unnecessary data transmission, and improves the emergency response capabilities and management efficiency of agricultural production.

[0036] Specifically, the transmission priority satisfies the relationship: , Represents clusters The transmission priority of Represents the cluster after normalization The number of inner crop images, Represents clusters The dimensions of the result vector in The probability value of Indicates the total number of dimensions in the result vector.

[0037] Based on image transmission priority, edge nodes first transmit images that, after cluster analysis, have reliable identification results and indicate severe pest infestations to the central server. Upon receiving these images, the central server uses a more precise and sophisticated pest and disease identification model for further analysis and identification. When the probability of a pest or disease identified exceeds a preset threshold, the area is deemed to be infested with pests and diseases, and an alert is issued promptly, prompting personnel to take appropriate preventive measures. This prioritized image transmission and precise identification mechanism effectively reduces misjudgments and improves the accuracy of pest detection. It also ensures timely intervention at the earliest stages of pest and disease outbreaks, maximizing crop health and enhancing the efficiency and stability of agricultural production.

[0038] In one embodiment, the distance satisfies the relationship: , Represents a crop image and crop image The distance between Represents a crop image The dimension of the result vector is The probability value of Represents a crop image The dimension of the result vector is The probability value of represents the vector function, Indicates the total number of dimensions in the result vector.

[0039] In which, in response to the probability values of the result vectors of the two crop images being equal in the same dimension, the value of the vector function of the dimension is 0; in response to the probability values of the result vectors of the two crop images being unequal in the same dimension, the value of the vector function of the dimension is 1.

[0040] In one embodiment, it should be noted that the calculation of transmission priority not only considers the number and similarity of crop images within a cluster, but also weights them according to the degree of damage caused by the pest. Certain pests are highly harmful, characterized by strong reproductive capacity, rapid spread, and significant potential for economic loss. Outbreaks of these pests can have extremely severe impacts on crops. Therefore, when these highly harmful pests are identified, image upload and processing should be prioritized so that control measures can be implemented as soon as possible to minimize crop losses.

[0041] When calculating the transmission priority of clusters, factors such as the severity of the pest damage, its spread speed, and its potential impact range are taken into account. This automatically adjusts the transmission priority based on the pest's harmfulness, ensuring that high-damage pests are detected and addressed promptly. This priority adjustment mechanism, based on pest damage and spread characteristics, will significantly improve the response efficiency of the pest detection system, optimize resource allocation and decision support, and ensure the sustainability and stability of agricultural production.

[0042] Specifically, those skilled in the art preset the hazard levels of different types of pests. The transmission priority satisfies the relationship: , Represents clusters The transmission priority of Represents the cluster after normalization The number of inner crop images, Represents clusters The dimensions of the result vector in The probability value of represents the total number of dimensions in the result vector, Represents clusters degree of harm.

[0043] It should be noted that those skilled in the art may make various modifications and improvements without departing from the scope of the present invention, and these modifications and improvements fall within the scope of protection of the present invention. Therefore, the scope of protection of the patent for this invention shall be based on the appended claims.

Claims

1. A method for monitoring crop diseases and insect pests based on the Internet of Things, characterized in that: include: The acquired crop image is input into the edge node to obtain the probability of different pests, and the probabilities of different types of pests in the same crop image are constructed as a result vector; The distance between any two crop images is calculated, and all crop images are clustered according to the distance to obtain several clusters. The transmission priority of each cluster is calculated, and the crop images in each cluster are transmitted according to the transmission priority.

2. The method for monitoring crop diseases and insect pests based on the Internet of Things according to claim 1, characterized in that: The edge node includes a lightweight pest and disease identification model.

3. The method for monitoring crop diseases and insect pests based on the Internet of Things according to claim 1, characterized in that: The distance satisfies the relationship: , Represents a crop image and crop image The distance between Represents a crop image The dimension of the result vector is The probability value of Represents a crop image The dimension of the result vector is The probability value of .

4. The method for monitoring crop diseases and insect pests based on the Internet of Things according to claim 1, characterized in that: The distance satisfies the relationship: , Represents a crop image and crop image The distance between Represents a crop image The dimension of the result vector is The probability value of Represents a crop image The dimension of the result vector is The probability value of represents the vector function, Indicates the total number of dimensions in the result vector.

5. The method for monitoring crop diseases and insect pests based on the Internet of Things according to claim 4, characterized in that: The value of the vector function is: In response to the probability values of the result vectors of the two crop images being equal in the same dimension, the value of the vector function of the dimension is 0; in response to the probability values of the result vectors of the two crop images being unequal in the same dimension, the value of the vector function of the dimension is 1.

6. The method for monitoring crop diseases and insect pests based on the Internet of Things according to claim 1, characterized in that: The clustering method is K-means clustering.

7. The method for monitoring crop diseases and insect pests based on the Internet of Things according to claim 1, characterized in that: The transmission priority satisfies the relationship: , Represents clusters The transmission priority of Represents the cluster after normalization The number of inner crop images, Represents clusters The dimensions of the result vector in The probability value of Indicates the total number of dimensions in the result vector.

8. The method for monitoring crop diseases and insect pests based on the Internet of Things according to claim 1, characterized in that: The transmission priority satisfies the relationship: Preset the damage levels of different types of pests; , Represents clusters The transmission priority of Represents the cluster after normalization The number of inner crop images, Represents clusters The dimensions of the result vector in The probability value of represents the total number of dimensions in the result vector, Represents clusters degree of harm.

Images