Intelligent detection and classification method and system for diseases and pests of dendrobium officinale

Abstract

This invention provides an intelligent detection and classification method and system for diseases and pests of Dendrobium officinale, relating to the field of data processing. The method includes: acquiring multiple sample data; establishing an optical flow calculation model, a feature extraction model, and a disease and pest classification model based on the multiple sample data; acquiring 3D point cloud data of Dendrobium officinale based on its 3D data using 3D reconstruction technology; determining the optical flow information of Dendrobium officinale based on its 2D image sequence using the optical flow calculation model; determining whether there is an optical flow generation area in the leaf region of Dendrobium officinale based on the 3D point cloud data and optical flow information; if so, extracting image features of Dendrobium officinale from its 2D image sequence using the feature extraction model; and determining the type of disease and pest of Dendrobium officinale based on its image features and optical flow information using the disease and pest classification model. This method has the advantage of improving the efficiency and accuracy of detecting and classifying diseases and pests of Dendrobium officinale.

Description

Technical Field

[0001] This invention relates to the field of data processing, and in particular to a method and system for intelligent detection and classification of diseases and pests of Dendrobium officinale. Background Technology

[0002] Dendrobium is a general term for perennial epiphytic herbaceous plants belonging to the genus Dendrobium in the family Orchidaceae. Classified by use, it can be divided into medicinal Dendrobium and ornamental Dendrobium. Generally, species with thin stems and small flowers, such as Dendrobium officinale, Dendrobium chrysanthum, and Dendrobium huoshanense, are used for medicinal purposes; while species with thick stems and large flowers are used for ornamental purposes. Both medicinal and ornamental Dendrobium have achieved a certain scale of industrialized production in the market and have yielded good economic benefits.

[0003] Existing intelligent detection and classification technologies for Dendrobium officinale diseases and pests cannot fully adapt to the diversity and complexity of these pests and diseases. The morphology and characteristics of diseases and pests may differ at different growth stages and under different environmental conditions, and existing intelligent detection and classification technologies for Dendrobium officinale diseases and pests may not be able to accurately capture these changes, resulting in limited identification accuracy.

[0004] Therefore, there is a need to provide intelligent detection and classification methods and systems for Dendrobium officinale diseases and pests to improve the efficiency and accuracy of detection and classification of Dendrobium officinale diseases and pests. Summary of the Invention

[0005] This invention provides an intelligent detection and classification method for diseases and pests of Dendrobium officinale, comprising: acquiring multiple sample data, wherein the sample data includes a two-dimensional image sequence of Dendrobium officinale samples, optical flow information, and disease and pest types; establishing an optical flow calculation model, a feature extraction model, and a disease and pest classification model based on the multiple sample data; acquiring three-dimensional data of Dendrobium officinale using a 3D camera; acquiring 3D point cloud data of Dendrobium officinale based on the three-dimensional data of Dendrobium officinale using 3D reconstruction technology; and acquiring a two-dimensional image sequence of Dendrobium officinale using a 2D camera, wherein the two-dimensional image sequence includes multiple time-series data. Two-dimensional images of the interstices; using the optical flow calculation model, determining the optical flow information of *Dendrobium officinale* based on the two-dimensional image sequence of the *Dendrobium officinale*; determining whether there is an optical flow generation region in the leaf area of ​​the *Dendrobium officinale* based on the 3D point cloud data and optical flow information of the *Dendrobium officinale*; if it is determined that there is an optical flow generation region in the leaf area of ​​the *Dendrobium officinale*, extracting the image features of *Dendrobium officinale* from the two-dimensional image sequence of the *Dendrobium officinale* using a feature extraction model; determining the type of pest or disease of *Dendrobium officinale* based on the image features and optical flow information of the *Dendrobium officinale* using the pest and disease classification model.

[0006] Further, based on the multiple sample data, a feature extraction model is established, including: determining multiple image feature factors; for each sample of Dendrobium officinale, extracting a first image feature of the sample of Dendrobium officinale according to the multiple image feature factors and the two-dimensional image sequence of the sample of Dendrobium officinale; for each image feature factor, calculating a difference parameter corresponding to the image feature factor according to the initial image features of each sample data; determining multiple target image feature factors from the multiple image feature factors according to the difference parameter corresponding to each image feature factor; for each sample of Dendrobium officinale, extracting a second image feature of the sample of Dendrobium officinale according to the multiple target image feature factors; and establishing the feature extraction model based on the two-dimensional image sequence and the second image feature of the multiple samples of Dendrobium officinale.

[0007] Further, using the optical flow calculation model, the optical flow information of *Dendrobium officinale* is determined based on the two-dimensional image sequence of the *Dendrobium officinale*, including: for each pair of adjacent two-dimensional images in the two-dimensional image sequence of the *Dendrobium officinale*, the optical flow calculation model determines the motion velocity and direction of each pixel corresponding to the two adjacent two-dimensional images; based on the motion velocity and direction of each pixel corresponding to each pair of adjacent two-dimensional images, the optical flow correlation of any two pixels, the mean motion velocity, the variance of motion velocity, the mean motion direction, and the variance of motion direction are determined, wherein the optical flow information of *Dendrobium officinale* includes the optical flow correlation of any two pixels, the mean motion velocity, the variance of motion velocity, the mean motion direction, and the variance of motion direction of each pixel.

[0008] Further, based on the 3D point cloud data and optical flow information of the Dendrobium officinale, it is determined whether there is an optical flow generation region in the leaf area of ​​the Dendrobium officinale, including: extracting leaf point cloud data from the 3D point cloud data of the Dendrobium officinale through a leaf recognition model; and determining whether there is an optical flow generation region in the leaf area of ​​the Dendrobium officinale based on the coordinate mapping relationship between the 3D point cloud and the two-dimensional image, the leaf point cloud data, and the optical flow information of the Dendrobium officinale.

[0009] Furthermore, based on the coordinate mapping relationship between the 3D point cloud and the 2D image, the leaf point cloud data, and the optical flow information of the Dendrobium officinale, it is determined whether there is an optical flow generation region in the leaf area of ​​the Dendrobium officinale. This includes: determining multiple target pixels based on the mean motion velocity, variance of motion velocity, mean motion direction, and variance of motion direction of each pixel; clustering the multiple target pixels according to the optical flow correlation and pixel distance between any two target pixels using a clustering algorithm to determine multiple optical flow generation regions; and determining whether there is an optical flow generation region in the leaf area of ​​the Dendrobium officinale based on the coordinate mapping relationship between the 3D point cloud and the 2D image, the leaf point cloud data, and the multiple optical flow generation regions.

[0010] Furthermore, the sample data also includes a sequence of growth environment information for the sample Dendrobium officinale, wherein the growth environment information sequence includes growth environment information at multiple time points: The pest and disease classification model determines the pest and disease types of Dendrobium officinale based on image features and optical flow information, including: establishing a pest and disease-environment association map based on the multiple sample data, wherein the pest and disease-environment association map is used to record the target environmental factors and environmental characteristics corresponding to various Dendrobium officinale pests and diseases; obtaining the growth environment information sequence of Dendrobium officinale; calculating the risk coefficient corresponding to each Dendrobium officinale pest and disease based on the pest and disease-environment association map and the growth environment information sequence of Dendrobium officinale; and determining the pest and disease types of Dendrobium officinale based on image features, optical flow information, and the risk coefficient corresponding to each Dendrobium officinale pest and disease using the pest and disease classification model.

[0011] Furthermore, based on the multiple sample data, a disease and pest-environment association map is established, including: grouping multiple Dendrobium officinale samples according to the disease and pest type of each sample to determine multiple Dendrobium officinale sample groups; for each Dendrobium officinale sample group, calculating the difference parameter corresponding to each growth environment factor based on the growth environment information sequence of the Dendrobium officinale samples included in the sample Dendrobium officinale sample group; screening the target environment factor corresponding to the sample Dendrobium officinale sample group according to the difference parameter corresponding to each growth environment factor; extracting the environmental characteristics corresponding to the sample Dendrobium officinale sample group according to the target environment factor corresponding to the sample Dendrobium officinale sample group; and establishing the disease and pest-environment association map based on the target environment factor and environmental characteristics corresponding to each sample Dendrobium officinale sample group.

[0012] Furthermore, based on the pest-disease-environment correlation map and the growth environment information sequence of Dendrobium officinale, the risk coefficient corresponding to each Dendrobium officinale pest is calculated, including: extracting the environmental characteristics of Dendrobium officinale from the growth environment information sequence of Dendrobium officinale according to the target environmental factors corresponding to each Dendrobium officinale pest; and calculating the risk coefficient corresponding to each Dendrobium officinale pest based on the environmental characteristics of Dendrobium officinale and the pest-disease-environment correlation map using a risk prediction model.

[0013] Furthermore, the optical flow calculation model is the SEA-RAFT optical flow calculation model; the pest and disease classification model includes the ResNet50 network.

[0014] This invention provides an intelligent detection and classification system for diseases and pests of Dendrobium officinale, applying the aforementioned intelligent detection and classification method for diseases and pests of Dendrobium officinale, comprising: a sample acquisition module for acquiring multiple sample data, wherein the sample data includes a two-dimensional image sequence of Dendrobium officinale samples, optical flow information, and disease / pest type; a model building module for establishing an optical flow calculation model, a feature extraction model, and a disease / pest classification model based on the multiple sample data; a data acquisition module for acquiring three-dimensional data of Dendrobium officinale using a 3D camera; a three-dimensional reconstruction module for acquiring the 3D outline of Dendrobium officinale based on the three-dimensional data of Dendrobium officinale using three-dimensional reconstruction technology; and an image acquisition module for acquiring a two-dimensional image sequence of Dendrobium officinale using a 2D camera, wherein the... The two-dimensional image sequence includes two-dimensional images at multiple time points; the optical flow calculation module is used to determine the optical flow information of *Dendrobium officinale* based on the two-dimensional image sequence using the optical flow calculation model; the region determination module is used to determine whether there is an optical flow generation region in the leaf area of ​​*Dendrobium officinale* based on the 3D contour and optical flow information of *Dendrobium officinale*; the feature extraction module is used to extract image features of *Dendrobium officinale* from the two-dimensional image sequence using the feature extraction model if the region determination module determines that there is an optical flow generation region in the leaf area of ​​*Dendrobium officinale*; and the pest and disease classification module is used to determine the pest and disease type of *Dendrobium officinale* based on the image features and optical flow information of *Dendrobium officinale* using the pest and disease classification model.

[0015] Compared with existing technologies, the intelligent detection and classification method and system for diseases and pests of Dendrobium officinale provided by this invention has at least the following beneficial effects:

[0016] 1. By combining 3D point cloud data, 2D image sequences, and optical flow information of *Dendrobium officinale*, a comprehensive analysis of its growth status and pest and disease conditions can be conducted from multiple dimensions, thereby improving the accuracy of identification. Feature extraction models are used to extract image features of *Dendrobium officinale* from 2D image sequences. These features can include color, texture, shape, and other aspects, helping to describe the growth status and pest and disease characteristics of *Dendrobium officinale* in greater detail. Optical flow calculation models can monitor optical flow information in the leaf areas of *Dendrobium officinale* in real time. Abnormal changes in optical flow in the leaf areas may indicate the occurrence or development of pests and diseases. Once early signs of pests and diseases are detected, immediate control measures can be taken to prevent further spread and damage. Using 3D and 2D cameras for data acquisition, combined with optical flow calculation models and feature extraction models, automated monitoring and identification of pests and diseases in *Dendrobium officinale* can be achieved, reducing manual intervention and costs.

[0017] 2. By identifying multiple image feature factors, various possible features in the two-dimensional images of Dendrobium officinale can be comprehensively considered, such as color, texture, shape, and edges, thus more accurately reflecting the growth status and pest and disease characteristics of Dendrobium officinale. For each image feature factor, its corresponding difference parameter can be calculated to quantify the ability of different feature factors to distinguish between pest and disease types or growth states. This helps to screen out the feature factors most influential for pest and disease detection, improving the targeting and effectiveness of feature extraction. Determining multiple target image feature factors from multiple image feature factors based on the difference parameter can simplify the structure of the feature extraction model, reducing its complexity and computational cost. Establishing a feature extraction model based on the two-dimensional image sequence and second image features of multiple sample Dendrobium officinale can ensure that the model can accurately capture the key features of Dendrobium officinale, providing strong support for subsequent pest and disease classification.

[0018] 3. 3D point cloud data, composed of three-dimensional coordinate points on the surface of actual objects, accurately reflects the shape, size, and position of Dendrobium officinale leaves. Extracting leaf point cloud data from 3D point cloud data using a leaf recognition model allows for precise identification of leaf regions, providing an accurate foundation for subsequent optical flow analysis. Optical flow information reflects the movement of pixels in an image and is highly sensitive to changes in leaf movement caused by pests and diseases. By determining whether optical flow regions exist within leaf areas, signs of pests and diseases can be detected promptly, improving detection accuracy. Optical flow analysis can process image data in real time and quickly extract leaf movement information. Combining 3D point cloud data and optical flow information enables real-time monitoring and identification of Dendrobium officinale pests and diseases, providing strong support for timely control measures. Clustering target pixels using a clustering algorithm identifies multiple optical flow regions. This automated process reduces the time and cost of manual intervention and improves the real-time performance of pest and disease identification. Different types of pests and diseases may cause different leaf movement patterns, which are represented by different optical flow regions in the optical flow information. By analyzing the characteristics of the light flow generation area, the classification of pests and diseases can be further refined, improving the accuracy of classification. 3D point cloud data provides three-dimensional information about the leaves, which, combined with light flow information, can more comprehensively describe the movement state of the leaves, providing richer feature information for pest and disease classification.

[0019] 4. By introducing the growth environment information sequence, the influence of multiple environmental factors such as temperature, humidity, and light on the occurrence of diseases and pests in *Dendrobium officinale* can be comprehensively considered, thereby improving the accuracy of disease and pest prediction. A disease-pest-environment correlation map is established, which records the target environmental factors and characteristics corresponding to various *Dendrobium officinale* diseases and pests, providing a scientific basis for disease and pest prediction and helping to accurately determine the types of diseases and pests and their occurrence trends. Based on the disease-pest-environment correlation map and the growth environment information sequence of *Dendrobium officinale*, the risk coefficient corresponding to each *Dendrobium officinale* disease and pest is calculated, providing more dimensional information for the disease and pest classification model to determine the types of *Dendrobium officinale* diseases and pests, thus improving the accuracy of disease and pest type determination. Attached Figure Description

[0020] This specification will be further described by way of exemplary embodiments, which will be described in detail with reference to the accompanying drawings. These embodiments are not limiting; in these embodiments, the same reference numerals denote the same structures, wherein:

[0021] Figure 1 This is a flowchart illustrating the intelligent detection and classification method for diseases and pests of Dendrobium officinale according to some embodiments of this specification;

[0022] Figure 2This is a schematic diagram of the modules of the intelligent detection and classification system for diseases and pests of Dendrobium officinale, as shown in some embodiments of this specification. Detailed Implementation

[0023] To more clearly illustrate the technical solutions of the embodiments in this specification, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are merely some examples or embodiments of this specification. For those skilled in the art, these drawings can be applied to other similar scenarios without creative effort. Unless obvious from the context or otherwise specified, the same reference numerals in the drawings represent the same structures or operations.

[0024] Figure 1 This is a flowchart illustrating the intelligent detection and classification method for diseases and pests of Dendrobium officinale according to some embodiments of this specification, such as... Figure 1 As shown, the intelligent detection and classification method for diseases and pests of Dendrobium officinale may include the following steps.

[0025] Step 110: Obtain multiple sample data.

[0026] The sample data includes two-dimensional image sequences, optical flow information, and pest and disease types of Dendrobium officinale samples.

[0027] The sample data also includes the growth environment information sequence of the sample Dendrobium officinale, which includes growth environment information at multiple time points (e.g., environmental humidity, environmental temperature, environmental light, etc.).

[0028] Step 120: Based on multiple sample data, establish an optical flow calculation model, a feature extraction model, and a pest and disease classification model.

[0029] As a preferred approach, a feature extraction model is established based on multiple sample data, including:

[0030] Determine multiple image feature factors;

[0031] For each sample of Dendrobium officinale, the first image features of the sample Dendrobium officinale are extracted based on multiple image feature factors and the two-dimensional image sequence of the sample Dendrobium officinale, such as color features, texture features, shape features, etc.

[0032] For each image feature factor, the difference parameter corresponding to the image feature factor is calculated based on the initial image features of each sample data.

[0033] Based on the difference parameters corresponding to each image feature factor, multiple target image feature factors are determined from multiple image feature factors;

[0034] For each sample of Dendrobium officinale, the second image features of the sample Dendrobium officinale are extracted based on multiple target image feature factors;

[0035] A feature extraction model was established based on the two-dimensional image sequence and second image features of multiple samples of Dendrobium officinale.

[0036] For example, the difference parameters corresponding to image feature factors can be calculated using the following formula:

[0037]

[0038] in, Let be the difference parameter corresponding to the i-th image feature factor. The feature corresponding to the i-th image feature factor in the first image feature of the i-th sample Dendrobium officinale. The feature corresponding to the i-th image feature factor in the first image feature of the j-th sample Dendrobium officinale is... Let be the cosine similarity between the feature corresponding to the i-th image feature factor in the first image features of the i-th sample Dendrobium officinale and the feature corresponding to the i-th image feature factor in the first image features of the j-th sample Dendrobium officinale. This represents the total number of Dendrobium officinale samples.

[0039] Image feature factors whose difference parameter is greater than the first difference parameter threshold can be used as target image feature factors.

[0040] For each Dendrobium officinale sample, a first training sample and a first test sample can be established based on the two-dimensional image sequence of the sample, and the second image features of the sample can be used as the label of the first training sample or the first test sample. A feature extraction model is established, trained using multiple first training samples, and evaluated using multiple first test samples. Evaluation metrics may include feature extraction precision, recall, F1 score, etc.

[0041] The methods for establishing optical flow calculation models and pest classification models are similar to those for establishing feature extraction models, and will not be elaborated here.

[0042] Step 130: Use a 3D camera to collect three-dimensional data of Dendrobium officinale.

[0043] For example, a 3D linear light camera can be used to acquire three-dimensional data of Dendrobium officinale.

[0044] Step 140: Based on the three-dimensional reconstruction technology, obtain the 3D point cloud data of Dendrobium officinale according to the three-dimensional data of Dendrobium officinale.

[0045] For example, using PointNet 3D reconstruction technology, 3D point cloud data of Dendrobium officinale can be reconstructed based on the 3D data of Dendrobium officinale.

[0046] Step 150: Use a 2D camera to acquire a two-dimensional image sequence of Dendrobium officinale.

[0047] The two-dimensional image sequence includes two-dimensional images at multiple time points.

[0048] Step 160: Using the optical flow calculation model, determine the optical flow information of Dendrobium officinale based on the two-dimensional image sequence of Dendrobium officinale.

[0049] Specifically, it includes:

[0050] For each pair of two adjacent frames of a two-dimensional image sequence of Dendrobium officinale, the motion speed and direction of each pixel in the two adjacent frames are determined by an optical flow calculation model based on the two adjacent frames. The optical flow calculation model can be the SEA-RAFT optical flow calculation model.

[0051] Based on the motion velocity and motion direction of each pixel corresponding to two adjacent frames of two-dimensional images, the optical flow correlation of any two pixels, the mean motion velocity, the variance of motion velocity, the mean motion direction, and the variance of motion direction of each pixel are determined. Among them, the optical flow information of Dendrobium officinale includes the optical flow correlation of any two pixels, the mean motion velocity, the variance of motion velocity, the mean motion direction, and the variance of motion direction of each pixel.

[0052] Specifically, the optical flow correlation between two pixels can be calculated using the following formula:

[0053]

[0054]

[0055]

[0056] in, The optical flow correlation between the i-th pixel and the j-th pixel is given. Let be the correlation coefficient between the motion velocities of the i-th pixel and the j-th pixel. Let be the correlation coefficient between the motion directions of the i-th pixel and the j-th pixel. and To preset weights, and Greater than 0, , Let be the motion speed of the two-dimensional image corresponding to the i-th pixel in the n-th group of adjacent frames. Let be the motion speed of the two-dimensional image corresponding to the j-th pixel in the n-th group of adjacent frames. This represents the number of groups of two adjacent 2D images. Let represent the motion direction of the two-dimensional image corresponding to the i-th pixel in the n-th group of adjacent frames. Let be the motion direction of the two-dimensional image corresponding to the j-th pixel in the n-th group of adjacent frames.

[0057] Step 170: Based on the 3D point cloud data and optical flow information of Dendrobium officinale, determine whether there is an area where optical flow occurs in the leaf region of Dendrobium officinale.

[0058] Specifically, it includes:

[0059] Leaf point cloud data is extracted from 3D point cloud data of Dendrobium officinale using a leaf recognition model, where the leaf recognition model can be a convolutional neural network model.

[0060] Based on the coordinate mapping relationship between 3D point cloud and 2D image, leaf point cloud data and optical flow information of Dendrobium officinale, it is determined whether there is an optical flow generation region in the leaf area of ​​Dendrobium officinale.

[0061] Preferably, based on the coordinate mapping relationship between 3D point clouds and 2D images, leaf point cloud data, and optical flow information of *Dendrobium officinale*, it is determined whether there is an optical flow generation region in the leaf area of ​​*Dendrobium officinale*, including:

[0062] Based on the mean motion velocity, variance of motion velocity, mean motion direction, and variance of motion direction of each pixel, multiple target pixels are determined. For example, pixels whose mean motion velocity is greater than a threshold for mean motion velocity, whose variance of motion velocity is greater than a threshold for variance of motion velocity, whose mean motion direction is greater than a threshold for mean motion direction, and / or whose variance of motion direction is greater than a threshold for variance of motion direction are selected as target pixels.

[0063] Multiple optical flow regions are determined by clustering algorithms (e.g., K-means clustering, hierarchical clustering) based on the optical flow correlation and pixel distance between any two target pixels.

[0064] Based on the coordinate mapping relationship between 3D point cloud and 2D image, leaf point cloud data and multiple optical flow generation regions, it is determined whether there is an optical flow generation region in the leaf area of ​​Dendrobium officinale. For example, if the proportion of pixels in the optical flow generation region located in the leaf area of ​​Dendrobium officinale is greater than the pixel proportion threshold, it is determined that there is an optical flow generation region in the leaf area of ​​Dendrobium officinale.

[0065] Step 180: If it is determined that there is an optical flow generation area in the leaf area of ​​Dendrobium officinale, the image features of Dendrobium officinale are extracted from the two-dimensional image sequence of Dendrobium officinale through the feature extraction model.

[0066] The SWin-T model includes the ResNet50 network.

[0067] First, the two-dimensional image sequence of Dendrobium officinale is preprocessed, including cropping (384*256), scaling, mirroring, Gaussian noise, and occlusion. Then, the image features of Dendrobium officinale are extracted from the two-dimensional image sequence of Dendrobium officinale through a feature extraction model.

[0068] Step 190: Based on the image features and optical flow information of Dendrobium officinale, determine the types of diseases and pests of Dendrobium officinale using a disease and pest classification model.

[0069] Specifically, it includes:

[0070] Based on multiple sample data, a disease and pest-environment association map was established. The disease and pest-environment association map is used to record the target environmental factors and environmental characteristics corresponding to various Dendrobium officinale diseases and pests.

[0071] Obtain the growth environment information sequence of Dendrobium officinale;

[0072] Based on the correlation map between pests and diseases and the environment and the growth environment information sequence of Dendrobium officinale, the risk coefficient corresponding to each Dendrobium officinale pest and disease is calculated;

[0073] The pest and disease classification model is used to determine the types of pests and diseases of Dendrobium officinale based on image features, optical flow information, and risk coefficients corresponding to each type of Dendrobium officinale pest and disease. The pest and disease classification model includes the ResNet50 network.

[0074] As a preferred approach, a disease and pest-environment correlation map is constructed based on multiple sample data, including:

[0075] Based on the types of diseases and pests in each Dendrobium officinale sample, multiple Dendrobium officinale samples were grouped to determine multiple Dendrobium officinale sample groups, where one type of disease and pest corresponds to one Dendrobium officinale sample group.

[0076] For each Dendrobium officinale sample group, based on the growth environment information sequence of the Dendrobium officinale samples included in the sample Dendrobium officinale sample group, the difference parameter corresponding to each growth environment factor is calculated. The target environment factor corresponding to the sample Dendrobium officinale sample group is screened according to the difference parameter corresponding to each growth environment factor. Based on the target environment factor corresponding to the sample Dendrobium officinale sample group, the environmental characteristics corresponding to the sample Dendrobium officinale sample group are extracted.

[0077] Based on the target environmental factors and environmental characteristics corresponding to each Dendrobium officinale group, a disease and pest-environment correlation map was established.

[0078] The method for calculating the differential parameters corresponding to each growth environment factor is similar to the method for calculating the differential parameters corresponding to image feature factors, and will not be repeated here.

[0079] The growth environment factors with differential parameters less than the threshold of the second differential parameter can be used as the target environment factors corresponding to the sample Dendrobium officinale group.

[0080] As a preferred method, based on the correlation map between pests and diseases and the environment, and the growth environment information sequence of Dendrobium officinale, the risk coefficient corresponding to each Dendrobium officinale pest and disease is calculated, including:

[0081] Based on the target environmental factors corresponding to each Dendrobium officinale disease and pest, the environmental characteristics of Dendrobium officinale are extracted from the growth environment information sequence of Dendrobium officinale;

[0082] Based on the environmental characteristics of Dendrobium officinale and the correlation map between pests and diseases and the environment, the risk prediction model is used to calculate the risk coefficient corresponding to each Dendrobium officinale pest and disease. The risk prediction model can be a convolutional neural network model.

[0083] Figure 2 This is a schematic diagram of the modules of the intelligent detection and classification system for Dendrobium officinale diseases and pests, as shown in some embodiments of this specification. Figure 2 As shown, the intelligent detection and classification system for diseases and pests of Dendrobium officinale can include a sample acquisition module, a model building module, a data acquisition module, a 3D reconstruction module, an image acquisition module, an optical flow calculation module, a region determination module, a feature extraction module, and a disease and pest classification module.

[0084] The sample acquisition module can be used to acquire multiple sample data, including two-dimensional image sequences, optical flow information, and pest and disease types of Dendrobium officinale samples.

[0085] The model building module can be used to build optical flow calculation models, feature extraction models, and pest and disease classification models based on multiple sample data.

[0086] The data acquisition module can be used to collect three-dimensional data of Dendrobium officinale using a 3D camera;

[0087] The 3D reconstruction module can be used to obtain the 3D outline of Dendrobium officinale based on its 3D data using 3D reconstruction technology.

[0088] The image acquisition module can be used to acquire two-dimensional image sequences of Dendrobium officinale using a 2D camera, wherein the two-dimensional image sequence includes two-dimensional images at multiple time points;

[0089] The optical flow calculation module can be used to determine the optical flow information of Dendrobium officinale based on the two-dimensional image sequence of Dendrobium officinale through the optical flow calculation model;

[0090] The region determination module can be used to determine whether there is a region where light flow occurs in the leaf area of ​​Dendrobium officinale based on the 3D outline and light flow information of Dendrobium officinale.

[0091] The feature extraction module can be used to extract image features of Dendrobium officinale from the two-dimensional image sequence of Dendrobium officinale if the region determination module determines that there is an optical flow generation region in the leaf area of ​​Dendrobium officinale.

[0092] The pest and disease classification module can be used to determine the types of pests and diseases of Dendrobium officinale based on the image features and optical flow information of Dendrobium officinale using a pest and disease classification model.

[0093] The intelligent detection and classification system for Dendrobium officinale diseases and pests can apply intelligent detection and classification methods for Dendrobium officinale diseases and pests. For more details on the intelligent detection and classification system for Dendrobium officinale diseases and pests, please refer to the relevant descriptions of intelligent detection and classification methods for Dendrobium officinale diseases and pests, which will not be repeated here.

[0094] Finally, it should be understood that the embodiments described in this specification are merely illustrative of the principles of the embodiments described herein. Other variations may also fall within the scope of this specification. Therefore, alternative configurations of the embodiments described herein are intended to be illustrative rather than limiting, and should be considered consistent with the teachings of this specification. Accordingly, the embodiments described herein are not limited to those explicitly introduced and described herein.

Claims

1. A method for intelligent detection and classification of diseases and pests in Dendrobium officinale, characterized in that, include: Multiple sample data were acquired, including two-dimensional image sequences, optical flow information, and pest and disease types of Dendrobium officinale samples. Based on the aforementioned sample data, an optical flow calculation model, a feature extraction model, and a pest and disease classification model are established. Three-dimensional data of Dendrobium officinale were collected using a 3D camera; Based on the three-dimensional data of Dendrobium officinale, 3D point cloud data of Dendrobium officinale is obtained using three-dimensional reconstruction technology. A two-dimensional image sequence of Dendrobium officinale was acquired using a 2D camera, wherein the two-dimensional image sequence includes two-dimensional images at multiple time points; The optical flow information of Dendrobium officinale is determined based on the two-dimensional image sequence of the Dendrobium officinale using the optical flow calculation model. Based on the 3D point cloud data and optical flow information of the Dendrobium officinale, it is determined whether there is an optical flow generation area in the leaf region of the Dendrobium officinale; If it is determined that there is a region where light flow occurs in the leaf area of ​​the Dendrobium officinale, the image features of Dendrobium officinale are extracted from the two-dimensional image sequence of the Dendrobium officinale using a feature extraction model; The pest and disease classification model is used to determine the pest and disease types of Dendrobium officinale based on its image features and optical flow information. The feature is that, based on the multiple sample data, a feature extraction model is established, including: Determine multiple image feature factors; For each of the Dendrobium officinale samples, the first image feature of the Dendrobium officinale sample is extracted based on the plurality of image feature factors and the two-dimensional image sequence of the Dendrobium officinale sample; For each image feature factor, the difference parameter corresponding to the image feature factor is calculated based on the initial image features of each sample data. Based on the difference parameter corresponding to each of the image feature factors, a plurality of target image feature factors are determined from the plurality of image feature factors; For each of the aforementioned Dendrobium officinale samples, the second image features of the Dendrobium officinale samples are extracted based on the plurality of target image feature factors; The feature extraction model is established based on the two-dimensional image sequence and second image features of multiple samples of Dendrobium officinale. The sample data also includes a sequence of growth environment information for the Dendrobium officinale samples, wherein the growth environment information sequence includes growth environment information at multiple time points: Based on the image features and optical flow information of *Dendrobium officinale*, the pest and disease classification model determines the types of pests and diseases affecting *Dendrobium officinale*, including: Based on the multiple sample data, a disease and pest-environment association map is established, wherein the disease and pest-environment association map is used to record the target environmental factors and environmental characteristics corresponding to various Dendrobium officinale diseases and pests; Obtain the growth environment information sequence of Dendrobium officinale; Based on the disease and pest-environment correlation map and the growth environment information sequence of Dendrobium officinale, calculate the risk coefficient corresponding to each Dendrobium officinale disease and pest; The pest and disease classification model determines the pest and disease types of Dendrobium officinale based on its image features, optical flow information, and risk coefficients corresponding to each type of Dendrobium officinale pest and disease.

2. The intelligent detection and classification method for diseases and pests of Dendrobium officinale according to claim 1, characterized in that, Based on the two-dimensional image sequence of *Dendrobium officinale*, the optical flow information of *Dendrobium officinale* is determined using the optical flow calculation model, including: For each pair of two adjacent frames of the two-dimensional image sequence of Dendrobium officinale, the optical flow calculation model determines the motion speed and direction of each pixel corresponding to the two adjacent frames based on the two adjacent frames. Based on the motion speed and direction of each pixel corresponding to the two-dimensional images of two adjacent frames in each group, the optical flow correlation of any two pixels, the mean motion speed, the variance of motion speed, the mean of motion direction, and the variance of motion direction of each pixel are determined. The optical flow information of Dendrobium officinale includes the optical flow correlation of any two pixels, the mean motion speed, the variance of motion speed, the mean of motion direction, and the variance of motion direction of each pixel.

3. The intelligent detection and classification method for diseases and pests of Dendrobium officinale according to claim 1, characterized in that, Based on the 3D point cloud data and optical flow information of the Dendrobium officinale, it is determined whether there is an optical flow generation region in the leaf area of ​​the Dendrobium officinale, including: Leaf point cloud data is extracted from the 3D point cloud data of the Dendrobium officinale using a leaf recognition model. Based on the coordinate mapping relationship between 3D point cloud and 2D image, leaf point cloud data and optical flow information of Dendrobium officinale, it is determined whether there is an optical flow generation region in the leaf area of ​​Dendrobium officinale.

4. The intelligent detection and classification method for diseases and pests of Dendrobium officinale according to claim 3, characterized in that, Based on the coordinate mapping relationship between 3D point clouds and 2D images, leaf point cloud data, and optical flow information of the *Dendrobium officinale*, it is determined whether there is an optical flow generation region in the leaf area of ​​the *Dendrobium officinale*, including: Multiple target pixels are determined based on the mean motion velocity, variance of motion velocity, mean motion direction, and variance of motion direction of each pixel. The multiple target pixels are clustered using a clustering algorithm based on the optical flow correlation and pixel distance between any two target pixels to determine multiple optical flow occurrence regions; Based on the coordinate mapping relationship between 3D point cloud and 2D image, leaf point cloud data and the multiple optical flow generation regions, it is determined whether there are optical flow generation regions in the leaf area of ​​the Dendrobium officinale.

5. The intelligent detection and classification method for diseases and pests of Dendrobium officinale according to claim 4, characterized in that, Based on the aforementioned sample data, a correlation map between pests and diseases and the environment was established, including: Based on the types of diseases and pests affecting each sample of Dendrobium officinale, multiple samples of Dendrobium officinale were grouped to determine multiple Dendrobium officinale sample groups; For each Dendrobium officinale sample group, based on the growth environment information sequence of the Dendrobium officinale samples included in the sample Dendrobium officinale sample group, the difference parameter corresponding to each growth environment factor is calculated. The target environment factor corresponding to the sample Dendrobium officinale sample group is screened according to the difference parameter corresponding to each growth environment factor. The environmental characteristics corresponding to the sample Dendrobium officinale sample group are extracted according to the target environment factor corresponding to the sample Dendrobium officinale sample group. Based on the target environmental factors and environmental characteristics corresponding to each Dendrobium officinale sample group, a disease and pest-environment correlation map is established.

6. The intelligent detection and classification method for diseases and pests of Dendrobium officinale according to claim 5, characterized in that, Based on the disease / pest-environment correlation map and the growth environment information sequence of *Dendrobium officinale*, the risk coefficient corresponding to each *Dendrobium officinale* disease / pest is calculated, including: Based on the target environmental factors corresponding to each Dendrobium officinale disease and pest, the environmental characteristics of Dendrobium officinale are extracted from the growth environment information sequence of Dendrobium officinale; Based on the environmental characteristics of Dendrobium officinale and the correlation map between pests and diseases and the environment, the risk coefficient corresponding to each Dendrobium officinale pest and disease is calculated using a risk prediction model.

7. The intelligent detection and classification method for diseases and pests of Dendrobium officinale according to any one of claims 1-3, characterized in that, The optical flow calculation model is the SEA-RAFT optical flow calculation model; The pest and disease classification model includes the ResNet50 network.

8. An intelligent detection and classification system for diseases and pests of Dendrobium officinale, characterized in that, The method for intelligent detection and classification of diseases and pests of Dendrobium officinale according to any one of claims 1-7 includes: The sample acquisition module is used to acquire multiple sample data, wherein the sample data includes a two-dimensional image sequence, optical flow information and pest and disease types of Dendrobium officinale samples; The model building module is used to build an optical flow calculation model, a feature extraction model, and a pest and disease classification model based on the multiple sample data. The data acquisition module is used to collect three-dimensional data of Dendrobium officinale using a 3D camera; The three-dimensional reconstruction module is used to obtain the 3D outline of Dendrobium officinale based on the three-dimensional data of Dendrobium officinale using three-dimensional reconstruction technology. An image acquisition module is used to acquire a two-dimensional image sequence of Dendrobium officinale using a 2D camera, wherein the two-dimensional image sequence includes two-dimensional images at multiple time points; The optical flow calculation module is used to determine the optical flow information of Dendrobium officinale based on the two-dimensional image sequence of Dendrobium officinale using the optical flow calculation model; The region determination module is used to determine whether there is a region where light flow occurs in the leaf area of ​​the Dendrobium officinale based on the 3D outline and light flow information of the Dendrobium officinale. The feature extraction module is used to extract image features of Dendrobium officinale from the two-dimensional image sequence of Dendrobium officinale by means of a feature extraction model if the region determination module determines that there is an optical flow generation region in the leaf region of Dendrobium officinale. The pest and disease classification module is used to determine the pest and disease type of Dendrobium officinale based on the image features and optical flow information of Dendrobium officinale using the pest and disease classification model.

Images