A method for calculating the severity of kiwifruit diseases

Abstract

This invention provides a method for calculating the severity of kiwifruit disease, addressing issues such as determining the severity of kiwifruit tree diseases. The method includes the following steps: S1: data collection and preprocessing; S2: disease identification and scoring; and S3: disease prevention and control decision-making based on the scores. This method offers advantages such as high disease identification accuracy and the ability to automatically generate prevention and control measures.

Description

Technical Field

[0001] The invention belongs to the technical field of disease and insect pest control, and particularly relates to a method for calculating the severity of kiwifruit diseases. Background Art

[0002] Currently, the main diseases affecting kiwifruit trees include canker, bacterial blossom end rot, root knot nematode disease, root rot, leaf blight, and black spot. Key insect pests include scarab beetles, wax beetles, stink bugs, small firewood beetles, spider mites, peach borers, leafhoppers, and crickets. Chemical, biological, and physical control methods are used depending on the specific disease. However, in the actual kiwifruit disease control process, growers lack timely information on the type and severity of the disease, failing to promptly alert and implement appropriate control measures, resulting in economic losses during kiwifruit cultivation.

[0003] To address the shortcomings of existing technologies, researchers have conducted extensive research and proposed a variety of solutions. For example, a Chinese patent document [202021328897.8] discloses a tomato disease identification system based on a cuckoo search algorithm. The system comprises an image acquisition module and a disease identification module. The image acquisition module includes five submodules: an STM32F765 module, a camera module, a motor drive module, an SD card storage module, and a battery module. The disease identification module serves as the host computer analysis platform. The STM32F765 module is the primary control component of the image acquisition module, controlling the coordinated operation of its various submodules.

[0004] The above solution solves the problem of pest and disease early warning to a certain extent, but the solution still has many shortcomings, such as the inability to timely determine the type of disease and its severity. Summary of the Invention

[0005] The purpose of the present invention is to provide a kiwifruit disease severity calculation method with a reasonable design that can timely determine the type and severity of kiwifruit diseases in order to solve the above problems.

[0006] To achieve the above object, the present invention adopts the following technical solution: a method for calculating the severity of kiwifruit diseases, comprising the following steps:

[0007] S1: data collection and preprocessing;

[0008] S2: disease identification and scoring;

[0009] S3: Make disease prevention and control decisions based on the scores.

[0010] In the above-mentioned method for calculating the severity of kiwifruit diseases, step S1 includes the following steps:

[0011] S11: Collect kiwifruit tree image data in real time using a high-definition camera and build a dataset;

[0012] S12: Preprocess the images in the collected data set.

[0013] In the above-mentioned method for calculating the severity of kiwifruit diseases, step S11 includes the following steps:

[0014] S111: The high-definition camera collects images of kiwi fruit trees in different areas and deletes irrelevant images;

[0015] S112: Label the image location and acquisition time for the dataset.

[0016] In the above-mentioned method for calculating the severity of kiwifruit diseases, step S12 includes the following steps:

[0017] S121: filtering out images with abnormal sizes and performing denoising on the images;

[0018] S122: Standardize the size and pixels of the image;

[0019] S123: Binarize and perform morphological processing on the grayscale image.

[0020] In the above-mentioned method for calculating the severity of kiwifruit diseases, step S2 includes the following steps:

[0021] S21: Use convolutional neural networks to analyze the preprocessed images and identify the types of kiwifruit diseases;

[0022] S22: Based on the identified disease type and area size, combined with the preset scoring rules, a comprehensive score of the severity of the kiwifruit disease is calculated.

[0023] In the above-mentioned method for calculating the severity of kiwifruit diseases, step S21 includes the following steps:

[0024] S211: Extract feature maps from the original training samples, including conv1, conv2, conv3 and conv4, to obtain basic feature maps;

[0025] S212: The feature map extracted in the initial layer is processed by a convolution operator and a merging operator to extract a multi-scale feature map, including several feature merging maps and convolution feature maps;

[0026] S213: The initial layer of the multi-scale convolutional layer includes several convolutional layers and pooling layers;

[0027] S214: Integrate the multi-scale feature maps and use the Cancat function to enable the network model to obtain the disease characteristics of kiwifruit trees;

[0028] S215: Cancat function includes several convolutional layers;

[0029] S216: Provide the fused feature map to the convolutional layer conv5;

[0030] S217: Input the feature map of the convolutional layer conv5 into the Softmax classifier to identify the type of kiwi disease.

[0031] In the above-mentioned method for calculating the severity of kiwifruit diseases, step S22 includes the following steps:

[0032] S221: setting different scores for different types of diseases, and multiplying the scores by the area of the diseased area to obtain the severity score of the corresponding type of disease;

[0033] S222: Obtain the disease area density by the area of the diseased area and the total area of the region where it is located;

[0034] S223: Multiplying the disease severity scores of different types in each area by the diseased area to obtain a scoring matrix for different detection areas and different types of diseases;

[0035] S224: Establish a prevention and control cost matrix based on the types of diseases and different prevention and control measures;

[0036] S225: Multiply the scoring matrix and the prevention and control cost matrix to obtain the prevention and control cost matrix of each area.

[0037] In the above-mentioned method for calculating the severity of kiwifruit diseases, step S3 includes the following steps:

[0038] S31: By obtaining the matrix of each region and its prevention and control cost;

[0039] S32: Setting the total cost threshold for prevention and control;

[0040] S33: Calculate the total cost of each prevention and control measure and determine whether it exceeds a threshold;

[0041] S34: Set different control effect scores for different control measures;

[0042] S35: accumulating and calculating the total score of the control effect of each control measure;

[0043] S36: Select the control measure with the highest total score of control effect as the disease control decision.

[0044] In the above-mentioned method for calculating the severity of kiwifruit diseases, the prevention and control measures in steps S33-S36 include single measures and combined measures.

[0045] In the above-mentioned method for calculating the severity of kiwifruit diseases, the data in step S1 includes images of the trunk, main vines, side vines, fruiting mother branches, and fruiting branches of the kiwifruit tree.

[0046] Compared with existing technologies, the advantages of the present invention are: real-time collection of kiwifruit tree image data, timely identification of disease types using a convolutional neural network algorithm, and calculation of severity through scoring, so that managers can obtain the disease and insect pest status of kiwifruit trees in a timely manner; separate scores are given for diseases and insect pests in different areas, which more intuitively reflects the overall kiwifruit planting status; and prevention and control measures are automatically recommended based on the scores, selecting the best combination of prevention and control measures while limiting prevention and control costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is an overall diagram of the method of the present invention;

[0048] Figure 2 is a scoring flow chart of the present invention;

[0049] Figure 3 It is a flow chart of prevention and treatment decision making of the present invention. DETAILED DESCRIPTION

[0050] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0051] like Figure 1-3 As shown in FIG, a method for calculating the severity of kiwifruit diseases is provided. The method collects image information of kiwifruit trees to identify diseases, scores the severity of diseases in each area, and formulates reasonable countermeasures based on the degree and type of diseases in each area. The method specifically includes the following steps:

[0052] S1: data collection and preprocessing;

[0053] S2: disease identification and scoring;

[0054] S3: Make disease prevention and control decisions based on the scores.

[0055] Specifically, similar to the conventional data collection method, step S1 in this application includes the following steps:

[0056] S11: Collect kiwifruit tree image data in real time using a high-definition camera and build a dataset;

[0057] S12: Preprocess the images in the collected data set.

[0058] In depth, step S11 adopts a manual method to construct a data set, which includes the following steps:

[0059] S111: A high-definition camera collects images of kiwifruit trees in different areas and manually deletes irrelevant images. The images mainly capture images of kiwifruit branches, leaves, and trunks.

[0060] S112: Mark the image position and acquisition time for the data set, and establish an image data set based on the time axis according to the acquisition time. Image changes in the same image acquisition area are also included in the processing flow.

[0061] Furthermore, in order to ensure that the sizes and specifications of the images are consistent and facilitate subsequent convolutional neural network analysis and processing, step S12 adopts the following steps:

[0062] S121: filtering out images with abnormal sizes and performing denoising on the images;

[0063] S122: Standardize the size and pixels of the image;

[0064] S123: Binarize and perform morphological processing on the grayscale image.

[0065] Furthermore, step S2 includes the following steps:

[0066] S21: Use convolutional neural networks to analyze the preprocessed images and identify the types of kiwifruit diseases;

[0067] S22: Based on the identified disease type and area size, combined with the preset scoring rules, a comprehensive score of the severity of the kiwifruit disease is calculated.

[0068] In addition, the convolutional neural network algorithm used in step S21 specifically includes the following steps:

[0069] S211: Extract feature maps from the original training samples, including conv1, conv2, conv3 and conv4, to obtain basic feature maps;

[0070] S212: The feature map extracted in the initial layer is processed by a convolution operator and a merging operator to extract a multi-scale feature map, including several feature merging maps and convolution feature maps;

[0071] S213: The initial layer of the multi-scale convolutional layer includes several convolutional layers and pooling layers;

[0072] S214: Integrate the multi-scale feature maps and use the Cancat function to enable the network model to obtain the disease characteristics of kiwifruit trees;

[0073] S215: Cancat function includes several convolutional layers;

[0074] S216: Provide the fused feature map to the convolutional layer conv5;

[0075] S217: The feature map from convolutional layer conv5 is fed into the Softmax classifier to identify the type of kiwifruit disease. The algorithm above uses feature extraction to identify the disease type, and the specific severity is then comprehensively considered by the subsequent scoring system.

[0076] Meanwhile, step S22 includes the following steps:

[0077] S221: Setting different scores for different types of diseases, and multiplying the scores by the area of the diseased area to obtain the severity score of the corresponding type of disease;

[0078] S222: Obtain the disease area density by the area of the diseased area and the total area of the region where it is located;

[0079] S223: Multiplying the disease severity scores of different types in each area by the diseased area to obtain a scoring matrix for different detection areas and different types of diseases;

[0080] S224: Establish a prevention and control cost matrix based on the types of diseases and different prevention and control measures;

[0081] S225: Multiply the scoring matrix by the prevention and control cost matrix to obtain the prevention and control cost matrix for each region. The local scoring matrix is shown in the following table:

[0082]

[0083] The local matrix of different disease types and control measures costs is shown below:

[0084]

[0085] Multiplying the above scoring matrix with the cost matrix, the control cost of each region is obtained as

[0086]

[0087] It can be seen that step S3 includes the following steps:

[0088] S31: Obtain the cost matrix of each region and its prevention and control;

[0089] S32: Setting the total cost threshold for prevention and control;

[0090] S33: Calculate the total cost of each prevention and control measure and determine whether it exceeds a threshold;

[0091] S34: Set different control effect scores for different control measures;

[0092] S35: accumulating and calculating the total score of the control effect of each control measure;

[0093] S36: Select the control measure with the highest total score of control effect as the disease control decision.

[0094] In this example, the total cost threshold for control is set at 4. For Region 1, a combination of lime sulfur and streptomycin is chosen, with a control effectiveness score of 3.4. A combination of abamectin and streptomycin achieves a control effectiveness score of 4.6. Given the total cost, this combination is superior to the lime sulfur and streptomycin combination. Similarly, combinations of control measures are obtained for all total cost thresholds, and the optimal control ratio is selected as the recommended decision.

[0095] Obviously, the control measures in steps S33-S36 include both single measures and combined measures. Generally, the combined measures offer better overall effectiveness and cost-effectiveness than single measures. In addition to the aforementioned chemical treatments, biological and physical control measures can also be introduced, with comprehensive comparisons conducted to select the optimal control combination.

[0096] Preferably, the data in step S1 includes images of the trunk, main vines, side vines, fruiting mother branches, and fruiting branches of the kiwifruit tree. For different parts of the kiwifruit tree, the disease influencing factors are also different, and the prevention and control costs are also adjusted accordingly.

[0097] In summary, the principle of this embodiment is to collect image information of kiwifruit trees in different areas, use convolutional neural networks to determine the type and severity of diseases, then use a weighted algorithm to establish a corresponding prevention and control measure matrix, and select the optimal prevention and control combination after scoring for recommendation.

[0098] The specific embodiments described herein are merely illustrative of the spirit of the present invention. Persons skilled in the art may make various modifications, additions, or substitutions to the described specific embodiments without departing from the spirit of the present invention or exceeding the scope of the appended claims.

[0099] Although this document frequently uses terms such as convolutional neural network and classifier, the use of other terms is not excluded. These terms are used solely to more conveniently describe and explain the essence of the present invention; interpreting them as any additional limitations is contrary to the spirit of the present invention.

Claims

1. A method for calculating the severity of kiwifruit diseases, characterized in that: The steps include: S1: data collection and preprocessing; S2: disease identification and scoring; S21: Use convolutional neural networks to analyze the preprocessed images and identify the types of kiwifruit diseases; S211: Extract feature maps from the original training samples, including conv1, conv2, conv3 and conv4, to obtain basic feature maps; S212: The feature map extracted in the initial layer is processed by a convolution operator and a merging operator to extract a multi-scale feature map, including several feature merging maps and convolution feature maps; S213: The initial layer of the multi-scale convolutional layer includes several convolutional layers and pooling layers; S214: Integrate the multi-scale feature maps and use the Cancat function to enable the network model to obtain the disease characteristics of kiwifruit trees; S215: Cancat function includes several convolutional layers; S216: Provide the fused feature map to the convolutional layer conv5; S217: Input the feature map of the convolutional layer conv5 into the Softmax classifier to identify the disease types of kiwifruit; S22: Based on the identified disease type and area size, combined with the preset scoring rules, a comprehensive calculation is made to calculate the severity score of the kiwifruit disease; S221: setting different scores for different types of diseases, and multiplying the scores by the area of the diseased area to obtain the severity score of the corresponding type of disease; S222: Obtain the disease area density by the area of the diseased area and the total area of the region where it is located; S223: Multiplying the disease severity scores of different types in each area by the diseased area to obtain a scoring matrix for different detection areas and different types of diseases; S224: Establish a prevention and control cost matrix based on the types of diseases and different prevention and control measures; S225: Multiply the scoring matrix and the prevention and control cost matrix to obtain the prevention and control cost matrix of each area; S3: Make disease prevention and control decisions based on the scores.

2. A method for calculating the severity of kiwifruit diseases according to claim 1, characterized in that: The step S1 includes the following steps: S11: Collect kiwifruit tree image data in real time using a high-definition camera and build a dataset; S12: Preprocess the images in the collected data set.

3. A method for calculating the severity of kiwifruit diseases according to claim 2, characterized in that: The step S11 includes the following steps: S111: The high-definition camera collects images of kiwi fruit trees in different areas and deletes irrelevant images; S112: Label the image location and acquisition time for the dataset.

4. A method for calculating the severity of kiwifruit diseases according to claim 2, characterized in that: The step S12 includes the following steps: S121: filtering out images with abnormal sizes and performing denoising on the images; S122: Standardize the size and pixels of the image; S123: Binarize and perform morphological processing on the grayscale image.

5. The method for calculating the severity of kiwifruit diseases according to claim 1, wherein: The step S3 includes the following steps: S31: Obtain the cost matrix of each region and its prevention and control; S32: Setting the total cost threshold for prevention and control; S33: Calculate the total cost of each prevention and control measure and determine whether it exceeds a threshold; S34: Set different control effect scores for different control measures; S35: accumulating and calculating the total score of the control effect of each control measure; S36: Select the control measure with the highest total score of control effect as the disease control decision.

6. A method for calculating the severity of kiwifruit diseases according to claim 5, characterized in that: The control measures in steps S33-S36 include single measures and combined measures.

7. A method for calculating the severity of kiwifruit diseases according to claim 1, characterized in that: The data in step S1 include images of the trunk, main vines, side vines, fruiting mother branches, and fruiting branches of the kiwifruit tree.

Images