A method and system for accurate identification of crop diseases

Abstract

The application provides a crop disease precision identification method and system, and relates to the technical field of crop disease identification. The method comprises the following steps: establishing standard lesion feature image datasets of different crops and different organs; constructing a background feature database of common crop diseases; forming a crop type and organ identification algorithm model and an organ lesion identification algorithm model; determining the crop type and the organ of the crop in an image to be identified based on the crop type and organ identification algorithm model; determining the disease range of the image to be identified based on the organ of the crop and using the organ lesion identification algorithm model corresponding to the organ of the crop, and outputting a lesion disease with similar features; outputting a matched crop disease type based on the background feature database of common crop diseases; and obtaining a common disease type in the determined lesion disease and the determined crop disease type, so as to realize the precision identification of the disease. The application classifies and grades to form a structured disease identification process, and improves the identification precision.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of crop disease identification, and particularly relates to a crop disease accurate identification method and system. BACKGROUND

[0002] The statements in this section merely provide background information related to the application and do not necessarily constitute prior art.

[0003] As a typical representative of traditional open planting, field crops are extremely susceptible to external environmental interference, leading to disease occurrence and endangering food security.

[0004] In the traditional sense, disease identification and prevention basically rely on planting experience or analysis and judgment of plant pathology knowledge. Traditional disease identification consumes a lot of manpower and often leads to misjudgment due to empiricism, and cannot realize intelligentization.

[0005] Since the 21st century, advanced technologies such as artificial intelligence, big data and computer vision have become increasingly mature. Many scholars have attempted to study intelligent identification of crop diseases and have made many breakthroughs in intelligent technology. However, the identification results obtained by relying solely on artificial intelligence technology are not ideal. Existing intelligent agricultural crop disease identification technologies and methods are mostly focused on simple disease image feature recognition, establishment of general image data sets, and training and testing through accumulation of a large amount of image data. Due to the high similarity of phenotypic traits among crops, simple single-part image feature recognition has poor robustness, and the collection of a large amount of training data is time-consuming and labor-intensive, making it difficult to improve recognition accuracy and long cycle.

[0006] The inventors found that the defects of the prior art that the accuracy is difficult to improve are caused by unclear and unclear data sets, unclear characteristics and rules of disease occurrence and unclear identification points. Only seeking breakthroughs from the aspects of related algorithms and technologies of artificial intelligence recognition have little effect on improving the accuracy of disease identification. SUMMARY

[0007] To overcome the above-mentioned deficiencies of the prior art, the present application provides a crop disease accurate identification method and system, which is realized by combining artificial intelligence algorithm identification technology with disease occurrence positioning and agronomic characteristics, forming a structured disease identification process through classification and grading, and improving the identification accuracy.

[0008] To achieve the above-mentioned purpose, one or more embodiments of the present application provide the following technical solutions:

[0009] The present application provides a crop disease accurate identification method.

[0010] A crop disease accurate identification method, comprising the following steps:

[0011] Step 1: Collect standard feature images of common crop diseases and establish a dataset of standard lesion feature images for different organs of different crops;

[0012] Step 2: Construct a background characteristic database of common crop diseases, including the main plant organs infected by the diseases, the months in which they often occur, and the environmental conditions under which they are prone to occur.

[0013] Step 3: Using Faster R-CNN convolutional neural network and target detection technology, we can achieve algorithmic identification of the main crop species and organs, and organ-specific lesions in the target region of the standard lesion feature image dataset, thus forming a crop species and organ identification algorithm model and an organ-specific lesion identification algorithm model.

[0014] Step 4: Acquire the image to be identified, and determine the diseased crop type and diseased organ in the image based on the crop type and organ identification algorithm model; based on the diseased organ, determine the disease range of the image to be identified using the corresponding organ sub-organ lesion identification algorithm model, and output lesions with similar characteristics.

[0015] Step 5: Based on the background feature database of common crop diseases, associate the background month with the image acquisition time, and at the same time associate the environmental conditions in the background feature database with the environmental conditions at the time of image acquisition, and output the matching crop disease type;

[0016] Step Six: Identify the common disease types among the lesions and diseases determined in Step Four and the crop disease types determined in Step Five, to achieve accurate disease identification.

[0017] Preferably, step one specifically includes:

[0018] Collect complete plant images of the main crops in the target area and construct a crop feature dataset;

[0019] Collect characteristic images of lesions of common diseases in the main organs of crops in the target area, and construct a dataset of lesions in the organs.

[0020] The crop feature dataset and the organ-specific lesion dataset were used as the standard lesion feature image dataset.

[0021] Preferably, in step two, the background feature database is determined based on knowledge and experience related to plant pathology and agronomy.

[0022] Preferably, step four specifically includes:

[0023] Step S401: Obtain two images of the same crop to be identified, wherein image 1 contains the complete plant of the diseased crop and image 2 contains the specific lesions of the diseased crop in image 1.

[0024] Step S402: Input the image to be identified 1 into the crop type and organ identification algorithm model, and use target detection technology to determine the type of diseased crop and the organ where the disease occurs;

[0025] Step S403: Based on the organ where the disease occurs, input the image to be identified 2 into the sub-organ lesion identification algorithm model of the corresponding organ where the disease occurs, locate the disease range to which the lesion belongs, and output lesions with similar characteristics.

[0026] Preferably, step S402 specifically includes:

[0027] Input the image to be identified 1 into the crop type and organ identification algorithm model, and use target detection technology to identify the crop type of the disease to be identified;

[0028] On plants where the types of crops affected by the disease have been identified, the diseased organs of the affected crops are further identified, starting with the above-ground organs: stems, leaves, and fruits.

[0029] If no lesions are found on the above-ground organs, the crop needs to be pulled up, the roots marked, and the specific organ where the disease occurred determined by the crop type and organ identification algorithm model.

[0030] Preferably, step five specifically includes:

[0031] The acquisition time of the image to be identified is obtained, and the acquisition time of the image to be identified is associated with the background month in the background feature database. At the same time, the environmental conditions in the image to be identified, which are collected in real time by sensors and weather stations, are associated with the environmental conditions in the background feature database. The data are compared with the disease information database, and the matching crop disease type is output.

[0032] Preferably, step three further includes:

[0033] First, use the labelme annotation tool to annotate the standard lesion feature image dataset from step one:

[0034] Data annotation of crop species and organs in crop feature datasets;

[0035] Based on the unique visual characteristics of the disease, the data of organ-specific lesions in the dataset are labeled.

[0036] The second aspect of this invention provides a precise crop disease identification system.

[0037] A precise crop disease identification system includes:

[0038] The dataset construction module is configured to: collect standard feature images of common crop diseases and establish a dataset of standard lesion feature images of different organs of different crops;

[0039] The background feature database construction module is configured to: construct a background feature database of common crop diseases, including the main plant organs infected by the diseases, the months in which they occur, and the environmental conditions in which they occur.

[0040] The model determination module is configured to: use Faster R-CNN convolutional neural network and target detection technology to realize the algorithmic identification of the main crop species and organs and organ lesions in the target area of ​​the standard lesion feature image dataset, and form a crop species and organ identification algorithm model and an organ lesion identification algorithm model.

[0041] The model recognition module is configured to: acquire an image to be recognized; determine the diseased crop type and diseased organ in the image based on a crop type and organ recognition algorithm model; determine the disease range of the image based on the diseased organ using a sub-organ lesion recognition algorithm model corresponding to the diseased organ; and output lesions with similar characteristics.

[0042] The association matching module is configured to: based on the background feature database of common crop diseases, associate the background month and image acquisition time in the background feature database, the environmental conditions in the background feature database and the image acquisition environmental conditions, and output the matched crop disease type;

[0043] The common disease identification module is configured to identify common disease types among the lesions and diseases identified in the model identification module and the crop disease types identified in the association matching module, thereby achieving accurate disease identification.

[0044] A third aspect of the present invention provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps of the crop disease accurate identification method as described in the first aspect of the present invention.

[0045] A fourth aspect of the present invention provides an electronic device including a memory, a processor, and a program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps in the crop disease accurate identification method as described in the first aspect of the present invention.

[0046] The above one or more technical solutions have the following beneficial effects:

[0047] This invention provides a method and system for accurate identification of crop diseases. By combining artificial intelligence algorithm identification technology with disease occurrence location and agronomic characteristics, a four-level structured disease identification process is formed through classification and grading. It has low requirements for datasets, improves data annotation efficiency, enhances identification robustness, and quickly and effectively improves the accuracy of intelligent disease identification, saving time and effort and effectively improving the accuracy of disease identification.

[0048] Compared with the prior art, the present invention is more time-saving, and for the same dataset, it can improve the accuracy of disease identification by a large margin and improve efficiency quickly and effectively. The present invention is more labor-saving, highly targeted, has low requirements for datasets, and improves data annotation efficiency.

[0049] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0050] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0051] Figure 1 This is a flowchart of the method in the first embodiment. Detailed Implementation

[0052] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0053] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations of the present invention.

[0054] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0055] The overall concept proposed in this invention is as follows:

[0056] While conducting research on intelligent identification of field crop diseases, the inventors discovered that the difficulty in improving accuracy in existing technologies stemmed from the vague and unclear nature of the datasets and the lack of clarity regarding the characteristics, patterns, and key identification points of disease occurrence. These technologies merely sought breakthroughs at the level of AI-based algorithms and techniques. Through research and trial use of existing results, the inventors found that this deficiency can be addressed by combining AI algorithm identification technology with disease location and agronomic characteristics. This allows for a structured disease identification process based on classification and grading, thereby improving identification accuracy.

[0057] Therefore, this invention proposes a method and system for accurate identification of crop diseases, wherein the method for accurate identification of crop diseases includes the following steps:

[0058] (1) Constructing a disease dataset: Using ground-based image acquisition equipment or other channels, standard feature images of common diseases of specific crops are collected. Under existing network resources and planting and production conditions, data are collected to establish a standard lesion feature image dataset for different organs of different crops. This includes ① a crop feature dataset, containing complete plant images of the main crops in the target area, which must meet the resolution requirement for clearly labeling crop organs; ② a sub-organ lesion dataset, containing lesion feature images of common diseases in the sub-organs of the main crops in the target area, which must meet the requirement for labeling sub-organ lesion features.

[0059] (2) Construct a disease information database: Based on plant pathology and agronomic knowledge and experience, construct a background characteristic database of common diseases of the target crop, including the plant organs mainly infected by the diseases, the specific background months in which they often occur, and the environmental conditions in which they are prone to occur.

[0060] (3) Building an identification algorithm model: Use labeling tools such as labelme to label the features of the dataset described in (1), label the crop types and organs of dataset ①, label the organ lesions of dataset ②, and use Faster R-CNN convolutional neural network and target detection technology to realize the algorithm identification of the main crop types and organs and organ lesions in the target area of ​​(1), forming a crop type and organ identification algorithm model and an organ lesion identification algorithm model.

[0061] (4) Four-level structured classification and localization of diseases: To improve the accuracy of disease image recognition, a four-level structured classification and localization was defined. Two images to be identified were acquired using ground-based image acquisition devices such as mobile phones. Image 1 contains the complete diseased crop, and image 2 contains the specific lesions of the diseased crop in image 1. The so-called structured classification is to locate the disease layer by layer according to the disease's affiliation relationship, thereby locking in the scope of the disease.

[0062] The four-level structured classification positioning specifically includes:

[0063] First, the image to be identified 1 is input into the crop type and organ identification algorithm model, and the type of diseased crop is determined using target detection technology;

[0064] Secondly, target detection technology is used to locate the organs where the disease occurs;

[0065] Furthermore, based on the results of organ-specific identification, the image to be identified, image 2, is input into the organ-specific lesion identification algorithm model to locate the lesion's lesion range.

[0066] Finally, by combining the image acquisition time with the background month and environmental conditions of the current disease occurrence, and matching it with the disease characteristics in the disease information database in (2), the range of common diseases of the target crop under the existing conditions is obtained. This is then compared with the results obtained from lesion identification to determine the specific disease and achieve accurate disease identification.

[0067] Example 1

[0068] This embodiment discloses a method for accurate identification of crop diseases.

[0069] like Figure 1 As shown, a method for accurate identification of crop diseases includes the following steps:

[0070] Step 1: Collect standard feature images of common crop diseases and establish a dataset of standard lesion feature images for different organs of different crops;

[0071] Step 2: Construct a background characteristic database of common crop diseases, including the main plant organs infected by the diseases, the months in which they often occur, and the environmental conditions under which they are prone to occur.

[0072] Step 3: Using Faster R-CNN convolutional neural network and target detection technology, we can achieve algorithmic identification of the main crop species and organs, and organ-specific lesions in the target region of the standard lesion feature image dataset, thus forming a crop species and organ identification algorithm model and an organ-specific lesion identification algorithm model.

[0073] Step 4: Acquire the image to be identified, and determine the diseased crop type and diseased organ in the image based on the crop type and organ identification algorithm model; based on the diseased organ, determine the disease range of the image to be identified using the corresponding organ sub-organ lesion identification algorithm model, and output lesions with similar characteristics.

[0074] Step 5: Based on the background feature database of common crop diseases, associate the background month and image acquisition time in the background feature database, the environmental conditions in the background feature database and the image acquisition environmental conditions, and output the matching crop disease type;

[0075] Step Six: Identify the common disease types among the lesions and diseases determined in Step Four and the crop disease types determined in Step Five, to achieve accurate disease identification.

[0076] Furthermore, step one specifically includes:

[0077] Collect complete plant images of the main crops in the target area and construct a crop feature dataset;

[0078] Collect characteristic images of lesions of common diseases in the main organs of crops in the target area, and construct a dataset of lesions in the organs.

[0079] The crop feature dataset and the organ-specific lesion dataset were used as the standard lesion feature image dataset.

[0080] Furthermore, in step two, the background feature database is determined based on knowledge and experience related to plant pathology and agronomy.

[0081] Furthermore, in step three, the crop species and organ identification algorithm model is used to identify crop species and organs, and the organ lesion identification algorithm model is used to identify the extent of disease in the corresponding organ.

[0082] In this embodiment, the steps for forming the crop species and organ identification algorithm model and the organ-specific lesion identification algorithm model include:

[0083] (1) Data preparation and preprocessing: Extract the plant and organ parts and organ lesion parts from step one respectively. Use the diseased plant and organ parts and organ lesion parts as positive samples for model training, and the other parts as negative samples.

[0084] (2) Increase the number of positive samples;

[0085] (3) Constructing an image detection model: The image detection model includes using a Faster R-CNN convolutional neural network as the feature extraction layer to extract feature maps of the image, generating a candidate box extraction network (RegionProposal Networks), inputting the extracted feature maps into the candidate box extraction network, and using the candidate box extraction network to generate candidate boxes (region proposals); collecting the region of interest pooling layer (ROI Pooling) of the candidate boxes, using the pooling layer to integrate the feature maps and proposals, and using the integrated feature maps and proposals as input for subsequent target category discrimination; finally, passing the integrated feature maps and proposals through a classification and regression network to judge the target image and achieve classification.

[0086] (4) Divide the dataset samples in step (1) into training set, test set and validation set according to a certain ratio.

[0087] (5) Input the dataset from step (4) into the image detection model constructed in (3) to train the image detection model and obtain the crop type and organ identification algorithm model and the organ lesion identification algorithm model respectively.

[0088] The specific steps of the crop type and organ identification algorithm model and the organ-specific lesion identification algorithm model described above are existing technologies and will not be elaborated upon in this embodiment.

[0089] Furthermore, step four specifically includes:

[0090] Step S401: Obtain two images of the same crop to be identified, wherein image 1 contains the complete plant of the diseased crop and image 2 contains the specific lesions of the diseased crop in image 1.

[0091] Step S402: Input the image to be identified 1 into the crop type and organ identification algorithm model, and use target detection technology to determine the type of diseased crop and the organ where the disease occurs;

[0092] Step S403: Based on the organ where the disease occurs, input the image to be identified 2 into the sub-organ lesion identification algorithm model of the corresponding organ where the disease occurs, locate the disease range to which the lesion belongs, and output lesions with similar characteristics.

[0093] Furthermore, step S402 specifically includes:

[0094] Input the image to be identified 1 into the crop type and organ identification algorithm model, and use target detection technology to identify the crop type of the disease to be identified;

[0095] On plants where the types of crops affected by the disease have been identified, the diseased organs of the affected crops are further identified, starting with the above-ground organs: stems, leaves, and fruits.

[0096] If no lesions are found on the above-ground organs, the crop needs to be pulled up, the roots marked, and the specific organ where the disease occurred determined by the crop type and organ identification algorithm model.

[0097] Furthermore, step five specifically includes:

[0098] The acquisition time of the image to be identified is obtained, and the acquisition time of the image to be identified is associated with the background month in the background feature database. At the same time, the environmental conditions in the image to be identified, which are collected in real time by sensors and weather stations, are associated with the environmental conditions in the background feature database. The data are compared with the disease information database, and the matching crop disease type is output.

[0099] Furthermore, step three also includes:

[0100] First, use the labelme annotation tool to annotate the standard lesion feature image dataset from step one:

[0101] Data annotation of crop species and organs in crop feature datasets;

[0102] Based on the unique visual characteristics of the disease, the data of organ-specific lesions in the dataset are labeled.

[0103] This invention provides a technical solution:

[0104] A method for precise identification of crop diseases based on a four-level structured definition, used to identify common diseases of corn and soybeans, the main crops of Dahewan Farm, specifically includes the following steps:

[0105] (1) The first step in this invention is to construct a disease dataset. Ground-based image acquisition devices, such as inspection robots, mobile phones, and digital cameras, are used to collect standard feature images of common diseases in corn and soybean. Under existing network resources and planting and production conditions, data is collected to establish standard lesion feature image datasets for different organs of corn and soybean. These include: ① a crop feature dataset, containing complete plant images of corn and soybean, which must meet the resolution requirement of clearly labeling crop organs, including the roots, stems, leaves, and fruits of corn and soybean; ② a lesion dataset for each organ, containing lesion feature images of common diseases in the roots, stems, leaves, and fruits of corn and soybean, which must meet the requirement of labeling the disease characteristics of lesions in each organ.

[0106] (2) Constructing a disease information database: Based on the knowledge and experience of plant pathology and agronomy of corn and soybean, a background database of common diseases of corn and soybean is constructed respectively, including the main disease types of corn and soybean, the organs that are easily infected by the disease, the specific background months in which the disease is likely to occur, and the environmental conditions in which the disease is likely to occur.

[0107] (3) Building an identification algorithm model: First, use labeling tools such as LabelMe to label the features of the dataset described in (1). According to the different phenotypic traits and different visual manifestations of corn and soybean, the crop types and organs of dataset ① are labeled; according to the unique visual manifestation features of diseases, such as the shape, color, brightness, and depth of lesions, the organ-specific lesion data of dataset ② are labeled; then, using Faster R-CNN convolutional neural network, combined with target detection technology, the algorithm identification of the crop types and organs of corn and soybean and the organ-specific lesions in (1) is realized, forming a crop type and organ identification algorithm model and an organ-specific lesion identification algorithm model for identifying corn and soybean.

[0108] (4) This invention uses a four-level structured classification method to identify diseases in corn and soybeans. The so-called structured classification is to locate the disease layer by layer according to the disease affiliation, lock the disease range, and determine the disease type. That is, after obtaining the corn and soybean disease image information to be identified through ground acquisition devices such as mobile phones, the target detection technology is used with relevant algorithm recognition models to sequentially obtain the crop category and the crop organ where the lesion occurs. According to the unique characteristics of the disease lesion, the disease range is further located. Finally, ground sensors and weather stations are used, combined with the disease image acquisition time, to obtain the background month and environmental conditions of the current disease occurrence. Combined with the disease information database in (2), the current common disease range of corn and soybeans is obtained. The results are compared with the recognition results of the algorithm model to determine the specific disease in four steps.

[0109] Please see Figure 1 The four-level structured classification and positioning of disease identification involves the following specific steps:

[0110] a. Image acquisition: Use ground image acquisition devices such as mobile phones to acquire two images of diseased corn or soybeans to be identified. Image 1 contains the complete plant of diseased corn or soybeans, and image 2 contains the specific disease lesion features of the crop in image 1.

[0111] b. Input the image to be identified 1 into the crop type and organ identification algorithm model, use target detection technology to identify the crop with the disease to be identified, and determine whether the diseased crop is corn or soybean;

[0112] c. Further use target detection technology to continue to mark the diseased organs of the diseased crop on the marked plants, including stems, leaves and fruits. If there are no diseased lesions on the above-ground organs, they need to be pulled out and the roots marked. Through the crop type and organ recognition algorithm model, the specific organ where the disease occurs can be finally determined.

[0113] d. Based on the organ identification results of c, input the image to be identified 2 into the crop category sub-organ lesion identification algorithm model obtained by b, locate the disease range of the crop obtained by b and the organ obtained by c to which the lesion belongs, and output the lesion with similar characteristics.

[0114] e. Combine the image acquisition time with the background month in which the disease occurred and the environmental conditions collected in real time by sensors and weather stations, and compare them with the disease information database in (2) to output the matching crop disease type;

[0115] The results obtained from fe are compared with those obtained from d to output the common types of diseases, and finally the diseases are determined, thus achieving accurate identification of diseases.

[0116] Example 2

[0117] This embodiment discloses a precise crop disease identification system.

[0118] A precise crop disease identification system includes:

[0119] The dataset construction module is configured to: collect standard feature images of common crop diseases and establish a dataset of standard lesion feature images of different organs of different crops;

[0120] The background feature database construction module is configured to: construct a background feature database of common crop diseases, including the main plant organs infected by the diseases, the months in which they occur, and the environmental conditions in which they occur.

[0121] The model determination module is configured to: use Faster R-CNN convolutional neural network and target detection technology to realize the algorithmic identification of the main crop species and organs and organ lesions in the target area of ​​the standard lesion feature image dataset, and form a crop species and organ identification algorithm model and an organ lesion identification algorithm model.

[0122] The model recognition module is configured to: acquire an image to be recognized; determine the diseased crop type and diseased organ in the image based on a crop type and organ recognition algorithm model; determine the disease range of the image based on the diseased organ using a sub-organ lesion recognition algorithm model corresponding to the diseased organ; and output lesions with similar characteristics.

[0123] The association matching module is configured to: based on the background feature database of common crop diseases, associate the background month and image acquisition time in the background feature database, the environmental conditions in the background feature database and the image acquisition environmental conditions, and output the matched crop disease type;

[0124] The common disease identification module is configured to identify common disease types among the lesions and diseases identified in the model identification module and the crop disease types identified in the association matching module, thereby achieving accurate disease identification.

[0125] Example 3

[0126] The purpose of this embodiment is to provide a computer-readable storage medium.

[0127] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the crop disease accurate identification method as described in Embodiment 1 of this disclosure.

[0128] Example 4

[0129] The purpose of this embodiment is to provide an electronic device.

[0130] An electronic device includes a memory, a processor, and a program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps in the crop disease accurate identification method as described in Embodiment 1 of this disclosure.

[0131] The steps and methods involved in the apparatuses of Embodiments 2, 3, and 4 above correspond to those in Embodiment 1. For specific implementation details, please refer to the relevant description section of Embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood as including any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and enabling the processor to perform any of the methods in this invention.

[0132] Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computer device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. The present invention is not limited to any particular combination of hardware and software.

[0133] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A method for accurate identification of crop diseases, characterized in that, Includes the following steps: Step 1: Collect standard feature images of common crop diseases and establish a dataset of standard lesion feature images for different organs of different crops; Step 2: Construct a background characteristic database of common crop diseases, including the main plant organs infected by the diseases, the months in which they often occur, and the environmental conditions under which they are prone to occur. Step 3: Using Faster R-CNN convolutional neural network and target detection technology, we can achieve algorithmic identification of the main crop species and organs, and organ-specific lesions in the target region of the standard lesion feature image dataset, thus forming a crop species and organ identification algorithm model and an organ-specific lesion identification algorithm model. Step 4: Acquire the image to be identified, and determine the diseased crop type and diseased organ in the image based on the crop type and organ identification algorithm model; based on the diseased organ, determine the disease range of the image to be identified using the corresponding organ sub-organ lesion identification algorithm model, and output lesions with similar characteristics. Specifically, it includes: Step S401: Obtain two images of the same crop to be identified, wherein image 1 contains the complete plant of the diseased crop and image 2 contains the specific lesions of the diseased crop in image 1. Step S402: Input the image to be identified 1 into the crop type and organ identification algorithm model, and use target detection technology to determine the type of diseased crop and the organ where the disease occurs; Step S403: Based on the organ where the disease occurs, input the image to be identified 2 into the sub-organ lesion identification algorithm model of the corresponding organ where the disease occurs, locate the disease range to which the lesion belongs, and output lesions with similar features; Step 5: Based on the background feature database of common crop diseases, associate the background month with the image acquisition time, and simultaneously associate the environmental conditions in the background feature database with the environmental conditions at the time of image acquisition, outputting the matching crop disease type; specifically including: The acquisition time of the image to be identified is obtained, and the acquisition time of the image to be identified is associated with the background month in the background feature database. At the same time, the environmental conditions in the image to be identified, which are collected in real time by sensors and weather stations, are associated with the environmental conditions in the background feature database. The data is compared with the disease information database, and the matching crop disease type is output. Step Six: Identify the common disease types among the lesions and diseases determined in Step Four and the crop disease types determined in Step Five, to achieve accurate disease identification.

2. The method for accurate identification of crop diseases as described in claim 1, characterized in that, Step one specifically includes: Collect complete plant images of the main crops in the target area and construct a crop feature dataset; Collect characteristic images of lesions of common diseases in the main organs of crops in the target area, and construct a dataset of lesions in the organs. The crop feature dataset and the organ-specific lesion dataset were used as the standard lesion feature image dataset.

3. The method for accurate identification of crop diseases as described in claim 1, characterized in that, In step two, the background feature database is determined based on knowledge and experience related to plant pathology and agronomy.

4. The method for accurate identification of crop diseases as described in claim 1, characterized in that, Step S402 specifically includes: Input the image to be identified 1 into the crop type and organ identification algorithm model, and use target detection technology to identify the crop type of the disease to be identified; On plants where the types of crops affected by the disease have been identified, the diseased organs of the affected crops are further identified, starting with the above-ground organs: stems, leaves, and fruits. If no lesions are found on the above-ground organs, the crop needs to be pulled up, the roots marked, and the specific organ where the disease occurred determined by the crop type and organ identification algorithm model.

5. The method for accurate identification of crop diseases as described in claim 2, characterized in that, Step three also includes: First, use the labelme annotation tool to annotate the standard lesion feature image dataset from step one: Data annotation of crop species and organs in crop feature datasets; Based on the unique visual characteristics of the disease, the data of organ-specific lesions in the dataset are labeled.

6. A precise crop disease identification system, characterized in that: include: The dataset construction module is configured to: collect standard feature images of common crop diseases and establish a dataset of standard lesion feature images of different organs of different crops; The background feature database construction module is configured to: construct a background feature database of common crop diseases, including the main plant organs infected by the diseases, the months in which they occur, and the environmental conditions in which they occur. The model determination module is configured to: use Faster R-CNN convolutional neural network and target detection technology to realize the algorithmic identification of the main crop species and organs and organ lesions in the target area of ​​the standard lesion feature image dataset, and form a crop species and organ identification algorithm model and an organ lesion identification algorithm model. The model recognition module is configured to: acquire an image to be recognized; determine the diseased crop type and diseased organ in the image based on a crop type and organ recognition algorithm model; determine the disease range of the image based on the diseased organ using a sub-organ lesion recognition algorithm model corresponding to the diseased organ; and output lesions with similar characteristics. Specifically, it includes: Two images of the same crop to be identified are obtained, where image 1 contains the complete plant of the diseased crop and image 2 contains the specific lesions of the diseased crop in image 1. The image to be identified 1 is input into the crop type and organ identification algorithm model, and the target detection technology is used to determine the type of diseased crop and the organ where the disease occurs. Based on the organ where the disease occurs, the image to be identified 2 is input into the sub-organ lesion identification algorithm model of the corresponding organ where the disease occurs, the lesion range to which it belongs is located, and lesions with similar characteristics are output. The association matching module is configured to: based on a background feature database of common crop diseases, associate the background month and image acquisition time in the background feature database, the environmental conditions in the background feature database, and the image acquisition environmental conditions, and output the matched crop disease type; specifically including: The acquisition time of the image to be identified is obtained, and the acquisition time of the image to be identified is associated with the background month in the background feature database. At the same time, the environmental conditions in the image to be identified, which are collected in real time by sensors and weather stations, are associated with the environmental conditions in the background feature database. The data is compared with the disease information database, and the matching crop disease type is output. The common disease identification module is configured to identify common disease types among the lesions and diseases identified in the model identification module and the crop disease types identified in the association matching module, thereby achieving accurate disease identification.

7. A computer-readable storage medium having a program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in the crop disease accurate identification method as described in any one of claims 1-5.

8. An electronic device, comprising a memory, a processor, and a program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the crop disease accurate identification method as described in any one of claims 1-5.

Images