Extraneous and local pest image recognition multi-dimensional database construction method and system

Abstract

The invention specifically relates to the technical field of computers and the technical field of biological information, and discloses a method and a system for constructing a multi-dimensional database for image recognition of foreign and local diseases and insect pests, and the method comprises the steps: obtaining basic information, hazard feature information and other information of foreign and local diseases and insect pests through self-sampling and public resource collection and processing; analyzing the collected original data set and the scanning image set, and carrying out image preprocessing, data set division, model evaluation and improved construction on the three-dimensional internal and external models and the image recognition model; a foreign and local pest and disease damage multi-dimensional information access rule is constructed, a foreign and local pest and disease damage image recognition multi-dimensional database is developed, data support can be provided for research and development of technologies such as rapid and accurate detection and one-time detection, and the system can become a port entry plant quarantine information support system. The method can certainly play an important role in improving port quarantine and extraneous potential pest control ability, effectively reducing extraneous quarantine pest introduction and protecting agricultural production and ecological safety.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and more specifically, to a method and system for constructing a multidimensional database for image recognition of foreign and native pests and diseases. Background Art

[0002] Pests and diseases are the combination of diseases and insect pests. Diseases refer to the stunted growth, withering or death of plants caused by bacteria, fungi, viruses, algae, unsuitable climate or soil, while insect pests refer to the damage to plant growth caused by harmful insects. They have caused serious harm to agricultural production and the ecological environment, such as reducing crop yields and quality, destroying plant diversity, and spreading zoonotic diseases. While affecting human health, they have also caused huge economic losses.

[0003] In most parts of my country, pest identification mainly relies on the experience of agricultural experts and farmers. However, due to the large cultivated land area and many types of pests and diseases in my country, agricultural experts have problems such as high labor intensity, poor timeliness and strong subjectivity when they go to the site to identify and count pests and diseases. In contrast, pest identification methods based on image processing can achieve pest identification and detection without human intervention. However, due to the complex field environment and the high similarity of some pests, early image recognition software and systems have not been able to be promoted. In recent years, computer vision technology has developed rapidly. The application of hyperspectral technology, multispectral technology and deep learning methods has enabled image processing technology to identify similar targets under complex backgrounds, but there is currently no classification and identification algorithm for large numbers and types of foreign and local pests and diseases. DNA barcode molecular detection is also an emerging technology that uses a conservative segment in the DNA of an organism to quickly and accurately identify species. It can identify samples of different growth periods, different parts and different processing degrees. It is not affected by morphological characteristics, has high accuracy, and can be used for non-expert species identification.

[0004] At the same time, the existing databases of alien and native pests and diseases mainly focus on the collection of basic information and damage characteristic information. There are few databases that combine image recognition and molecular detection, and the collected species are incomplete and the information is not comprehensive and standard. Some image recognition software can identify species, but cannot obtain complete basic information, damage characteristic information and phenotypic information, especially three-dimensional structure. The three-dimensional structure can not only be used for fine observation of pests and diseases and comparative analysis of the internal and external morphological structures of pests and diseases, but also provide a data basis for the pest identification system as a training data set for deep learning algorithms to improve the accuracy of pest identification. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present invention provide a method and system for constructing a multi-dimensional database for image recognition of foreign and native pests and diseases, so as to solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solution: a method for constructing a multidimensional database for image recognition of foreign and native pests and diseases, comprising the following steps:

[0007] S1. Collect data from self-sampled samples, public resources, relevant news, laws and regulations, books, and papers to construct original datasets and scanned image sets of exotic and native pests and diseases. The original datasets include the basis and features.

[0008] S2. Analyze the collected original data sets and scanned image sets, and perform image preprocessing, data set division, model evaluation and construction improvement to build three-dimensional internal and external models and image recognition models.

[0009] S3. Design multidimensional data tables for foreign and local pests and diseases, and build data access rules to handle the interaction between front-end and back-end data;

[0010] S4. According to the data access rules, the basic information, damage characteristic information, phenotypic information, omics information and image recognition modules of alien and native pests and diseases are constructed respectively.

[0011] Preferably, the foreign and native pests and diseases collected in S1 cover all foreign and native pests and diseases within China, and the specific information should also be comprehensive, covering the basic information modules of species name, foreign name, host, morphological description, and pictures, including quarantine status, hazard characteristics, hazard level, place of origin, distribution location, transmission route, prevention and control methods, and entry quarantine requirements. The phenotypic module contains high-definition images and multispectral data, and the omics module contains genome data, DNA barcode data, and primer sequences. Preferably, the resources queried for the basic information of foreign and native pests and diseases collected in S1 are websites such as NCBI and the China Foreign and Native Pests and Diseases Information System, as well as books and papers on related foreign and native pests and diseases, and related news reports or investigation reports.

[0012] Preferably, the pictures of foreign and native pests and diseases collected in S1 should be comprehensive, including pictures collected by drones, mobile phones, cameras, microscopes, CT, three-dimensional, and hyperspectral.

[0013] Preferably, in said S2, the collected original data set is analyzed, and the collected image data needs to be subjected to image size compression and data augmentation processing. After the data processing, the data set is divided into a training set, a validation set, and a test set according to the proportion.

[0014] Preferably, in S2, the collected original data set is analyzed, the projection images of foreign and native pests and diseases are reconstructed into a data volume consisting of a slice image sequence, posture correction and effective signal amplification are performed, and a segmentation method is selected.

[0015] Preferably, in S2, a three-dimensional model is constructed using micro-computed tomography technology, that is, an X-ray generator emits X-rays to penetrate the specimen, and a detector placed at the other end records the process, generating radiographs, and obtaining three-dimensional fine structures that are difficult to obtain using traditional morphological dissection and tissue sectioning methods. The three-dimensional model is used for fine observation of foreign and native pests and diseases and comparative analysis of the internal and external morphological structures of foreign and native pests and diseases, so that the model can be directly used to distinguish between pest species. The modeling of internal morphology can achieve species distinction by using the difference in internal structure when the foreign and native pests and diseases are similar in appearance.

[0016] Preferably, the constructed internal and external structure model database of foreign and native pests and diseases can provide a data basis for pest identification and serve as a training data set for deep learning algorithms to improve the accuracy of pest identification.

[0017] Preferably, DensetNet121, MobileNetV3, ResNet101 and ShuffleNet models are selected for training experiments to construct evaluation indicators for multi-recognition object classification models. To ensure the reliability of evaluation parameters, the model sets the average precision, average recall rate and average F1-score as macro-average, and the overall accuracy rate as micro-average, and selects MobileNetV3 with better comprehensive performance.

[0018] Preferably, in S3, constructing data access rules for image recognition of foreign and native pests and diseases, that is, designing a data table for image recognition of foreign and native pests and diseases, and constructing data access rules, specifically including:

[0019] Construct basic information access rules for alien and native pests and diseases, access rules for hazard information of alien and native pests and diseases, construct phenotypic information access rules for alien and native pests and diseases, access rules for omics data modules of alien and native pests and diseases, and construct image recognition access rules for alien and native pests and diseases.

[0020] Preferably, in S4, a standardized storage is constructed for the collected data, and a database of foreign and native pests and diseases is constructed according to the omics division module. The front-end and back-end separation technology is used to read the data involving multi-dimensional modules and generate visualization.

[0021] The present invention provides a system for constructing a multidimensional database for image recognition of foreign and native pests and diseases, which specifically includes:

[0022] Data pre-collection module: collect basic information, damage characteristics, omics information and phenotypic information of foreign and native pests and diseases through public resources or self-sampled samples;

[0023] Data integration module: integrates data from different experiments or studies, preprocesses them, and sends them to the data analysis module;

[0024] Data analysis module: Analyze the collected original data sets and scanned image sets, and perform image preprocessing, data set division, model evaluation and selection to build three-dimensional internal and external models and image recognition models.

[0025] Model building module: build three-dimensional models, image recognition models, and perform model improvement and verification;

[0026] Data storage module: used to store the pre-processed basic information, damage characteristics information, and phenotypic information of foreign and local pests and diseases, and send it to the image recognition unit;

[0027] Omics processing module: reads the sequence fragments transmitted by the data interaction module, uses DNA barcoding, Blast alignment and other functions to identify pests and diseases, and sends the identification results to the data interaction module.

[0028] Image recognition module: reads the photos transmitted by the data interaction module, calls the model for recognition, and sends it to the data interaction module;

[0029] Data interaction module: based on the access rules of the basic information modules of alien and native pests and diseases, the damage characteristic information of alien and native pests and diseases, the access rules of the phenotypic modules of alien and native pests and diseases, the access rules of the omics modules of alien and native pests and diseases, and the access rules of the image recognition modules of alien and native pests and diseases, the interaction between the data storage module, the image recognition module and the data visualization module are processed, and standardized information is returned as needed.

[0030] Data Visualization Module: It is used to provide users with a search form for all functions of the interactive foreign language native pest and disease multidimensional database, visualize the analysis results through various charts, graphs and tools, and provide data mining functions.

[0031] The significant technical progress achieved by the present invention:

[0032] First, the present invention constructs a three-dimensional internal and external model, which can be directly used to distinguish pest species. The modeling of internal morphology can achieve species distinction by using the difference in internal structure when the appearance of similar foreign and native pests and diseases is similar. It provides a data basis for pest identification and serves as a training data set for deep learning algorithms to improve the accuracy of pest identification. The present invention implements an image recognition algorithm based on MobileNetV3, which can detect pests and diseases without human intervention. The present invention constructs a database system for storing image recognition information of foreign and native pests and diseases, strictly controls the data quality of basic species information, hazard characteristic information quality and phenotypic data quality, and provides the most comprehensive foreign and native pest and disease data.

[0033] Second, the multidimensional database construction method for image recognition of foreign and native pests and diseases provided by the present invention not only covers a large amount of pest and disease information, but also introduces efficient image recognition technology, molecular detection and data processing methods. The significant technological advances brought about include:

[0034] 1) Comprehensiveness of data: The present invention ensures that the information in the database is comprehensive and accurate by comprehensively collecting information from various resources (such as self-sampled samples, public resources, relevant news, laws and regulations, books, papers, etc.). This means that users can obtain comprehensive data on exotic and native pests and diseases from a centralized location.

[0035] 2) Efficient pest and disease identification: Traditional pest and disease identification mostly relies on the experience and intuitive judgment of experts. The present invention provides image recognition and molecular detection. The former utilizes advanced image processing and deep learning technology, and the latter compares DNA barcodes to automatically, quickly and accurately identify pests and diseases, greatly improving the efficiency and accuracy of identification.

[0036] 3) Multi-dimensional data analysis: In addition to basic pest and disease information, the present invention also provides damage characteristic information, phenotypic information, omics information, image recognition modules, etc., providing users with a multi-angle and multi-dimensional pest and disease data analysis method.

[0037] 4) Real-time data update and interaction: The real-time and interactive nature of the database is ensured through the designed data access rules and front-end and back-end data interaction mechanisms, allowing users to obtain the latest pest and disease information at any time and interact with the system in real time.

[0038] 5) Optimized time and resource utilization: Through effective preprocessing of acquired images, data set partitioning, and model evaluation and construction improvements, data redundancy and invalid operations are greatly reduced, and the operating efficiency of the system is improved.

[0039] 6) Ecological and economic value: The invention not only helps to better identify and prevent pests and diseases, but also provides strong support for agricultural production, ecological protection and economic development. By effectively identifying and controlling foreign pests and diseases, their impact on native ecosystems and agricultural production can be reduced, thus bringing significant ecological and economic benefits.

[0040] In short, this multidimensional database construction method for image recognition of foreign and native pests and diseases represents the cutting-edge progress of current pest and disease identification and data processing technology. It not only provides researchers and agricultural workers with an efficient and accurate tool, but also has important significance for maintaining ecological balance and promoting economic development.

[0041] Third, the expected benefits and commercial value after the transformation of the technical solution of the present invention are: the technical solution of the present invention can provide rich, complete and standard high-quality pest and disease data for customs, provide data support for the development of rapid and accurate detection, one-time detection, tracking and tracing technologies, and play an important role in protecting agricultural production and ecological security. The present invention can provide efficient, accurate and comprehensive pest and disease identification methods for agricultural production and environmental protection, specifically including image recognition and molecular detection methods, improve port quarantine and foreign potential harmful organisms, reduce agricultural input and losses, and increase agricultural income and competitiveness. The technical solution of the present invention can provide a rich, complete and standard pest and disease database and image recognition software for agricultural scientific research and education, promote agricultural innovation and development, cultivate agricultural talents and experts, and improve the level and influence of agriculture. The present invention provides two types of identification methods, image recognition of pests and diseases and molecular detection methods, instead of manually looking at them, and standardizes the collection of data to provide comprehensive, accurate and reliable data.

[0042] The technical solution of the present invention fills the technical gap in the industry at home and abroad:

[0043] In the technical solution of the present invention, in the image collection part of the database, in addition to conventional microscopes and cameras, drone photography, collection photography, CT, and hyperspectral technology are also used to fill the problem of incomplete application of image collection technology in the database of foreign and local pests and diseases.

[0044] In the technical solution of the present invention, on the premise that the alien and native pests and diseases database provides basic species data and comprehensive pictures, it provides high-precision classification and identification of a large number of alien and native pests and diseases with various types based on deep learning methods and a molecular detection method based on DNA barcodes, which solves the problem of the lack and singleness of the existing pest and disease database in identification and detection functions.

[0045] The technical solution of the present invention adopts three-dimensional structure reconstruction technology to achieve detailed observation and comparative analysis of the three-dimensional structure of pests and diseases, and solves the problems of incomplete, non-standard and non-visualized information in existing pest and disease databases.

[0046] The technical solution of the present invention adopts cloud computing and Internet of Things technologies to achieve real-time monitoring and early warning of pests and diseases, and solves the problems of poor timeliness, poor effect, high cost, etc. of existing pest and disease control.

[0047] The technical solution of the present invention solves the technical problems that people have been eager to solve but have never succeeded in solving: the technical solution of the present invention solves the problem that people have been eager to solve, how to quickly, accurately and comprehensively obtain the basic data, epidemic prevention data, phenotypic data and omics data of foreign and pests and diseases, as well as rich and comprehensive image data required for the development of reliable rapid and accurate detection, one-time detection, tracking and tracing technologies on the same platform, and realizes the function of providing data support. The technical solution of the present invention solves the problem that people have been eager to solve, how to use a variety of methods to quickly, accurately and comprehensively identify and prevent foreign and native pests and diseases in complex fields and customs quarantine situations, so as to ensure the safety and sustainable development of agricultural production and the ecological environment.

[0048] The technical solution of the present invention overcomes technical bias, does not rely on the experience of agricultural experts and farmers, and uses image recognition and molecular detection methods to improve the technical level and technical value of disease and insect pest identification and prevention.

[0049] Fourth, the present invention collects data on exotic and native pests and diseases from different sources (self-sampled samples, public resources, relevant news, laws and regulations, books, and papers). These data include basic information such as species name, foreign name, host, morphological description, pictures, and quarantine status, hazard characteristics, hazard level, origin, distribution location, transmission route, prevention and control methods, entry quarantine requirements and other hazard characteristics. In addition, it also includes high-definition images, multispectral data, genomic data, DNA barcode data, primer sequences, etc.

[0050] The collected data is preprocessed, including image size compression and data augmentation. The dataset is then divided into training, validation, and test sets. This step also involves reconstructing the projection image into a slice image sequence, performing posture correction and effective signal amplification, and selecting an appropriate segmentation method.

[0051] Design a multidimensional data table of foreign and native pests and diseases for processing front-end and back-end data interaction, and build data access rules. According to the data access rules, build a database of foreign and native pests and diseases containing basic information, damage characteristic information, phenotypic information, omics information and image recognition modules.

[0052] The advantage of the present invention is that it provides a comprehensive and detailed way to collect, process, analyze and store information about pests and diseases, which is essential for the identification, monitoring and management of pests and diseases. By integrating different types of data and using advanced image processing technology, this method can significantly improve the accuracy and efficiency of pest and disease identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 The species basic information module of the database system of the present invention Figure 1 .

[0054] Figure 2 The species basic information module of the database system of the present invention Figure 2 .

[0055] Figure 3 This is a test result diagram of the insect recognition model provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0056] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0057] The present invention adopts an image recognition model based on a convolutional neural network to solve the problem of identifying foreign and native pests and diseases, constructs an internal and external three-dimensional model of foreign and native pests and diseases to solve the problem of missing model data, designs a multidimensional database for image recognition of foreign and native pests and diseases, provides image recognition functions and basic data of species, hazard characteristic data, omics data, phenotypic data and interfaces to solve the problem of incomplete and non-standard species data in traditional recognition software. It can provide data support for the development of technologies such as rapid and accurate detection, one-time detection, and tracking and tracing, and will also become a plant quarantine information support system for my country's ports of entry, which will surely play an important role in improving the port's quarantine and foreign potential harmful organisms, effectively reducing the introduction of foreign quarantine harmful organisms, and protecting my country's agricultural production and ecological security.

[0058] See also Figure 1 As shown, the present invention provides a multidimensional database construction system for image recognition of foreign and native pests and diseases, including: a data pre-collection module, a data integration module, a data analysis module, a model construction module, a data storage module, an omics processing module, an image recognition module, a data interaction module and a data visualization module.

[0059] The data pre-collection module is connected to the data integration module, the data integration module is connected to the data analysis module, the data analysis module is connected to the data storage module, the data storage module is connected to the data interaction module, the image recognition module is connected to the data visualization module, and the data interaction module is connected to the data visualization module.

[0060] Data pre-collection module: collect basic information, damage characteristics, omics information and phenotypic information of foreign and native pests and diseases through public resources or self-sampled samples;

[0061] The units collect basic information, damage characteristic information, omics information and phenotypic information of foreign and native pests and diseases through public resources or self-sampled samples, and construct original data sets and scanned image sets of foreign and native pests and diseases, and send them to the data integration module;

[0062] The data integration module integrates data from different experiments or studies, performs preprocessing, and sends it to the data analysis module;

[0063] Data analysis module: In one design, when constructing the three-dimensional internal and external models of foreign and native pests and diseases, measurement planning and data collection are performed to obtain the projection images of foreign and native pests and diseases. The projection images of foreign and native pests and diseases are reconstructed into a data volume composed of slice image sequences, and then the posture correction and effective signal amplification of the foreign and native pests and diseases data volume are performed. Manually draw some target areas, perform deep learning, and evaluate the accuracy of deep learning segmentation through segmentation preview images and functions.

[0064] In one design, when building an image recognition algorithm for foreign and native pests and diseases, the image data of various species collected in the original data set are very different, such as the image data collection sources are diverse, the image size and original pixels are different, and the sample size is large. Therefore, the existing image data cannot meet the production of the network model training data set. In the analysis module, the collected image data needs to be compressed and unified, and the data is augmented. After data processing, the data set is divided into training set, validation set, and test set according to the proportion. And select macro average and micro average as model performance evaluation indicators, and record training accuracy, recall rate, F1-score and accuracy.

[0065] In one design, DensetNet121, MobileNetV3, ResNet101 and ShuffleNet models were selected for training experiments to construct evaluation indicators for multi-recognition object classification models. To ensure the reliability of the evaluation parameters, the model set the average precision, average recall rate and average F1-score as macro-average, and the overall accuracy rate as micro-average.

[0066] Model building module: In one design, according to the data body after posture correction and effective signal amplification, manual segmentation or deep learning segmentation is selected, and then the internal and external three-dimensional models of foreign and local pests and diseases are constructed. Image recognition function is constructed according to the MobileNetV3 algorithm.

[0067] In one design, a three-dimensional model is constructed using micro-computed tomography technology, in which an X-ray generator emits X-rays to penetrate the specimen, and a detector placed at the other end records the process, generating radiographs to obtain three-dimensional fine structures that are difficult to obtain using traditional morphological dissection and tissue sectioning methods.

[0068] Based on the above segmentation process, an automated process was developed for large-scale modeling of the internal and external structures of exotic and native pests and diseases.

[0069] Data storage module: used to store the pre-processed basic information, damage characteristics information, and phenotypic information of foreign and local pests and diseases, and send it to the image recognition unit;

[0070] Omics processing module: reads the sequence fragments transmitted by the data interaction module, uses DNA barcoding, Blast alignment and other functions to identify pests and diseases, and sends the identification results to the data interaction module.

[0071] Image recognition module: reads the photos passed by the data interaction module, calls the model for recognition, and sends it to the data interaction module;

[0072] Data interaction module: based on the access rules of basic information modules of alien and native pests and diseases, the damage characteristic information of alien and native pests and diseases, the access rules of phenotypic modules of alien and native pests and diseases, the access rules of omics data modules of alien and native pests and diseases, and the access rules of image recognition modules of alien and native pests and diseases, the interaction between data storage module, image recognition module and data visualization module is processed, and standardized information is returned as needed.

[0073] In one design, the construction of data access rules for image recognition of foreign and native pests and diseases, that is, designing a data table for image recognition of foreign and native pests and diseases, and constructing data access rules, specifically includes:

[0074] According to the generated species search needs, search in the species table, hazard level table, hazard characteristic table, biological characteristic table, epidemic prevention status, three-dimensional model, multispectral image, phenotypic data table, genome data table, DNA barcode table, sequence table, primer table, process into a standardized format, and return the species' basic biological information, biological images, biological characteristics, invasion characteristics, invasion hazard degree, and epidemic prevention status, three-dimensional model, multispectral image, phenotypic data and omics data objects.

[0075] Data visualization module: It is used to provide users with a search form for all functions of the interactive multidimensional database of exotic and native pests and diseases, visualize the analysis results through various charts, graphs and tools, and provide data mining functions.

[0076] In one design, it is used to provide users with interactive selection forms for foreign and native pests and diseases and basic biological information such as species name, foreign name, host, morphological description, and pictures; the hazard characteristic information module and the phenotypic information module are used to provide users with interactive selection forms for foreign and native pests and diseases and hazard information such as quarantine status, hazard characteristics, hazard level, origin, distribution location, transmission route, prevention and control methods, and entry quarantine requirements; the phenotypic information module is used to provide users with interactive selection forms for foreign and native pests and diseases and phenotypic information such as high-definition images, multispectral data, and three-dimensional internal and external models; the image recognition module is used to provide users with interactive recognition of foreign and native pests and diseases and display related information; the omics module is used to provide users with interactive genome data, DNA barcode data, primer sequences, and Blast comparison functions.

[0077] See also Figure 2 As shown, the present invention provides a method for constructing a multidimensional database for image recognition of foreign and native pests and diseases, which specifically includes the following steps:

[0078] S1. Collect and construct original datasets and scanned image sets of exotic and native pests and diseases from self-sampled samples, public resources, relevant news, laws and regulations, books, and papers;

[0079] S2. Analyze the collected original data sets and scanned image sets, and perform image preprocessing, data set division, model evaluation and construction improvement to build three-dimensional internal and external models and image recognition models.

[0080] S3. Design multidimensional data tables for foreign and local pests and diseases, and build data access rules to handle the interaction between front-end and back-end data;

[0081] S4. According to the data access rules, the basic information, damage characteristic information, phenotypic information, omics information and image recognition modules of alien and native pests and diseases are constructed respectively.

[0082] Example 1: Identification and prevention of foreign pests and diseases

[0083] Data collection: First, data on exotic pests and diseases, especially those related to China, were collected from the China Exotic and Native Pest Information System, NCBI website, relevant books, papers, news reports and survey reports. This includes basic information such as species name, foreign name, host, morphological description, pictures, etc.; descriptions of hazard characteristics such as quarantine status, hazard characteristics, hazard level, origin, distribution location, transmission route, prevention and control methods, entry quarantine requirements, etc.; as well as phenotypic modules such as high-definition images and multispectral data, including genomic data, DNA barcode data, and primer sequence omics data modules.

[0084] Data preprocessing and model building: The collected image data is compressed and unified, and the data is augmented, and then the data set is divided into training set, validation set, and test set in proportion. After that, a deep learning model is used to build an image recognition model to distinguish between foreign pests and diseases and local pests and diseases.

[0085] Data volume reconstruction and segmentation: The projection images of foreign and local pests and diseases are reconstructed into a data volume composed of a sequence of slice images. This data volume allows the pests and diseases to be observed from different perspectives and levels, and their characteristics to be better identified. Then posture correction and effective signal amplification are performed, and then an appropriate segmentation method is selected for image segmentation.

[0086] Example 2: Maintaining the ecological balance of local pests and diseases

[0087] Data collection: Collect data on China's native pests and diseases from China's agricultural departments, research institutions, university research centers, environmental monitoring stations, and related news reports and survey reports. Pay special attention to data related to native ecosystems and biodiversity.

[0088] Data preprocessing and model building: Use the data preprocessing methods mentioned above, including image size compression, data augmentation, etc. Then use deep learning or other machine learning algorithms to classify and identify native pests and foreign pests. In addition, unsupervised learning methods, such as cluster analysis, can be used to explore the different types of pests and diseases and their roles in the ecosystem.

[0089] Ecological model construction: Using the collected data, an ecological model is built to simulate the behavior of native pests and diseases in the ecosystem and their interactions with other organisms. This model can help understand the importance of native pests and diseases, as well as how to maintain ecological balance and prevent the invasion of foreign pests and diseases.

[0090] The technical solution of the present invention can be applied to a system for identifying foreign and native pests and diseases, which includes the following modules:

[0091] Data collection module: This module is responsible for collecting data from public resources and self-sampled samples. Image data of pests and diseases collected in the field or customs quarantine environment can be collected using hyperspectral cameras, multispectral cameras or ordinary cameras and microscopes.

[0092] Data preprocessing module: This module is responsible for denoising, enhancing, segmenting, aligning and other operations on the collected image data to improve the quality and consistency of the image.

[0093] Omics module: This module is responsible for receiving sequence fragments sent by front-end users, using DNA barcoding, Blast alignment and other functions to identify pests and diseases, and providing other omics information, such as genome sequences and gene annotation functions.

[0094] Data storage module: This module is responsible for storing and managing the database of foreign and local pests and diseases, including basic information of pests and diseases, damage characteristics information, phenotypic information, three-dimensional structure information and omics information data, etc. Cloud computing and Internet of Things technologies can be used to achieve remote access and update of data.

[0095] Image recognition module: This module is responsible for receiving photos uploaded by front-end users’ mobile phones or other devices, and using deep learning to classify and identify pests and diseases on the pre-processed image data.

[0096] Data interaction module: This module is responsible for outputting the results of image recognition to users. It can use visualization, voice, text and other methods to provide the names, information, and prevention and control measures of pests and diseases. It can use cloud computing and Internet of Things technologies to achieve remote transmission and sharing of results.

[0097] Table 1 Test results of insect recognition model

[0098]

[0099]

[0100] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for constructing a multidimensional database for image recognition of foreign and native pests and diseases, characterized in that: The following steps are involved: S1. Collect data from self-sampled samples, public resources, relevant news, laws and regulations, books, and papers to construct original datasets and scanned image sets of exotic and native pests and diseases; S2, analyze the collected original data sets and scanned image sets, perform image preprocessing, data set division, model evaluation and improvement construction of three-dimensional internal and external models and image recognition models; S3. Design multidimensional data tables for foreign and local pests and diseases, and build data access rules to handle the interaction between front-end and back-end data; S4. According to the data access rules, the basic information, damage characteristic information, phenotypic information, omics data and image recognition modules of alien and native pests and diseases are constructed respectively.

2. The method for constructing a multidimensional database for image recognition of foreign and native pests and diseases according to claim 1, characterized in that: The alien and native pests and diseases collected in S1 cover all alien and native pests and diseases within China. The specific information must also be comprehensive, covering the basic information modules of species name, foreign name, host, morphological description, and pictures; the hazard characteristic description module including quarantine status, hazard characteristics, hazard level, place of origin, distribution location, transmission route, prevention and control methods, and entry quarantine requirements; the phenotypic module including high-definition images and multi-spectral data; and the omics data module including species genome data, DNA barcodes, and primer sequences.

3. The images of foreign and native pests and diseases as claimed in claim 1 should be comprehensive, including pictures collected by drones, mobile phones, cameras, microscopes, CT, three-dimensional, and hyperspectral. The original data of the three-dimensional model is measured and planned and collected by the control instrument to obtain the corresponding projection map.

4. The method for constructing a multidimensional database for image recognition of foreign and native pests and diseases according to claim 1, characterized in that: The resources used to collect basic information on exotic and native pests and diseases in S1 include resources from all over the world, including the U.S. National Center for Biotechnology Information, the China Exotic and Native Pests and Diseases Information System website, books and papers on related exotic and native pests and diseases, and related news reports or investigation reports.

5. The method for constructing a multidimensional database for image recognition of foreign and native pests and diseases according to claim 1, characterized in that: In S2, the collected original data sets are analyzed. When constructing the three-dimensional model, the projection images of foreign and native pests and diseases are reconstructed into a data body composed of slice image sequences, posture correction and effective signal amplification are performed, and the segmentation method is selected. The constructed database of the internal and external structure models of foreign and native pests and diseases can provide a data basis for pest identification and serve as a training data set for the deep learning algorithm to improve the accuracy of pest identification. In S2, the collected original data set is analyzed. When building the image recognition model, the collected image data needs to be compressed and unified in size and data augmented; After data processing, the dataset is divided into training set, validation set, and test set in proportion, and the evaluation indicators of the multi-recognition object classification model are constructed. To ensure the reliability of the evaluation parameters, the model sets the average precision, average recall rate, and average F1-score as macro-average, and the overall accuracy rate as micro-average.

6. The method for constructing a multidimensional database for image recognition of foreign and native pests and diseases according to claim 1, characterized in that: When constructing the model in S2, micro-computed tomography technology is used to construct a three-dimensional model, that is, the X-ray generator emits X-rays to penetrate the specimen, and the detector placed at the other end records the process, generating radiographs to obtain three-dimensional fine structures that are difficult to obtain using traditional morphological dissection and tissue sectioning methods.

7. The method for constructing a multidimensional database for image recognition of foreign and native pests and diseases according to claim 1, characterized in that: When building the model in S2, DensetNet121, MobileNetV3, ResNet101 and ShuffleNet models were selected for training experiments. After comprehensive comparison, the selected model MobileNetV3 was improved.

8. The method for constructing a multidimensional database for image recognition of foreign and native pests and diseases according to claim 1, characterized in that: The construction access rules in S3 include the construction access rules of the basic information module of alien and native pests and diseases, the access rules of the damage characteristics module of alien and native pests and diseases, the access rules of the phenotypic module of alien and native pests and diseases, the access rules of the omics data module of alien and native pests and diseases, and the access rules of the image recognition module of alien and native pests and diseases.

9. The method for constructing a multidimensional database for image recognition of foreign and native pests and diseases according to claim 1, characterized in that: The basic information module in S4 is used to provide users with interactive selection forms for exotic and native pests and diseases and displays of basic biological information such as species name, foreign name, host, morphological description, and pictures; The phenotypic information module is used to provide users with interactive selection forms for foreign and native pests and diseases and display of hazard information such as quarantine status, hazard characteristics, hazard level, place of origin, distribution location, transmission route, prevention and control methods, and entry quarantine requirements; the phenotypic information module is used to provide users with interactive selection forms for foreign and native pests and diseases and display of phenotypic information such as high-definition images, multispectral data, and three-dimensional internal and external models; the image recognition module is used to provide users with interactive identification of foreign and native pests and diseases and display related information; the omics data module is used to provide users with omics information to support interactive molecular detection of foreign and native pests and diseases, such as genome data, DNA barcode data, primer sequences, and Blast comparison functions.

10. A system for constructing a multidimensional database for image recognition of foreign and native pests and diseases, using the method for constructing a multidimensional database for image recognition of foreign and native pests and diseases as claimed in any one of claims 1 to 9, characterized in that: include: Data pre-collection module: including basic information units of foreign and native pests and diseases, damage characteristics information units, omics information units and phenotypic units; The units collect basic information, damage characteristics, omics information and phenotypic information of foreign and native pests and diseases through public resources or self-sampled samples, and construct original data sets and scanned image sets of foreign and native pests and diseases; Data integration module: integrates data from different experiments or studies, preprocesses them, and sends them to the data analysis module; Data analysis module: preprocessing for 3D model construction, reconstructing the projection images of foreign and local pests and diseases into a data volume composed of slice image sequences, performing posture correction and effective signal amplification, and selecting the segmentation method; preprocessing for image recognition model construction, and performing image size compression and data augmentation processing on the collected image data; Construct training data set, validation data set and test data set; And select macro-average and micro-average as the model performance evaluation indicators, and record the training accuracy, recall rate, F1-score and accuracy; Model building module: build three-dimensional models, image recognition models, and perform model selection and verification; Data storage module: used to store basic information, damage characteristics information, phenotypic information and information of image recognition module of foreign and local pests and diseases and send them to image recognition unit; Omics processing module: reads the sequence fragments transmitted by the data interaction module, uses DNA barcodes, Blast comparison and other functions to identify pests and diseases, and sends the identification results to the data interaction module; Image recognition module: reads the photos transmitted by the data interaction module, calls the model building module for recognition, calls the data storage module to read the corresponding data and sends it to the data interaction module; Data interaction module: based on the access rules of basic information modules of foreign and native pests and diseases, the access rules of damage characteristics of foreign and native pests and diseases, the access rules of phenotypic modules of foreign and native pests and diseases, the access rules of omics data modules of foreign and native pests and diseases, and the access rules of image recognition modules of foreign and native pests and diseases, the interaction between data storage modules, image recognition modules and data visualization modules is processed, and standardized information is returned as needed; Data visualization module: It is used to provide users with a search form for all functions of the interactive multidimensional database of exotic and native pests and diseases, visualize the analysis results through various charts, graphs and tools, and provide data mining functions.

Images