A method for intelligent identification and precise prevention and control of citrus diseases

Abstract

The present invention relates to a method for intelligent identification and precise prevention and control of citrus diseases, and belongs to the technical field of citrus diseases. The present invention identifies remote sensing image data of citrus trees in the current target area based on a damage identification model, obtains identification results, and obtains probability values ​​of various citrus disease types according to the identification results and the knowledge graph, and finally obtains location information of damaged citrus trees according to the identification results, and issues early warnings based on the probability values ​​of various citrus disease types and the location information of damaged citrus trees, and performs precise prevention and control according to the generated early warning information. The present invention identifies the damage to citrus trees, thereby evaluating the possibility of disease according to the damage and environmental data, and then issues early warnings and early prevention and control according to the possibility of disease, perceives the possibility of disease in advance, and reduces the economic losses of citrus caused by disease.

Description

Technical Field

[0001] The present invention relates to the technical field of citrus disease prevention and control, and in particular to a method for intelligent identification and precise prevention and control of citrus diseases. Background Art

[0002] Citrus has a long history of cultivation and is one of the important economic fruits. It grows better in relatively humid places with suitable temperatures. However, for a long time, citrus diseases have seriously threatened the growth of citrus, affecting its yield and quality, and have had a huge impact on the citrus industry. Therefore, how to strengthen the control of citrus disease prevention and control technology, effectively prevent the occurrence of diseases, and thus achieve high yield and high quality of citrus is an issue that people need to focus on at present. Citrus diseases mainly harm leaves, new shoots and fruits, and are especially susceptible to infecting young and tender tissues. When the leaves are infected, waxy yellow oily spots appear at first, which gradually expand to form grayish white to dark brown conical scabs. Later, the spots become lignified and protruding, the back of the leaves are prominent, and the leaf surface is sunken. When the new shoots are infected, the symptoms are similar to those of the leaves. When the young fruit is infected, the fruit surface is densely covered with tea-brown spots, which then expand to form yellow-brown conical, lignified tumor-like protrusions on the peel. Among them, the production of citrus diseases is mainly that the mycelium overwinters in the diseased tissue. In the next spring, when it is rainy and humid, the pathogen begins to move and produce conidia, which are spread by wind and rain or insects, infecting new shoots and tender leaves, producing new conidia for re-infection, harming new shoots and young fruits, and spring shoots and young trees are more seriously affected. The more rain there is during the citrus disease-susceptible period, the more serious the disease. In the prior art, citrus diseases are often treated by drugs or biological control methods after they occur, but this has already caused certain economic losses. How to reduce the occurrence of citrus diseases is one of the major problems in the planting process. Summary of the invention

[0003] The present invention overcomes the deficiencies of the prior art and provides a method for intelligent identification and precise prevention and control of citrus diseases.

[0004] To achieve the above object, the technical solution adopted by the present invention is:

[0005] The first aspect of the present invention provides a method for intelligent identification and precise prevention and control of citrus diseases, comprising the following steps:

[0006] Obtain the number of occurrences of each citrus disease type within a preset time, calculate the probability value of each citrus disease type occurring under the citrus tree damage factor based on the number of occurrences of each citrus disease type within the preset time, and construct a knowledge graph;

[0007] Acquire remote sensing image data of citrus fruit trees through unmanned aerial vehicle remote sensing technology, and construct a damage recognition model based on the remote sensing image data of citrus fruit trees;

[0008] Based on the damage recognition model, remote sensing image data of citrus trees in the current target area are recognized to obtain recognition results, and probability values ​​of the occurrence of various citrus disease types are obtained according to the recognition results and the knowledge graph;

[0009] The location information of the damaged citrus trees is obtained based on the recognition results, an early warning is issued based on the probability values ​​of the occurrence of each type of citrus disease and the location information of the damaged citrus trees, and precise prevention and control is carried out based on the generated early warning information.

[0010] Furthermore, in this method, the number of occurrences of each citrus disease type within a preset time is obtained, and the probability value of each citrus disease type occurring under the citrus tree damage factor is calculated based on the number of occurrences of each citrus disease type within the preset time, and a knowledge graph is constructed, specifically:

[0011] Obtain the number of occurrences of each citrus disease type within a preset time, and obtain the number of times citrus trees are damaged when each citrus disease type occurs, and calculate the probability value of each citrus disease type occurring under the citrus tree damage factor based on the number of occurrences of each citrus disease type within the preset time and the number of times citrus trees are damaged;

[0012] Counting the probability values ​​of each citrus disease type under the citrus tree damage factor, constructing a topological structure graph, taking the citrus tree damage factor as the first graph node of the topological structure graph, and taking the citrus tree damage factor as the second graph node of the topological structure graph;

[0013] Connecting the first graph node and the second graph node through a directed connection relationship, and obtaining a final topological structure graph after connecting the first graph node and the second graph node;

[0014] Based on the final topological structure diagram, a corresponding adjacency matrix is ​​obtained, a knowledge graph is constructed, the corresponding adjacency matrix is ​​input into a storage space in the knowledge graph for storage, and a query address is configured for each storage space.

[0015] Furthermore, in this method, remote sensing image data of citrus trees are obtained by using UAV remote sensing technology, and a damage recognition model is constructed based on the remote sensing image data of citrus trees, which specifically includes:

[0016] Acquire remote sensing image data of citrus trees by using UAV remote sensing technology, extract remote sensing image data of damaged citrus trees from the remote sensing image data of the citrus trees, filter and denoise the remote sensing image data of damaged citrus trees, and acquire pre-processed remote sensing image data;

[0017] Introducing a feature pyramid network, inputting the preprocessed remote sensing image data into the feature pyramid network, obtaining feature image data of citrus trees when damaged, and constructing a damage recognition model based on a deep neural network;

[0018] Constructing a plurality of damage type training sets based on the characteristic image data of the citrus tree when it is damaged, and inputting the damage type training sets into the damage recognition model for training;

[0019] A prediction accuracy threshold is set. When the prediction accuracy of the damage identification model is greater than the prediction accuracy threshold, the model parameters of the damage identification model are saved, the damage identification model is output, and the training is completed.

[0020] Furthermore, in this method, the remote sensing image data of citrus trees in the current target area is identified based on the damage identification model to obtain an identification result, and the probability value of each citrus disease type is obtained according to the identification result and the knowledge graph, specifically:

[0021] Acquire remote sensing image data of citrus fruit trees in the current target area, and input the remote sensing image data of citrus fruit trees in the current target area into the damage recognition model for recognition, to obtain recognition results;

[0022] Determine whether there is damage in the recognition result. When there is damage in the recognition result, obtain the probability value of each citrus disease type through the knowledge graph, and output the probability value of each citrus disease type.

[0023] Furthermore, in the present method, the location information of the damaged citrus fruit trees is obtained according to the recognition results, and an early warning is given based on the probability values ​​of the occurrence of each type of citrus disease and the location information of the damaged citrus fruit trees, specifically including:

[0024] When there is damage in the identification result, the location information of the damaged citrus fruit tree is obtained, and the probability threshold of the occurrence of citrus disease types is set, and it is determined whether the probability value of each citrus disease type is greater than the probability threshold of the occurrence of the citrus disease type;

[0025] When the probability value of the occurrence of a citrus disease type is greater than the probability threshold of the occurrence of the citrus disease type, the real-time environmental information corresponding to the possible occurrence of the citrus disease type and the location information of the damaged citrus fruit tree are obtained;

[0026] Obtain the colony type of the possible disease type according to the corresponding possible citrus disease type, and obtain the survival rate data of the colony type of the possible disease type under the real-time environmental information through big data;

[0027] When the survival rate data of the colony type that may cause the disease type under the real-time environmental information is greater than the preset survival rate threshold, warning information is generated according to the location information of the corresponding citrus disease type and the damaged citrus trees, and a warning is issued.

[0028] Furthermore, in this method, precise prevention and control is carried out according to the generated early warning information, specifically including:

[0029] Obtain the prevention and control measures for each type of citrus disease through big data, and obtain the location information of the corresponding citrus disease types and damaged citrus trees based on the generated early warning information;

[0030] Based on the prevention and control measures for each type of citrus disease and the corresponding types of citrus diseases that may occur, the prevention and control measures for the current type of citrus disease are obtained, and the prevention and control points are generated based on the location information of the damaged citrus trees;

[0031] Based on the control points, control marks are generated, and control marks are arranged for control according to the control measures for the current citrus disease type.

[0032] A second aspect of the present invention provides a citrus disease intelligent identification and precise prevention and control system, the system comprising a memory and a processor, the memory comprising a citrus disease intelligent identification and precise prevention and control method program, the citrus disease intelligent identification and precise prevention and control method program when executed by the processor, implements the following steps:

[0033] Obtain the number of occurrences of each citrus disease type within a preset time, calculate the probability value of each citrus disease type occurring under the citrus tree damage factor based on the number of occurrences of each citrus disease type within the preset time, and construct a knowledge graph;

[0034] Acquire remote sensing image data of citrus fruit trees through unmanned aerial vehicle remote sensing technology, and construct a damage recognition model based on the remote sensing image data of citrus fruit trees;

[0035] Based on the damage recognition model, remote sensing image data of citrus trees in the current target area are recognized to obtain recognition results, and probability values ​​of the occurrence of various citrus disease types are obtained according to the recognition results and the knowledge graph;

[0036] The location information of the damaged citrus trees is obtained based on the recognition results, an early warning is issued based on the probability values ​​of the occurrence of each type of citrus disease and the location information of the damaged citrus trees, and precise prevention and control is carried out based on the generated early warning information.

[0037] The third aspect of the present invention provides a computer-readable storage medium, which includes a program for the method for intelligent identification and precise prevention and control of citrus diseases. When the program for the method for intelligent identification and precise prevention and control of citrus diseases is executed by a processor, the steps of any one of the methods for intelligent identification and precise prevention and control of citrus diseases are implemented.

[0038] The present invention solves the defects existing in the background technology and has the following beneficial effects:

[0039] The present invention obtains the number of occurrences of each citrus disease type within a preset time, calculates the probability value of each citrus disease type under the citrus tree damage factor based on the number of occurrences of each citrus disease type within the preset time, and constructs a knowledge graph, and then obtains remote sensing image data of citrus trees through unmanned aerial vehicle remote sensing technology, and constructs a damage recognition model based on the remote sensing image data of citrus trees, thereby identifying the remote sensing image data of citrus trees in the current target area based on the damage recognition model, obtaining the recognition result, and obtaining the probability value of each citrus disease type according to the recognition result and the knowledge graph, and finally obtaining the location information of the damaged citrus trees according to the recognition result, and giving an early warning based on the probability value of each citrus disease type and the location information of the damaged citrus trees, and performing precise prevention and control according to the generated early warning information. The present invention identifies the damage of citrus trees, thereby evaluating the possibility of disease according to the damage and environmental data, and then giving an early warning and early prevention and control according to the possibility of disease, and perceives the possibility of disease in advance, thereby reducing the economic losses of citrus caused by disease. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, drawings of other embodiments can be obtained based on these drawings without paying creative work.

[0041] Figure 1 An overall flow chart of a method for intelligent identification and precise prevention and control of citrus diseases is shown;

[0042] Figure 2 A partial method flow chart of a method for intelligent identification and precise prevention and control of citrus diseases is shown;

[0043] Figure 3 The system block diagram of the citrus disease intelligent identification and precise prevention and control system is shown. DETAILED DESCRIPTION

[0044] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0045] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0046] like Figure 1 As shown, the first aspect of the present invention provides a method for intelligent identification and precise prevention and control of citrus diseases, comprising the following steps:

[0047] S102: Obtain the number of occurrences of each citrus disease type within a preset time, calculate the probability value of each citrus disease type occurring under the citrus tree damage factor based on the number of occurrences of each citrus disease type within the preset time, and construct a knowledge graph;

[0048] like Figure 2 As shown, further, in step S102 of the method, the following steps are specifically included:

[0049] S202: Obtain the number of occurrences of each citrus disease type within a preset time, and obtain the number of times citrus trees are damaged when each citrus disease type occurs, and calculate the probability value of each citrus disease type occurring under the citrus tree damage factor based on the number of occurrences of each citrus disease type within the preset time and the number of times citrus trees are damaged;

[0050] S204: Counting the probability values ​​of each citrus disease type under the citrus tree damage factor, constructing a topological structure graph, taking the citrus tree damage factor as the first graph node of the topological structure graph, and taking the citrus tree damage factor as the second graph node of the topological structure graph;

[0051] S206: Connect the first graph node and the second graph node through a directed connection relationship, and obtain a final topological structure graph after connecting the first graph node and the second graph node;

[0052] S208: Based on the final topological structure diagram, obtain the corresponding adjacency matrix, construct a knowledge graph, input the corresponding adjacency matrix into the storage space in the knowledge graph for storage, and configure a query address for each storage space.

[0053] It should be noted that the main diseases of citrus actually include citrus black spot, citrus canker, citrus anthracnose, citrus yellow spot, citrus gray mold and other types of diseases. In fact, many diseases are only attacked by pathogens after citrus is damaged (such as bark damage, frostbite, burns, cut wounds, insect wounds, etc.). For example, citrus black spot disease, the pathogen is weakly parasitic, and spore germination and invasion require water droplets. It can only invade from the wounds of the host (frostbite, burns, cut wounds, insect wounds, etc.). On the surface of new tissues such as young leaves and young fruits without wounds and strong vitality, the infection of the pathogen is blocked in the epidermis of the host, forming many small black spots of colloid. When the host has a large number of wounds, there is a lot of rain and the temperature is suitable, symptoms of gum flow, dryness and fruit stalk rot will occur, and finally cause a large area of ​​disease. This method can be used to calculate the probability of each type of citrus disease occurring under citrus tree damage factors, thereby constructing a knowledge graph.

[0054] S104: acquiring remote sensing image data of citrus trees by using UAV remote sensing technology, and constructing a damage recognition model based on the remote sensing image data of citrus trees;

[0055] Furthermore, in step S104 of the method, the following steps are specifically included:

[0056] Acquire remote sensing image data of citrus trees through UAV remote sensing technology, extract remote sensing image data of damaged citrus trees from the remote sensing image data of citrus trees, filter and denoise the remote sensing image data of damaged citrus trees, and acquire pre-processed remote sensing image data;

[0057] The feature pyramid network was introduced, and the preprocessed remote sensing image data was input into the feature pyramid network to obtain the feature image data of citrus trees when they were damaged, and a damage recognition model was constructed based on a deep neural network.

[0058] Based on the characteristic image data of citrus trees when damaged, several damage type training sets are constructed, and the damage type training sets are input into the damage recognition model for training;

[0059] A prediction accuracy threshold is set. When the prediction accuracy of the damage identification model is greater than the prediction accuracy threshold, the model parameters of the damage identification model are saved, the damage identification model is output, and the training is completed.

[0060] It should be noted that the characteristic image data of citrus trees when damaged include frostbite, burn, cut injury, insect injury and other data, and the damage recognition model can be constructed through this method.

[0061] S106: Identify the remote sensing image data of citrus trees in the current target area based on the damage identification model, obtain the identification result, and obtain the probability value of each citrus disease type according to the identification result and the knowledge graph;

[0062] Furthermore, in step S106 of this method, specifically:

[0063] Acquire remote sensing image data of citrus fruit trees in the current target area, and input the remote sensing image data of citrus fruit trees in the current target area into a damage recognition model for recognition, and obtain recognition results;

[0064] Determine whether there is damage in the recognition result. When there is damage in the recognition result, obtain the probability value of each citrus disease type through the knowledge graph, and output the probability value of each citrus disease type.

[0065] S108: Obtain the location information of the damaged citrus trees based on the identification results, issue an early warning based on the probability values ​​of each type of citrus disease and the location information of the damaged citrus trees, and perform precise prevention and control based on the generated early warning information.

[0066] Furthermore, in this method, the location information of the damaged citrus fruit trees is obtained according to the recognition results, and an early warning is given based on the probability values ​​of the occurrence of each citrus disease type and the location information of the damaged citrus fruit trees, specifically including:

[0067] When there is damage in the identification result, the location information of the damaged citrus fruit tree is obtained, and the probability threshold of the occurrence of citrus disease type is set, and it is determined whether the probability value of each citrus disease type is greater than the probability threshold of the occurrence of citrus disease type;

[0068] When the probability value of the occurrence of a citrus disease type is greater than the probability threshold of the occurrence of a citrus disease type, the real-time environmental information corresponding to the possible occurrence of the citrus disease type and the location information of the damaged citrus fruit tree are obtained;

[0069] According to the corresponding possible citrus disease types, the colony types of possible disease types are obtained, and the survival rate data of the colony types of possible disease types under real-time environmental information are obtained through big data;

[0070] When the survival rate data of the colony type that may cause the disease type under the real-time environmental information is greater than the preset survival rate threshold, warning information is generated and issued based on the location information of the corresponding citrus disease type and the damaged citrus trees.

[0071] It should be noted that, for example, citrus canker, the pathogenic colonies of the disease mainly invade through stomata, water pores, lenticels and wounds. The disease occurs in the temperature range of 20-35°C, and the optimum temperature is 25-30°C. High temperature and high humidity weather are necessary conditions for the epidemic. Therefore, the survival rate data of the colony type that may cause the disease type under real-time environmental information is obtained through big data. When the survival rate data of the colony type that may cause the disease type under real-time environmental information is greater than the preset survival rate threshold, it means that the probability of the disease is very high. Real-time environmental data includes temperature, humidity, etc.

[0072] Furthermore, in this method, precise prevention and control is carried out according to the generated early warning information, specifically including:

[0073] Obtain the prevention and control measures for each type of citrus disease through big data, and obtain the location information of the corresponding citrus disease types and damaged citrus trees based on the generated early warning information;

[0074] Based on the prevention and control measures of each citrus disease type and the corresponding possible citrus disease types, the prevention and control measures of the current citrus disease type are obtained, and the prevention and control points are generated based on the location information of the damaged citrus trees;

[0075] Generate prevention and control marks based on the prevention and control points, and arrange the prevention and control marks according to the prevention and control measures for the current citrus disease type.

[0076] It should be noted that prevention and control measures include biological control, drug control, physical control and other methods.

[0077] In summary, it is worth mentioning that the present invention identifies the damage to citrus trees, thereby evaluating the possibility of disease occurrence based on the damage conditions and environmental data, and then providing early warning and early prevention and control based on the possibility of disease occurrence, thereby perceiving the possibility of disease occurrence in advance and reducing the economic losses of citrus caused by disease.

[0078] In addition, obtaining remote sensing image data of citrus trees through UAV remote sensing technology can include the following steps:

[0079] Constructing a UAV remote sensing monitoring network, obtaining the position information of the extreme position points in the area of ​​the target area to be collected, and initializing the position information of the communication equipment in the UAV remote sensing monitoring network;

[0080] According to the location information of the communication equipment in the UAV remote sensing monitoring network and the location information of the extreme location points in the area of ​​the target area to be collected, a plurality of communication Euclidean distance values ​​are calculated, and a genetic algorithm is introduced, and a genetic algebra is set based on the genetic algorithm;

[0081] Determine whether the communication Euclidean distance values ​​are all not greater than a preset communication Euclidean distance value, and when the communication Euclidean distance values ​​are all not greater than the preset communication Euclidean distance value, output the location information of the communication device in the UAV remote sensing monitoring network, and control the UAV to obtain remote sensing image data of the citrus tree through the UAV remote sensing monitoring network;

[0082] When at least one of the communication Euclidean distance values ​​is greater than a preset communication Euclidean distance value, genetic iteration is performed based on the genetic algebra to readjust the location information of the communication equipment in the UAV remote sensing monitoring network until all the communication Euclidean distance values ​​are no greater than the preset communication Euclidean distance value.

[0083] It should be noted that, in fact, the UAV remote sensing monitoring network is a network for controlling the flight of UAVs, and in the predetermined area, the installation location of the communication equipment will affect the control delay of the UAV remote sensing monitoring network. When the Euclidean distance between the location information of the communication equipment in the UAV remote sensing monitoring network and the location information of the extreme position point in the area of ​​the target area to be collected for data is larger, the increase in distance will lead to an increase in the time for information transmission and an increase in control delay. This method can improve the rationality of UAVs in collecting citrus disease images.

[0084] In addition, constructing several damage type training sets based on the characteristic image data of citrus trees when damaged may also include the following steps:

[0085] Building several training sets of damage types based on the characteristic image data of the citrus tree when it is damaged, randomly selecting training items in the training set of each damage type, and extracting damage features in each training item;

[0086] Determine whether there are at least two damage features in the damage features in each training item. If there is only one damage feature in the damage features in each training item, use the corresponding sample data as a training item in the training set;

[0087] If there are at least two damage features in the damage features in each training item, the Mahalanobis distance value between each damage feature is calculated, and when the Mahalanobis distance value is not greater than a preset Mahalanobis distance threshold, the corresponding sample data is used as a training item in the training set;

[0088] When the Mahalanobis distance value is greater than the preset Mahalanobis distance threshold, the damage features in the training item are segmented and one of the features is eliminated. When the Mahalanobis distance value is not greater than the preset Mahalanobis distance threshold, the training item after the damage feature is eliminated is output, and the training item after the damage feature is eliminated is used as the training item in the training set.

[0089] It should be noted that, in fact, the collected image data may be images with multiple damage conditions, such as images with damage features such as frostbite, burns, cuts, and insect damage. At this time, when the model is trained, it is divided into multiple frostbite, burns, cuts, and insect damage training sets. If irrelevant features are not removed, it will cause interference in multi-scale data. This method can suppress the multi-scale interference of the model during training and improve the prediction accuracy of the damage recognition model.

[0090] like Figure 3 As shown, the second aspect of the present invention provides a citrus disease intelligent identification and precise prevention and control system 4, the system 4 includes a memory 41 and a processor 42, the memory 41 includes a citrus disease intelligent identification and precise prevention and control method program, and when the citrus disease intelligent identification and precise prevention and control method program is executed by the processor 42, the following steps are implemented:

[0091] Obtain the number of occurrences of each citrus disease type within a preset time, calculate the probability value of each citrus disease type under the citrus tree damage factor based on the number of occurrences of each citrus disease type within the preset time, and construct a knowledge graph;

[0092] The remote sensing image data of citrus trees is obtained through UAV remote sensing technology, and a damage recognition model is constructed based on the remote sensing image data of citrus trees;

[0093] Based on the damage recognition model, the remote sensing image data of citrus trees in the current target area is identified to obtain the recognition results, and the probability values ​​of each type of citrus disease are obtained based on the recognition results and the knowledge graph;

[0094] The location information of the damaged citrus trees is obtained based on the identification results, and an early warning is issued based on the probability values ​​of each type of citrus disease and the location information of the damaged citrus trees. Precise prevention and control are carried out based on the generated early warning information.

[0095] Furthermore, in this system, the number of occurrences of each citrus disease type within a preset time is obtained, and the probability value of each citrus disease type under the citrus tree damage factor is calculated based on the number of occurrences of each citrus disease type within the preset time, and a knowledge graph is constructed, specifically:

[0096] Obtain the number of occurrences of each citrus disease type within a preset time, and obtain the number of times citrus trees are damaged when each citrus disease type occurs, and calculate the probability value of each citrus disease type occurring under the citrus tree damage factor based on the number of occurrences of each citrus disease type within the preset time and the number of times citrus trees are damaged;

[0097] Statistically calculate the probability value of each citrus disease type under the citrus tree damage factor, construct a topological structure diagram, and use the citrus tree damage factor as the first graph node of the topological structure diagram, and use the citrus tree damage factor as the second graph node of the topological structure diagram;

[0098] The first graph node and the second graph node are connected through a directed connection relationship, and a final topological structure graph is obtained after the first graph node and the second graph node are connected;

[0099] Based on the final topological structure diagram, the corresponding adjacency matrix is ​​obtained, a knowledge graph is constructed, the corresponding adjacency matrix is ​​input into the storage space in the knowledge graph for storage, and a query address is configured for each storage space.

[0100] Furthermore, in this system, remote sensing image data of citrus trees are obtained through UAV remote sensing technology, and a damage recognition model is constructed based on the remote sensing image data of citrus trees, which specifically includes:

[0101] Acquire remote sensing image data of citrus trees through UAV remote sensing technology, extract remote sensing image data of damaged citrus trees from the remote sensing image data of citrus trees, filter and denoise the remote sensing image data of damaged citrus trees, and acquire pre-processed remote sensing image data;

[0102] The feature pyramid network was introduced, and the preprocessed remote sensing image data was input into the feature pyramid network to obtain the feature image data of citrus trees when they were damaged, and a damage recognition model was constructed based on a deep neural network.

[0103] Based on the characteristic image data of citrus trees when damaged, several damage type training sets are constructed, and the damage type training sets are input into the damage recognition model for training;

[0104] A prediction accuracy threshold is set. When the prediction accuracy of the damage identification model is greater than the prediction accuracy threshold, the model parameters of the damage identification model are saved, the damage identification model is output, and the training is completed.

[0105] Furthermore, in this system, the remote sensing image data of citrus trees in the current target area is identified based on the damage recognition model to obtain the recognition results, and the probability values ​​of each type of citrus disease are obtained based on the recognition results and the knowledge graph, specifically:

[0106] Acquire remote sensing image data of citrus fruit trees in the current target area, and input the remote sensing image data of citrus fruit trees in the current target area into a damage recognition model for recognition, and obtain recognition results;

[0107] Determine whether there is damage in the recognition result. When there is damage in the recognition result, obtain the probability value of each citrus disease type through the knowledge graph, and output the probability value of each citrus disease type.

[0108] Furthermore, in this system, the location information of the damaged citrus trees is obtained according to the recognition results, and an early warning is issued based on the probability values ​​of the occurrence of each type of citrus disease and the location information of the damaged citrus trees, specifically including:

[0109] When there is damage in the identification result, the location information of the damaged citrus fruit tree is obtained, and the probability threshold of the occurrence of citrus disease type is set, and it is determined whether the probability value of each citrus disease type is greater than the probability threshold of the occurrence of citrus disease type;

[0110] When the probability value of the occurrence of a citrus disease type is greater than the probability threshold of the occurrence of a citrus disease type, the real-time environmental information corresponding to the possible occurrence of the citrus disease type and the location information of the damaged citrus fruit tree are obtained;

[0111] According to the corresponding possible citrus disease types, the colony types of possible disease types are obtained, and the survival rate data of the colony types of possible disease types under real-time environmental information are obtained through big data;

[0112] When the survival rate data of the colony type that may cause the disease type under the real-time environmental information is greater than the preset survival rate threshold, warning information is generated and issued based on the location information of the corresponding citrus disease type and the damaged citrus trees.

[0113] Furthermore, in this system, precise prevention and control is carried out according to the generated early warning information, including:

[0114] Obtain the prevention and control measures for each type of citrus disease through big data, and obtain the location information of the corresponding citrus disease types and damaged citrus trees based on the generated early warning information;

[0115] Based on the prevention and control measures of each citrus disease type and the corresponding possible citrus disease types, the prevention and control measures of the current citrus disease type are obtained, and the prevention and control points are generated based on the location information of the damaged citrus trees;

[0116] Generate prevention and control marks based on the prevention and control points, and arrange the prevention and control marks according to the prevention and control measures for the current citrus disease type.

[0117] The third aspect of the present invention provides a computer-readable storage medium, which includes a program for the method for intelligent identification and precise prevention and control of citrus diseases. When the program for the method for intelligent identification and precise prevention and control of citrus diseases is executed by a processor, any step of the method for intelligent identification and precise prevention and control of citrus diseases is implemented.

[0118] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0119] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed on multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0120] In addition, all functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately used as a unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0121] A person of ordinary skill in the art can understand that: all or part of the steps of implementing the above method embodiment can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiment; and the aforementioned storage medium includes: a mobile storage device, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and other media that can store program codes.

[0122] Alternatively, if the above-mentioned integrated unit of the present invention is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention can be essentially or partly reflected in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the methods of each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROM, RAM, magnetic disks or optical disks.

[0123] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art who is familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for intelligent identification and precise prevention and control of citrus diseases, characterized in that: The following steps are involved: Obtain the number of occurrences of each citrus disease type within a preset time, calculate the probability value of each citrus disease type occurring under the citrus tree damage factor based on the number of occurrences of each citrus disease type within the preset time, and construct a knowledge graph; Acquire remote sensing image data of citrus fruit trees through unmanned aerial vehicle remote sensing technology, and construct a damage recognition model based on the remote sensing image data of citrus fruit trees; Based on the damage recognition model, remote sensing image data of citrus trees in the current target area are recognized to obtain recognition results, and probability values ​​of the occurrence of various citrus disease types are obtained according to the recognition results and the knowledge graph; Acquire the location information of the damaged citrus fruit trees according to the recognition results, issue an early warning based on the probability values ​​of the occurrence of each type of citrus disease and the location information of the damaged citrus fruit trees, and perform precise prevention and control according to the generated early warning information; Acquiring remote sensing image data of citrus trees through UAV remote sensing technology, and constructing a damage recognition model based on the remote sensing image data of citrus trees, specifically including: Acquire remote sensing image data of citrus trees by using UAV remote sensing technology, extract remote sensing image data of damaged citrus trees from the remote sensing image data of the citrus trees, filter and denoise the remote sensing image data of damaged citrus trees, and acquire pre-processed remote sensing image data; Introducing a feature pyramid network, inputting the preprocessed remote sensing image data into the feature pyramid network, obtaining feature image data of citrus trees when damaged, and constructing a damage recognition model based on a deep neural network; Constructing a plurality of damage type training sets based on the characteristic image data of the citrus tree when it is damaged, and inputting the damage type training sets into the damage recognition model for training; Setting a prediction accuracy threshold, when the prediction accuracy of the damage identification model is greater than the prediction accuracy threshold, saving the model parameters of the damage identification model, outputting the damage identification model, and completing the training; Acquiring remote sensing image data of citrus trees through UAV remote sensing technology can include the following steps: Constructing a UAV remote sensing monitoring network, obtaining the position information of the extreme position points in the area of ​​the target area to be collected, and initializing the position information of the communication equipment in the UAV remote sensing monitoring network; According to the location information of the communication equipment in the UAV remote sensing monitoring network and the location information of the extreme location points in the area of ​​the target area to be collected, a plurality of communication Euclidean distance values ​​are calculated, and a genetic algorithm is introduced, and a genetic algebra is set based on the genetic algorithm; Determine whether the communication Euclidean distance values ​​are all not greater than a preset communication Euclidean distance value, and when the communication Euclidean distance values ​​are all not greater than the preset communication Euclidean distance value, output the location information of the communication device in the UAV remote sensing monitoring network, and control the UAV to obtain remote sensing image data of the citrus tree through the UAV remote sensing monitoring network; When at least one of the communication Euclidean distance values ​​is greater than a preset communication Euclidean distance value, genetic iteration is performed based on the genetic algebra to readjust the location information of the communication device in the UAV remote sensing monitoring network until all the communication Euclidean distance values ​​are no greater than the preset communication Euclidean distance value; Constructing several damage type training sets based on the characteristic image data of citrus trees when damaged may also include the following steps: Building several training sets of damage types based on the characteristic image data of the citrus tree when it is damaged, randomly selecting training items in the training set of each damage type, and extracting damage features in each training item; Determine whether there are at least two damage features in the damage features in each training item. If there is only one damage feature in the damage features in each training item, use the corresponding sample data as a training item in the training set; If there are at least two damage features in the damage features in each training item, the Mahalanobis distance value between each damage feature is calculated, and when the Mahalanobis distance value is not greater than a preset Mahalanobis distance threshold, the corresponding sample data is used as a training item in the training set; When the Mahalanobis distance value is greater than the preset Mahalanobis distance threshold, the damage features in the training item are segmented and one of the features is eliminated. When the Mahalanobis distance value is not greater than the preset Mahalanobis distance threshold, the training item after the damage feature is eliminated is output, and the training item after the damage feature is eliminated is used as the training item in the training set.

2. A method for intelligent identification and precise prevention and control of citrus diseases according to claim 1, characterized in that: Based on the damage recognition model, the remote sensing image data of citrus trees in the current target area is recognized to obtain a recognition result, and the probability value of each citrus disease type is obtained according to the recognition result and the knowledge graph, specifically: Acquire remote sensing image data of citrus fruit trees in the current target area, and input the remote sensing image data of citrus fruit trees in the current target area into the damage recognition model for recognition, to obtain recognition results; Determine whether there is damage in the recognition result. When there is damage in the recognition result, obtain the probability value of each citrus disease type through the knowledge graph, and output the probability value of each citrus disease type.

3. The method for intelligent identification and precise prevention and control of citrus diseases according to claim 1, characterized in that: The method comprises: obtaining the location information of the damaged citrus fruit trees according to the identification results, and issuing an early warning based on the probability values ​​of the occurrence of each citrus disease type and the location information of the damaged citrus fruit trees. Specifically, the method comprises: When there is damage in the identification result, the location information of the damaged citrus fruit tree is obtained, and the probability threshold of the occurrence of citrus disease types is set, and it is determined whether the probability value of each citrus disease type is greater than the probability threshold of the occurrence of the citrus disease type; When the probability value of the occurrence of a citrus disease type is greater than the probability threshold of the occurrence of the citrus disease type, the real-time environmental information corresponding to the possible occurrence of the citrus disease type and the location information of the damaged citrus fruit tree are obtained; Obtain the colony type of the possible disease type according to the corresponding possible citrus disease type, and obtain the survival rate data of the colony type of the possible disease type under the real-time environmental information through big data; When the survival rate data of the colony type that may cause the disease type under the real-time environmental information is greater than the preset survival rate threshold, warning information is generated according to the location information of the corresponding citrus disease type and the damaged citrus trees, and a warning is issued.

4. The method for intelligent identification and precise prevention and control of citrus diseases according to claim 1, characterized in that: Carry out precise prevention and control based on the early warning information generated, including: Obtain the prevention and control measures for each type of citrus disease through big data, and obtain the location information of the corresponding citrus disease types and damaged citrus trees based on the generated early warning information; Based on the prevention and control measures for each type of citrus disease and the corresponding types of citrus diseases that may occur, the prevention and control measures for the current type of citrus disease are obtained, and the prevention and control points are generated based on the location information of the damaged citrus trees; Based on the control points, control marks are generated, and control marks are arranged for control according to the control measures for the current citrus disease type.

5. A citrus disease intelligent identification and precise prevention and control system, characterized in that: The system includes a memory and a processor, wherein the memory includes a program for intelligent identification and precise prevention and control of citrus diseases, and when the program for intelligent identification and precise prevention and control of citrus diseases is executed by the processor, the following steps are implemented: Obtain the number of occurrences of each citrus disease type within a preset time, calculate the probability value of each citrus disease type occurring under the citrus tree damage factor based on the number of occurrences of each citrus disease type within the preset time, and construct a knowledge graph; Acquire remote sensing image data of citrus fruit trees through unmanned aerial vehicle remote sensing technology, and construct a damage recognition model based on the remote sensing image data of citrus fruit trees; Based on the damage recognition model, remote sensing image data of citrus trees in the current target area are recognized to obtain recognition results, and probability values ​​of the occurrence of various citrus disease types are obtained according to the recognition results and the knowledge graph; Acquire the location information of the damaged citrus fruit trees according to the recognition results, issue an early warning based on the probability values ​​of the occurrence of each type of citrus disease and the location information of the damaged citrus fruit trees, and perform precise prevention and control according to the generated early warning information; Acquiring remote sensing image data of citrus trees through UAV remote sensing technology, and constructing a damage recognition model based on the remote sensing image data of citrus trees, specifically including: Acquire remote sensing image data of citrus trees by using UAV remote sensing technology, extract remote sensing image data of damaged citrus trees from the remote sensing image data of the citrus trees, filter and denoise the remote sensing image data of damaged citrus trees, and acquire pre-processed remote sensing image data; Introducing a feature pyramid network, inputting the preprocessed remote sensing image data into the feature pyramid network, obtaining feature image data of citrus trees when damaged, and constructing a damage recognition model based on a deep neural network; Constructing a plurality of damage type training sets based on the characteristic image data of the citrus tree when it is damaged, and inputting the damage type training sets into the damage recognition model for training; Setting a prediction accuracy threshold, when the prediction accuracy of the damage identification model is greater than the prediction accuracy threshold, saving the model parameters of the damage identification model, outputting the damage identification model, and completing the training; Acquiring remote sensing image data of citrus trees through UAV remote sensing technology can include the following steps: Constructing a UAV remote sensing monitoring network, obtaining the position information of the extreme position points in the area of ​​the target area to be collected, and initializing the position information of the communication equipment in the UAV remote sensing monitoring network; According to the location information of the communication equipment in the UAV remote sensing monitoring network and the location information of the extreme location points in the area of ​​the target area to be collected, a plurality of communication Euclidean distance values ​​are calculated, and a genetic algorithm is introduced, and a genetic algebra is set based on the genetic algorithm; Determine whether the communication Euclidean distance values ​​are all not greater than a preset communication Euclidean distance value, and when the communication Euclidean distance values ​​are all not greater than the preset communication Euclidean distance value, output the location information of the communication device in the UAV remote sensing monitoring network, and control the UAV to obtain remote sensing image data of the citrus tree through the UAV remote sensing monitoring network; When at least one of the communication Euclidean distance values ​​is greater than a preset communication Euclidean distance value, genetic iteration is performed based on the genetic algebra to readjust the location information of the communication device in the UAV remote sensing monitoring network until all the communication Euclidean distance values ​​are no greater than the preset communication Euclidean distance value; Constructing several damage type training sets based on the characteristic image data of citrus trees when damaged may also include the following steps: Building several training sets of damage types based on the characteristic image data of the citrus tree when it is damaged, randomly selecting training items in the training set of each damage type, and extracting damage features in each training item; Determine whether there are at least two damage features in the damage features in each training item. If there is only one damage feature in the damage features in each training item, use the corresponding sample data as a training item in the training set; If there are at least two damage features in the damage features in each training item, the Mahalanobis distance value between each damage feature is calculated, and when the Mahalanobis distance value is not greater than a preset Mahalanobis distance threshold, the corresponding sample data is used as a training item in the training set; When the Mahalanobis distance value is greater than the preset Mahalanobis distance threshold, the damage features in the training item are segmented and one of the features is eliminated. When the Mahalanobis distance value is not greater than the preset Mahalanobis distance threshold, the training item after the damage feature is eliminated is output, and the training item after the damage feature is eliminated is used as the training item in the training set.

6. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a program for the method for intelligent identification and precise prevention and control of citrus diseases. When the program for the method for intelligent identification and precise prevention and control of citrus diseases is executed by a processor, the steps of the method for intelligent identification and precise prevention and control of citrus diseases as described in any one of claims 1 to 4 are implemented.

Images