Precise targeted pesticide application method and device based on multi-source information perception

Abstract

The invention provides an accurate targeted pesticide application method and device based on multi-source information perception, and relates to the technical field of agricultural plant protection, and the method comprises the steps: obtaining multi-source information and a target image of a to-be-applied target in a planting environment; when it is identified that the to-be-applied target is infected with diseases and insect pests from the target image, a first pesticide application parameter of the pesticide application system is regulated and controlled to be a treatment mode pesticide application parameter according to the identified disease and insect pest information and a constructed treatment pesticide application decision model, so that treatment pesticide application is conducted on the disease and insect pest infected area of the to-be-applied target; and according to the multi-source information and the constructed preventive pesticide application decision model, the second pesticide application parameter of the pesticide application system is regulated and controlled to be a prevention mode pesticide application parameter, so that preventive pesticide application is performed on the non-pest-disease-infection area of the to-be-applied target. According to the invention, the problems of large target identification error, simple pesticide application decision and delayed pesticide application parameter regulation in the target pesticide application method in the prior art are solved.

Description

Technical Field

[0001] This invention relates to the field of agricultural plant protection technology, and in particular to a method and apparatus for precise targeted pesticide application based on multi-source information sensing. Background Technology

[0002] Traditional pesticide application methods rely on manual experience or uniform spraying, leading to excessive pesticide use, environmental pollution, and increased pesticide resistance. Precision application technologies, on the other hand, often rely on a single sensor (such as vision or lidar), which has low accuracy in identifying targets (pests and diseases) in complex environments and lacks real-time dynamic application decision-making.

[0003] Existing technologies propose a target-identification precision pesticide application system and method. This method determines the spraying amount, spraying direction, and spraying distance based on canopy contours, density distribution, target pest type, and pest characteristics. However, this method uses relatively limited data, lacks environmental information, and makes simplistic pesticide application decisions. Furthermore, it fails to detect latent pests, leading to significant target identification errors. Existing technologies also propose an agricultural robot-based regional target application method. This method utilizes a microprocessor equipped with a deep learning model of pests and weeds to predict the types of pests and weeds captured by a detection camera. It can quickly identify the types of pests and weeds and further control the opening and closing of corresponding solenoid valves based on the location of the pests and weeds for regional target application. However, it cannot control the application parameters in a timely manner according to the specific application situation, resulting in problems such as lag in parameter adjustment. Summary of the Invention

[0004] This invention provides a precise target-based drug delivery method and apparatus based on multi-source information perception, which solves the problems of large target identification errors, simple drug delivery decision-making, and lagging drug delivery parameter control in existing target delivery methods.

[0005] This invention provides a precise target-directed drug delivery method based on multi-source information sensing, the method comprising the following steps: Acquire multi-source information and target images of the target in the planting environment, including temperature and humidity information, light information, soil information, pathogen characteristic factors and pest pheromones; When it is identified from the target image that the target to be treated is infected with pests and diseases, the first application parameter of the application system is adjusted to the treatment mode application parameter based on the identified pest and disease information and the constructed treatment application decision model, and the second application parameter of the application system is adjusted to the prevention mode application parameter based on the multi-source information and the constructed prevention application decision model. According to the treatment mode application parameters, the pest-infected areas of the target to be treated are treated with pesticides, and according to the prevention mode application parameters, the non-pest-infected areas of the target to be treated are treated with pesticides.

[0006] In some embodiments, before acquiring multi-source information about the target in the planting environment and the target image, the method further includes: The system collects historical multi-source information about the target plant in its historical planting environment using sensors, including a temperature and humidity sensor for collecting temperature and humidity information, a light sensor for collecting light information, and a microbial sensor for collecting pathogen characteristic factors and pest pheromones. A preventative drug application decision model is constructed based on the aforementioned historical multi-source information; Historical images of the target to be treated are acquired using cameras, including RGB cameras and near-infrared cameras; A governance-based drug application decision model is constructed based on the historical images.

[0007] In some embodiments, constructing a governance-based drug application decision model based on the historical multi-source information includes: A first mapping relationship is established between the historical multi-source information and the amount of preventive pesticide application, and a preventive pesticide application decision model is constructed based on the first mapping relationship. The amount of preventive pesticide application is determined according to the preventive mode pesticide application parameters of the pesticide application system. The preventive mode pesticide application parameters include the on / off status of the pesticide nozzle, the pressure inside the nozzle, and the flow rate of the nozzle. The step of constructing a governance-based drug application decision model based on the historical images includes: Extract the historical pest and disease identification results from the historical images; A second mapping relationship is established between the historical pest and disease identification results and the severity level of pests and diseases, and a third mapping relationship is established between the severity level of pests and diseases and the application degree of the treatment. The application degree of the treatment is determined according to the treatment mode application parameters of the application system, which include the on / off status of the application nozzle, droplet size, and application rate. Based on the second mapping relationship and the third mapping relationship, a governance-based drug application decision model is constructed.

[0008] In some embodiments, the target image includes an RGB image and a near-infrared image, and the process of identifying the pest and disease information includes: Pest and disease detection is performed on the RGB image and the near-infrared image respectively to obtain the corresponding visible pest and disease information and potential pest and disease information, and to determine the pest and disease infected area and non-pest and disease infected area of ​​the target to be treated. The pest and disease information and potential pest and disease information are input into a pre-trained pest and disease classification model for prediction to obtain the severity level of pest and disease of the target to be treated.

[0009] In some embodiments, the method further includes: Point cloud data of the target to be treated is acquired using lidar; The growth information of the target to be treated is determined based on the point cloud data, and the growth information includes the canopy volume, leaf area and density distribution information of the target to be treated. When it is identified from the target image that the target to be treated is infected with pests and diseases, the application parameters of the application system are adjusted to the application parameters of the treatment mode based on the pest and disease information, the growth information and the treatment application decision model. The application parameters of the application system are adjusted to preventive application parameters based on the multi-source information, the growth information, and the preventive application decision model.

[0010] In some embodiments, the method further includes: When it is identified from the target image that the target to be treated is not infected with pests or diseases, the application system is adjusted to a preventive application mode based on the growth information, the multi-source information, and the preventive application decision model. The preventive application mode is used to automatically adjust the application parameters to preventive application parameters to preventive application to the target to be treated.

[0011] The present invention also provides a precise target-directed drug delivery device based on multi-source information sensing, the device comprising the following modules: An information sensing system is used to acquire multi-source information and target images of the target to be sprayed in the planting environment. The multi-source information includes temperature and humidity information, light information, soil information, pathogen characteristic factors and pest pheromones. The pesticide application control system is used to adjust the first pesticide application parameter of the pesticide application system to the control mode pesticide application parameter based on the identified pest and disease information and the constructed control pesticide application decision model when the target to be pesticided is identified from the target image. The system also adjusts the second pesticide application parameter of the pesticide application system to the prevention mode pesticide application parameter based on the multi-source information and the constructed prevention pesticide application decision model. The pesticide application system is used to perform therapeutic pesticide application on the pest-infected area of ​​the target to be treated according to the pesticide application parameters of the treatment mode, and to perform preventive pesticide application on the non-pest-infected area of ​​the target to be treated according to the pesticide application parameters of the prevention mode.

[0012] In some embodiments, the apparatus further includes: The pesticide application decision system is used to collect historical multi-source information of the target to be pesticided in the historical planting environment through sensors. The sensors include a temperature and humidity sensor for collecting temperature and humidity information, a light sensor for collecting light information, and a microbial sensor for collecting pathogen characteristic factors and pest pheromones. A preventative drug application decision model is constructed based on the aforementioned historical multi-source information; The drug application decision system is also used to acquire historical images of the target to be drugged via cameras, including RGB cameras and near-infrared cameras; A governance-based drug application decision model is constructed based on the historical images.

[0013] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the precise target drug delivery method based on multi-source information perception as described above.

[0014] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the precise target-directed drug delivery method based on multi-source information perception as described above.

[0015] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the precise target drug delivery method based on multi-source information perception as described above.

[0016] This invention provides a precise targeted pesticide application method and apparatus based on multi-source information perception. By acquiring multi-source information and images of the target in its planting environment, when the target is identified as infected with pests or diseases from the image, pesticide application decisions are made using the identified pest / disease information, multi-source information, and a pre-built decision model. The decision model enables real-time, automatic, and precise control of the pesticide application system's parameters, overcoming the shortcomings of existing technologies where pesticide application decisions are simple and parameter adjustment is lagging. Furthermore, utilizing target images to detect pests and diseases and distinguishing between infected and uninfected areas for targeted pesticide application solves the problem of large target identification errors. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced one by one below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating the precise targeted drug delivery method based on multi-source information perception provided by the present invention.

[0019] Figure 2 This is a schematic diagram of multi-source information sensing provided by the present invention.

[0020] Figure 3This is a schematic diagram illustrating the principle of the precise targeted drug delivery method based on multi-source information perception provided by the present invention.

[0021] Figure 4 This is a schematic diagram of the structure of the precision target-directed drug delivery device based on multi-source information perception provided by the present invention.

[0022] Figure 5 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0024] The following description, in conjunction with the accompanying drawings, outlines the precise targeted drug delivery method and apparatus of the present invention based on multi-source information sensing. Figure 1 This is a flowchart illustrating the precise target-directed drug delivery method based on multi-source information sensing provided by the present invention, as shown below. Figure 1 As shown, the method includes steps 101 to 103, which are described in detail below.

[0025] Step 101: Obtain multi-source information and target images of the target in the planting environment. The multi-source information includes temperature and humidity information, light information, soil information, pathogen characteristic factors and pest pheromones.

[0026] Here, the target of pesticide application can be fruit trees, crops, or plants that may be infected with pests or diseases. In this embodiment of the invention, fruit trees will be used as an example for explanation.

[0027] First, multi-source information and images of the target plant in its growing environment are obtained. Multi-source information refers to environmental information, specifically including temperature and humidity, light intensity, soil conditions, pathogen characteristic factors, and pest pheromones. This multi-source information can be obtained through methods such as... Figure 2 The multi-source information sensing device shown in the figure senses and collects the data.

[0028] like Figure 2As shown, the multi-source information sensing device includes a temperature and humidity sensor, a light sensor, and a microbial sensor. The temperature and humidity sensor collects temperature and humidity information, such as ambient temperature and humidity; the light sensor collects light information, such as light intensity; and the microbial sensor senses pathogenic factors and pest pheromones, such as airborne spores, insect eggs, and other microorganisms that can cause diseases in humans or animals, as well as harmful trace chemicals produced by these microorganisms.

[0029] The target image specifically includes RGB images and near-infrared images, such as... Figure 2 As shown, near-infrared images are acquired by a near-infrared camera in a multi-source information sensing device, while RGB images are acquired by an RGB camera. During image acquisition, both the near-infrared and RGB cameras target the same object and are acquired simultaneously; that is, both RGB and near-infrared images are acquired simultaneously for the fruit tree.

[0030] Here, by performing pest and disease detection on the target image, the corresponding pest and disease detection results are obtained. The detection results can mark the areas and locations of pests and diseases in the image, which can further determine whether the target to be treated has been infected with pests and diseases. Pest and disease detection can be achieved through deep learning models of target detection or image segmentation.

[0031] Step 102: When it is identified from the target image that the target to be treated is infected with pests and diseases, the first application parameter of the application system is adjusted to the treatment mode application parameter based on the identified pest and disease information and the constructed treatment application decision model, and the second application parameter of the application system is adjusted to the prevention mode application parameter based on multi-source information and the constructed prevention application decision model.

[0032] When the target image reveals that the target to be treated is infected with pests and diseases, it indicates that a pesticide application system is needed to treat and prevent the infestation. Pest and disease prevention typically involves applying pesticides to the non-pest-infested areas of the target to achieve preventative measures. Pest and disease treatment, on the other hand, involves applying pesticides to the infested areas of the target to treat and eliminate the pests and diseases.

[0033] For example Figure 2 As shown, pest and disease detection is performed on RGB and near-infrared images of fruit trees. If pests and diseases are found, a pesticide application system is needed to treat the affected areas, while preventative measures are taken for the non-affected (normal) areas. The pesticide application system can be an orchard sprayer or other intelligent spraying equipment.

[0034] The application of pesticides by the system relies on application decisions. In this embodiment of the invention, these decisions are made through pre-constructed control and preventative application decision models. These models are used to adjust the application parameters of the system to apply pesticides to diseased and pest-infested fruit trees, thereby achieving targeted control or prevention.

[0035] Here, based on the identified pest and disease information and the constructed control-oriented pesticide application decision model, the first application parameter of the pesticide application system is adjusted to the control-mode application parameter. Pest and disease information is obtained from target images of the targets to be treated, specifically including the severity level of the pests and diseases and their distribution. The distribution includes pest-infested and non-pest-infested areas of the target, which can be determined based on the pest and disease identification results of the target image. The severity level of the pests and diseases needs to be further predicted using a pest and disease grading model. The pest and disease identification results are input into the pest and disease grading model, which can assess and classify the severity of the pests and diseases, for example, into multiple levels such as severe, moderate, and mild.

[0036] The identified pest and disease information is input into the constructed control-oriented pesticide application decision model. The model then predicts the degree of pesticide application, which can be represented by the application level or the application rate. The application level can be either intensive or light application. Furthermore, the pesticide application system can adjust its application parameters based on this information to achieve the desired control mode, specifically including the on / off status of the spray nozzles, droplet size, and application rate.

[0037] Furthermore, based on multi-source information and the constructed preventative application decision model, the second application parameter of the application system is adjusted to the preventative application parameter. Preventative application differs from controllable application; preventative application requires controlling the appropriate application rate to achieve the purpose of prevention, while controllable application aims to eliminate pests and diseases and does not require controlling the application rate. Multi-source information includes temperature and humidity information, light information, soil information, pathogen characteristic factors, and pest pheromones. This multi-source information is input into the constructed preventative application decision model, which predicts the preventative application rate based on this information. For example, if there are many pathogen characteristic factors and pest pheromones, and the temperature, humidity, light, and soil information is poor, the application rate will be relatively higher; conversely, if there are few pathogen characteristic factors and pest pheromones, and the temperature, humidity, light, and soil information is good, the application rate will be relatively lower. At this point, the application system can adjust its application parameters according to the predicted application amount, switching to the prevention mode application parameters, specifically including the on / off status of the application nozzle, the internal pressure of the nozzle, and the flow rate of the nozzle.

[0038] Step 103: Apply pesticides to the pest-infected areas of the target area according to the pesticide application parameters of the treatment mode, and apply pesticides to the non-pest-infected areas of the target area according to the pesticide application parameters of the prevention mode.

[0039] Having determined the corresponding application parameters through the application decision in step 102, the final step is to execute the application process of the application system. The application process consists of two parts. The first part involves applying a treatment pesticide to the pest-infected area of ​​the target pest based on the treatment mode application parameters. These parameters allow for adjustments to the on / off status of the application nozzles, control of droplet size, and achieving appropriate pesticide dosage, ensuring effective pest control. The application targets the pest-infected area, for example... Figure 2 In this way, precise and targeted pesticide application can be achieved for the diseased and pest-infested areas of fruit trees.

[0040] The second part involves preventative application of pesticides to the non-pest-infested areas of the target area based on the preventative application parameters. These parameters allow for adjustments to the on / off status of the spray nozzles, control of the nozzle pressure, and management of the flow rate, ensuring effective pest and disease prevention with minimal pesticide application. The target area is the non-pest-infested area of ​​the target area, for example... Figure 2 In the non-pest-infested areas of fruit trees infected with pests and diseases, this allows for preventative and precise targeted application of pesticides.

[0041] Because treatment and prevention need to be carried out separately for the diseased and non-disease-affected areas of the target area, the application parameters of the pesticide application system need to be adjusted to both treatment and prevention modes during implementation to ensure targeted application. When there is only one spray nozzle, the application parameters can be switched back and forth between the diseased and non-disease-affected areas. However, when there are multiple spray nozzles, the application parameters of some nozzles can be adjusted to treatment mode, while those of others can be adjusted to prevention mode, thus applying the appropriate pesticide to the corresponding areas.

[0042] The following is combined with Figure 3 This describes the overall process of drug application in the above embodiments. For example... Figure 3As shown, by inputting multi-source information from temperature and humidity sensors, light sensors, and microbial sensors into the preventative pesticide application decision model, the application parameters for the preventative mode are determined, achieving precise targeted preventative pesticide application. Pest and disease identification is performed based on target images captured by RGB and near-infrared cameras, and the results are further used to determine pest and disease information, including the distribution of pests and diseases on the target area. Furthermore, the pest and disease identification results are input into a pest and disease grading model to predict the severity level of the pest and disease. Finally, the severity level of the pest and disease can be input into the control pesticide application decision model to determine the application parameters for the control mode, achieving precise targeted control pesticide application.

[0043] When applying pesticides, the first step is to determine whether the area is already infected with pests or diseases. If so, it is necessary to further determine whether the area is indeed infected. If so, targeted pesticide application is implemented for treatment. Otherwise, targeted pesticide application is implemented for prevention.

[0044] This invention, through acquiring multi-source information and images of the target plant in its growing environment, determines pesticide application based on the identified pest and disease information, multi-source information, and a pre-built decision model when the target plant is identified as infected with pests or diseases from the image. The decision model enables real-time, automatic, and precise control of the pesticide application system's parameters, overcoming the shortcomings of existing technologies where pesticide application decisions are simple and parameter adjustments are lagging. Furthermore, utilizing the target image for pest and disease detection and distinguishing between infected and uninfected areas for targeted pesticide application solves the problem of large target identification errors.

[0045] In the above embodiments, the pesticide application decision is made through a preventive pesticide application decision model and a remedial pesticide application decision model. These models are pre-built before the pesticide application decision is executed. Therefore, in some embodiments, before obtaining multi-source information on the target in the planting environment and the target image, it is also necessary to build a preventive pesticide application decision model and a remedial pesticide application decision model, which will be explained in detail below.

[0046] One is a preventative pesticide application decision model, which uses sensors to collect historical multi-source information about the target plant in its historical growing environment. These sensors include temperature and humidity sensors for collecting temperature and humidity information, light sensors for collecting light information, and microbial sensors for collecting pathogen characteristic factors and pest pheromones.

[0047] Here, historical multi-source information on the target pesticide in its historical planting environment can be obtained in real time through sensors over a historical period. This historical multi-source information still includes temperature and humidity information, light information, pathogen characteristic factors, and pest pheromones. The data collection process still uses sensors with fixed parameters.

[0048] Then, a preventative pesticide application decision model is constructed based on historical multi-source information. This model can be a neural network model capable of receiving multiple parameters. Using this historical multi-source information, the preventative pesticide application dosage is determined based on application experience. This dosage, along with the historical multi-source information, forms the data to train a neural network model. This neural network model can accurately predict the preventative pesticide application dosage based on the input multi-source information. The trained neural network model is the preventative pesticide application decision model.

[0049] The second is the preventative pesticide application decision model, which uses cameras to acquire historical images of the target area to be treated. These cameras include both RGB and near-infrared cameras. The historical images consist of both RGB and near-infrared images. The acquisition process still uses cameras with fixed parameters.

[0050] Then, a decision-making model for controlled pesticide application is constructed based on historical images. This model can also be a neural network model. For these historical images, relevant pest and disease information can be extracted first. Then, based on application experience, the application level for controlled pesticide application can be determined. The application level is used as a data label, forming a dataset with the pest and disease information. This dataset is then used to train the neural network model, enabling it to accurately predict the application level for controlled pesticide application based on the input pest and disease information. The trained neural network model is the controlled pesticide application decision-making model.

[0051] In this embodiment of the invention, by constructing a treatment-oriented drug application decision model and a preventive drug application decision model, drug application decisions can be made for the target to be treated, thereby enabling targeted drug application and achieving real-time, automatic, and precise control of the drug application parameters of the drug application system. This overcomes the shortcomings of the prior art, which has simple drug application decisions and lagging drug application parameter control.

[0052] In some embodiments, constructing a governance-based pesticide application decision model based on historical multi-source information can also be achieved in the following ways, which are described below.

[0053] First, a primary mapping relationship is established between historical multi-source information and the dosage for preventative pesticide application. Here, based on experience with preventative pesticide application, the corresponding dosage can be determined for different historical multi-source information sets, thus establishing the primary mapping relationship. The dosage for preventative pesticide application is determined based on the preventative mode application parameters of the pesticide application system, which include the on / off status of the spray nozzle, the internal pressure of the nozzle, and the nozzle flow rate. This ensures that the pesticide application system can adjust the preventative mode application parameters according to the preventative dosage, achieving precise targeted preventative pesticide application.

[0054] Next, a preventative medication decision-making model is constructed based on the first mapping relationship. This model can be implemented using a knowledge graph, where the first mapping relationship can be embedded as empirical knowledge into the knowledge graph used for medication decision-making. Based on the input historical multi-source information and leveraging the decision-making reasoning capabilities of the knowledge graph, the dosage for preventative medication can be accurately inferred.

[0055] The construction of a pest control decision-making model based on historical images can also be achieved in the following way. First, extract historical pest detection results from the historical images. This involves detecting pests in historical images and outputting the corresponding historical pest detection results, specifically images with marked areas and locations of pests. Then, a second mapping relationship can be established between the historical pest detection results and the severity level of the pests. Here, the severity level can be determined based on prior knowledge using the historical pest detection results, or it can be predicted using a pest classification model. Furthermore, a third mapping relationship can be established between the severity level of the pests and the application level of the pest control application. The application level of the pest control application is determined based on the application parameters of the pest control mode of the application system. These parameters include the on / off status of the application nozzles, droplet size, and application rate. This ensures that the application system can adjust the application parameters of the pest control mode according to the application rate, achieving precise and targeted pest control application.

[0056] Finally, based on the second and third mapping relationships, a governance-based drug application decision model is constructed. This model can be implemented using a knowledge graph, where the second and third mapping relationships can be considered as empirical knowledge and embedded into the knowledge graph used for drug application decisions. Based on the input historical multi-source information, the reasoning capabilities of the knowledge graph can accurately infer the appropriate dosage for governance-based drug application.

[0057] In this embodiment of the invention, a mapping relationship between the model prediction results is constructed by using historical multi-source information and historical images of the target to be treated, and this mapping relationship is implemented by using a knowledge graph, thereby ensuring that the governance-based drug application decision model can make effective and accurate drug application decisions.

[0058] In some embodiments, the identification process of pest and disease information can also be implemented in the following ways, as detailed below.

[0059] First, the target images of the objects to be treated include RGB images and near-infrared images. Pest and disease detection can be performed on both RGB and near-infrared images to obtain corresponding visible and potential pest and disease information. Pest and disease detection can be achieved using deep learning models for image target detection and image segmentation. By performing pest and disease detection on the RGB image, visible pest and disease information can be extracted, and visible pest and disease areas in the target object can be marked, such as insect eggs and infested areas on fruit tree leaves. Similarly, by performing pest and disease detection on the near-infrared image, potential pest and disease information can be extracted, and potentially pest and disease areas in the target object can be marked, such as yellowing areas on fruit tree leaves. These marked areas make it easy to determine the infested and non-infested areas of the target object.

[0060] Then, the existing pest and disease information, as well as potential pest and disease information, are input into a pre-trained pest and disease classification model for prediction, yielding the severity level of the pest and disease affecting the target area to be treated. Here, the existing pest and disease information and potential pest and disease information can be fused using image fusion or feature fusion methods. The resulting fused features or fused image are then input into the pre-trained pest and disease classification model, which can be implemented using classification algorithms or neural network models. The pest and disease classification model ultimately predicts the severity level of the pest and disease affecting the target area to be treated, thus concluding the pest and disease information identification process.

[0061] In this embodiment of the invention, visible pest and disease information and potential pest and disease information are divided by pest and disease detection, and further predicted by a pre-trained pest and disease classification model. Finally, the severity level of pest and disease, the area of ​​pest and disease infection and the area of ​​non-pest and disease infection of the target to be treated are determined, thus realizing accurate judgment of pest and disease and solving the problem of large target identification error in existing methods.

[0062] In some embodiments, when applying pesticides to a target, the growth information of the target also needs to be considered, as this information may affect the effectiveness of the application. Therefore, this embodiment of the invention also acquires point cloud data of the target using lidar, and determines the target's growth information based on the point cloud data. This growth information includes the target's canopy volume, leaf area, and density distribution information. Specifically, canopy volume refers to the three-dimensional spatial volume occupied by the target's canopy, leaf area is the distribution area of ​​the target's leaves, and density distribution information represents the density of the target's growth.

[0063] like Figure 2As shown, the multi-source information sensing device is also equipped with a lidar to collect point cloud data of the fruit trees. Then, the point cloud data is used to perform three-dimensional reconstruction of the fruit trees, thereby calculating the canopy volume, leaf area, and density distribution information of the fruit trees based on the reconstructed three-dimensional model.

[0064] Next, when it is identified from the target image that the target to be treated is infected with pests and diseases, the application parameters of the application system are adjusted to the application parameters of the treatment mode based on pest and disease information, growth information and the treatment application decision model, and the application parameters of the application system are adjusted to the application parameters of the prevention mode based on multi-source information, growth information and the prevention application decision model.

[0065] like Figure 3 As shown, the growth information determined based on the point cloud data collected by lidar is further used to help perform targeted drug delivery for both therapeutic and preventative purposes.

[0066] Specifically, pest and disease information and growth information are simultaneously input into the control-based pesticide application decision model for prediction. This model, combining the characteristics of pest and disease information and growth information, can predict a more accurate application level for control-based pesticide application. Then, based on this application level, the application parameters of the pesticide application system are adjusted to the control-based application parameters, achieving precise targeted pesticide application for control. Conversely, multi-source information and growth information are simultaneously input into the preventative pesticide application decision model for prediction. This model, combining the characteristics of multi-source information and growth information, can predict a more accurate application dosage for preventative pesticide application. Then, based on this dosage, the application parameters of the pesticide application system are adjusted to the preventative application parameters, achieving precise targeted preventative pesticide application.

[0067] Of course, in building both the governance-based and preventative pesticide application decision models, it is also necessary to collect historical growth information of the target to be detected using lidar with the same equipment parameters, and combine historical images and historical multi-source data to train the corresponding models, so that the trained models can further combine growth information to make more accurate pesticide application decisions.

[0068] In this embodiment of the invention, the growth information of the target to be treated is collected by lidar, and the growth information is combined to help the treatment and preventive treatment decision models make treatment decisions. This makes the decision results more accurate, which can not only improve the treatment effect but also save on the amount of medicine to be applied.

[0069] Furthermore, in some embodiments, when it is identified from the target image that the target to be treated is not infected with pests or diseases, the application system is adjusted to a preventive application mode based on growth information, multi-source information, and a preventive application decision model. The preventive application mode is used to automatically adjust the application parameters to preventive application parameters for preventive application to the target.

[0070] Here, when the target image of the target to be treated is detected as uninfected by pests during pest detection, it indicates that there are no pest-infected areas in the pest distribution, and the severity level of pests is low or nonexistent. Therefore, targeted application of pesticides is unnecessary, but latent pests may still exist. Thus, in this embodiment of the invention, even when the target is identified as uninfected, preventative targeted application of pesticides is still performed. That is, based on growth information, multi-source information, and the preventative application decision model, the application system is adjusted to a preventative application mode. This preventative application mode automatically adjusts the application parameters to preventative application parameters for preventative application of pesticides to the target. The specific preventative application process can be referred to in the above embodiment, and will not be repeated here.

[0071] like Figure 3 As shown, when applying pesticides, it is first determined whether the plant is already infected with pests or diseases. If not, there is no need to perform targeted pesticide application for treatment, but preventive targeted pesticide application is still performed through the preventive pesticide application decision model to achieve the effect of preventing pests and diseases.

[0072] In this embodiment of the invention, if the target to be treated is identified as not infected with pests or diseases, appropriate targeted pesticide application is performed instead of targeted pesticide application. This allows for flexible adjustment of the application mode based on the pest or disease infection status of the target, ensuring that the lowest possible dosage is applied to the target, thereby achieving timely pest and disease prevention.

[0073] The precise target-directed drug delivery device based on multi-source information perception provided by the present invention will be described below. The precise target-directed drug delivery device based on multi-source information perception described below can be referred to in correspondence with the precise target-directed drug delivery method based on multi-source information perception described above.

[0074] like Figure 4 As shown, the precision target application device based on multi-source information perception specifically includes: an information perception system 401, an application control system 402, and an application execution system 403.

[0075] Specifically, the information sensing system 401 is used to acquire multi-source information and target images of the target in the planting environment, including temperature and humidity information, light information, soil information, pathogen characteristic factors, and pest pheromones; the pesticide application control system 402 is used to adjust the first application parameter of the pesticide application system to the treatment mode application parameter according to the identified pest information and the constructed treatment application decision model when the target is identified as being infected with pests and diseases from the target image, and to adjust the second application parameter of the pesticide application system to the prevention mode application parameter according to the multi-source information and the constructed prevention application decision model; the pesticide application execution system 403 is used to perform treatment application on the pest-infected area of ​​the target according to the treatment mode application parameter, and to perform prevention application on the non-pest-infected area of ​​the target according to the prevention mode application parameter.

[0076] In some embodiments, such as Figure 4 As shown, the precision targeted pesticide application device based on multi-source information perception provided by the present invention also includes a pesticide application decision system 404. Specifically, the pesticide application decision system 404 is used to collect historical multi-source information of the target to be pesticided in the historical planting environment through sensors, including a temperature and humidity sensor for collecting temperature and humidity information, a light sensor for collecting light information, and a microbial sensor for collecting pathogen characteristic factors and pest pheromones; and to construct a preventive pesticide application decision model based on the historical multi-source information. The pesticide application decision system 404 is also used to acquire historical images of the target to be treated via a camera, including an RGB camera and a near-infrared camera; and to construct a governance-based pesticide application decision model based on the historical images.

[0077] It should be noted that the beneficial effects of the precision target-directed drug delivery device based on multi-source information perception mentioned above correspond to each other, so the beneficial effects of the precision target-directed drug delivery device based on multi-source information perception will not be elaborated here.

[0078] Figure 5 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 5As shown, the electronic device may include: a processor 510, a communications interface 520, a memory 530, and a communications bus 540, wherein the processor 510, the communications interface 520, and the memory 530 communicate with each other through the communications bus 540. The processor 510 can call logic instructions in the memory 530 to execute a precise target application method based on multi-source information perception. This method includes: acquiring multi-source information about the target in its planting environment and a target image, the multi-source information including temperature and humidity information, light information, soil information, pathogen characteristic factors, and pest pheromones; when the target is identified as being infected with pests or diseases from the target image, adjusting the first application parameter of the application system to the treatment mode application parameter based on the identified pest and disease information and a constructed treatment application decision model, and adjusting the second application parameter of the application system to the prevention mode application parameter based on the multi-source information and a constructed prevention application decision model; applying treatment medication to the pest-infected area of ​​the target according to the treatment mode application parameter, and applying prevention medication to the non-pest-infected area of ​​the target according to the prevention mode application parameter.

[0079] Furthermore, the logical instructions in the aforementioned memory 530 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0080] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the precise target application method based on multi-source information perception provided by the above methods. The method includes: acquiring multi-source information of the target to be applied in the planting environment and a target image, wherein the multi-source information includes temperature and humidity information, light information, soil information, pathogen characteristic factors, and pest pheromones; when it is identified from the target image that the target to be applied is infected with pests and diseases, adjusting the first application parameter of the application system to the treatment mode application parameter according to the identified pest and disease information and the constructed treatment application decision model, and adjusting the second application parameter of the application system to the prevention mode application parameter according to the multi-source information and the constructed prevention application decision model; performing treatment application on the pest-infected area of ​​the target to be applied according to the treatment mode application parameter, and performing prevention application on the non-pest-infected area of ​​the target to be applied according to the prevention mode application parameter.

[0081] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements the precise targeted pesticide application method based on multi-source information perception provided by the above methods. The method includes: acquiring multi-source information of the target to be pesticided in the planting environment and a target image, wherein the multi-source information includes temperature and humidity information, light information, soil information, pathogen characteristic factors, and pest pheromones; when it is identified from the target image that the target to be pesticided is infected with pests and diseases, adjusting the first pesticide application parameter of the pesticide application system to the treatment mode pesticide application parameter according to the identified pest and disease information and the constructed treatment pesticide application decision model, and adjusting the second pesticide application parameter of the pesticide application system to the prevention mode pesticide application parameter according to the multi-source information and the constructed prevention pesticide application decision model; performing treatment pesticide application on the pest and disease infected area of ​​the target to be pesticided according to the treatment mode pesticide application parameter, and performing prevention pesticide application on the non-pest and disease infected area of ​​the target to be pesticided according to the prevention mode pesticide application parameter.

[0082] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0083] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0084] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A precise targeted drug delivery method based on multi-source information sensing, characterized in that, The method includes: Acquire multi-source information and target images of the target in the planting environment, including temperature and humidity information, light information, soil information, pathogen characteristic factors and pest pheromones; When it is identified from the target image that the target to be treated is infected with pests and diseases, the first application parameter of the application system is adjusted to the treatment mode application parameter based on the identified pest and disease information and the constructed treatment application decision model, and the second application parameter of the application system is adjusted to the prevention mode application parameter based on the multi-source information and the constructed prevention application decision model. According to the treatment mode application parameters, the pest-infected areas of the target to be treated are treated with pesticides, and according to the prevention mode application parameters, the non-pest-infected areas of the target to be treated are treated with pesticides.

2. The method for precise targeted drug delivery based on multi-source information sensing according to claim 1, characterized in that, Before acquiring multi-source information and target images of the target in the planting environment, the method further includes: The system collects historical multi-source information about the target plant in its historical planting environment using sensors, including a temperature and humidity sensor for collecting temperature and humidity information, a light sensor for collecting light information, and a microbial sensor for collecting pathogen characteristic factors and pest pheromones. A preventative drug application decision model is constructed based on the aforementioned historical multi-source information; Historical images of the target to be treated are acquired using cameras, including RGB cameras and near-infrared cameras. A governance-based drug application decision model is constructed based on the historical images.

3. The method for precise targeted drug delivery based on multi-source information sensing according to claim 2, characterized in that, The step of constructing a governance-based drug application decision model based on the historical multi-source information includes: A first mapping relationship is established between the historical multi-source information and the amount of preventive pesticide application, and a preventive pesticide application decision model is constructed based on the first mapping relationship. The amount of preventive pesticide application is determined according to the preventive mode pesticide application parameters of the pesticide application system. The preventive mode pesticide application parameters include the on / off status of the pesticide nozzle, the pressure inside the nozzle, and the flow rate of the nozzle. The step of constructing a governance-based drug application decision model based on the historical images includes: Extract the historical pest and disease identification results from the historical images; A second mapping relationship is established between the historical pest and disease identification results and the severity level of pests and diseases, and a third mapping relationship is established between the severity level of pests and diseases and the application degree of the treatment. The application degree of the treatment is determined according to the treatment mode application parameters of the application system, which include the on / off status of the application nozzle, droplet size, and application rate. Based on the second mapping relationship and the third mapping relationship, a governance-based drug application decision model is constructed.

4. The method for precise targeted drug delivery based on multi-source information sensing according to claim 1, characterized in that, The target image includes RGB images and near-infrared images, and the process of identifying the pest and disease information includes: Pest and disease detection is performed on the RGB image and the near-infrared image respectively to obtain the corresponding visible pest and disease information and potential pest and disease information, and to determine the pest and disease infected area and non-pest and disease infected area of ​​the target to be treated. The pest and disease information and potential pest and disease information are input into a pre-trained pest and disease classification model for prediction to obtain the severity level of pest and disease of the target to be treated.

5. The method for precise targeted drug delivery based on multi-source information sensing according to claim 1, characterized in that, The method further includes: Point cloud data of the target to be treated is acquired using lidar; The growth information of the target to be treated is determined based on the point cloud data, and the growth information includes the canopy volume, leaf area and density distribution information of the target to be treated. When it is identified from the target image that the target to be treated is infected with pests and diseases, the application parameters of the application system are adjusted to the application parameters of the treatment mode based on the pest and disease information, the growth information and the treatment application decision model. The application parameters of the application system are adjusted to preventive application parameters based on the multi-source information, the growth information, and the preventive application decision model.

6. The method for precise targeted drug delivery based on multi-source information sensing according to claim 5, characterized in that, The method further includes: When it is identified from the target image that the target to be treated is not infected with pests or diseases, the application system is adjusted to a preventive application mode based on the growth information, the multi-source information, and the preventive application decision model. The preventive application mode is used to automatically adjust the application parameters to preventive application parameters to preventive application to the target to be treated.

7. A precise target-directed drug delivery device based on multi-source information sensing, characterized in that, The device includes: An information sensing system is used to acquire multi-source information and target images of the target to be sprayed in the planting environment. The multi-source information includes temperature and humidity information, light information, soil information, pathogen characteristic factors and pest pheromones. The pesticide application control system is used to adjust the first pesticide application parameter of the pesticide application system to the control mode pesticide application parameter based on the identified pest and disease information and the constructed control pesticide application decision model when the target to be pesticided is identified from the target image. The system also adjusts the second pesticide application parameter of the pesticide application system to the prevention mode pesticide application parameter based on the multi-source information and the constructed prevention pesticide application decision model. The pesticide application system is used to perform therapeutic pesticide application on the pest-infected area of ​​the target to be treated according to the pesticide application parameters of the treatment mode, and to perform preventive pesticide application on the non-pest-infected area of ​​the target to be treated according to the pesticide application parameters of the prevention mode.

8. The multi-source information sensing precision target delivery device according to claim 7, characterized in that, The device further includes: The pesticide application decision system is used to collect historical multi-source information of the target to be pesticided in the historical planting environment through sensors. The sensors include a temperature and humidity sensor for collecting temperature and humidity information, a light sensor for collecting light information, and a microbial sensor for collecting pathogen characteristic factors and pest pheromones. A preventative drug application decision model is constructed based on the aforementioned historical multi-source information; The drug application decision system is also used to acquire historical images of the target to be drugged via cameras, including RGB cameras and near-infrared cameras; A governance-based drug application decision model is constructed based on the historical images.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the precise target-directed drug delivery method based on multi-source information perception as described in any one of claims 1 to 6.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the precise target-directed drug delivery method based on multi-source information perception as described in any one of claims 1 to 6.

Images