Farmland management method and system, electronic equipment, storage medium and product

Abstract

The invention provides a farmland management method and system, an electronic device, a storage medium and a product, and the method comprises the steps: issuing an operation decision for a target land parcel to a target operation device under the condition of determining that the target land parcel needs to be operated, and indicating the target operation device to execute an operation task corresponding to the operation decision; the first operation characteristic data uploaded by the local monitoring server and the second operation characteristic data uploaded by the field monitoring device are compared and analyzed, and the operation execution condition of the target land parcel is determined and displayed in the farmland AR realistic picture; the first operation feature data is feature data used for representing the operation execution condition in the first farmland feature data, and the first farmland feature data is obtained by inputting the farmland AR live-action video stream into a first feature extraction model. According to the invention, the effectiveness of farmland data can be ensured, the accuracy of operation execution condition analysis is improved, the workload of manual offline investigation is saved, and the efficiency of farmland management is improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a farmland management method, system, electronic device, storage medium and product. Background Art

[0002] Smart farmland management is an important direction for the development of modern agriculture, which can effectively improve agricultural production efficiency, reduce costs, increase yields, and thus promote the sustainable development of agriculture.

[0003] Current smart farmland management solutions usually upload environmental data based on Internet of Things (IoT) sensing devices, and make corresponding decision-making to manage the farmland after platform statistical analysis. However, the existing solutions have the following disadvantages: (1) Most of the various environmental detection sensors transmit data through the 4G network. Due to the poor stability of 4G network transmission, there will be problems such as time delay and data loss caused by link damage, which directly affect the timeliness and integrity of environmental detection data, thereby reducing the effectiveness of the data, and further affecting the accuracy of subsequent data analysis; (2) The authenticity of the sensing data usually needs to be manually verified, and it is easy to make verification errors due to human factors, reducing the effectiveness of the data, and further resulting in a decrease in the accuracy of data analysis.

[0004] Therefore, how to improve the effectiveness of farmland data and the accuracy of data analysis, so as to improve the efficiency of farmland management is an urgent problem to be solved at present. Summary of the Invention

[0005] The present invention provides a farmland management method, system, electronic device, storage medium and product to improve the efficiency of farmland management.

[0006] The present invention provides a farmland management method applied to a cloud server, including: When it is determined that an operation needs to be performed on a target plot, issuing an operation decision for the target plot to a target operation device; the operation decision is used to instruct the target operation device to execute the operation task corresponding to the operation decision; Comparing and analyzing the first operation feature data uploaded by a local monitoring server with the second operation feature data uploaded by a field monitoring device to determine the operation execution situation of the target plot; the first operation feature data is the feature data in the first farmland feature data used to characterize the operation execution situation, and the first farmland feature data is obtained by inputting a farmland AR live video stream into a pre-trained first feature extraction model; Displaying the operation execution situation of the target plot in the farmland AR live map.

[0007] According to a farmland management method provided by the present invention, before sending an operation decision for a target plot to a target operation device when it is determined that an operation needs to be performed on the target plot, the method further includes: Obtain first farmland feature data uploaded by a local monitoring server and second farmland feature data uploaded by a field monitoring device; Input the first farmland feature data and the second farmland feature data into a farmland data analysis model to obtain a warning message; the warning message is data characterizing an adverse situation affecting the growth of crops; Determine a target plot corresponding to the warning message on a farmland AR real scene map.

[0008] According to a farmland management method provided by the present invention, the method further includes: When the field monitoring device is powered on, obtain first device location information uploaded by a local monitoring server and second device location information uploaded by the field monitoring device; wherein, the first device location information is obtained by inputting a farmland AR real scene video stream into a pre-trained second feature extraction model; If the first device location information is consistent with the second device location information, mark the field monitoring device on the farmland AR real scene map.

[0009] According to a farmland management method provided by the present invention, the method further includes: When performing inspections on each field monitoring device, obtain device exception information uploaded by a target field monitoring device and the working condition of an external status light of the target field monitoring device uploaded by a local monitoring server; wherein, the working condition of the external status light is obtained by inputting a farmland AR real scene video stream into a pre-trained third feature extraction model; Compare and analyze the device exception information with the working condition of the external status light to determine whether the target field monitoring device is abnormal; If the target field monitoring device is abnormal, issue a device exception alarm; the device exception alarm is used to prompt an operation and maintenance personnel that there is a device exception situation.

[0010] The present invention also provides another farmland management method, which is applied to a local monitoring server and includes: Obtain a farmland AR real scene video stream; Input the farmland AR real scene video stream into a pre-trained feature extraction model to obtain a feature extraction result; Wherein, the feature extraction model includes a first feature extraction model; the inputting the farmland AR real scene video stream into a pre-trained feature extraction model to obtain a feature extraction result includes: When it is determined that operations need to be performed on the target plot, input the farmland AR live video stream into a pre-trained first feature extraction model to obtain first farmland feature data; the first farmland feature data includes first operation feature data for characterizing the operation execution situation. Upload the first operation feature data to the cloud server; the first operation feature data is used to instruct the cloud server to compare and analyze the first operation feature data with second operation feature data uploaded by field monitoring devices, determine the operation execution situation of the target plot, and display the operation execution situation of the target plot in the farmland AR live map.

[0011] According to a farmland management method provided by the present invention, the feature extraction model further includes a second feature extraction model and a third feature extraction model; the step of inputting the farmland AR live video stream into the pre-trained feature extraction model to obtain a feature extraction result further includes: Input the farmland AR live video stream into a pre-trained second feature extraction model to obtain first device position information of the field monitoring device. Input the farmland AR live video stream into a pre-trained third feature extraction model to obtain the working condition of the external status light of the field monitoring device.

[0012] According to a farmland management method provided by the present invention, the step of obtaining the farmland AR live video stream includes: Obtain multiple farmland images taken by a farmland AR panoramic acquisition device in all directions at the same position at each moment. Stitch the multiple farmland images to obtain an initial farmland AR panoramic image. Perform post-processing on the initial farmland AR panoramic image to obtain farmland AR live images at each moment. Arrange the farmland AR live images at each moment in time sequence to obtain a farmland AR live video stream.

[0013] According to a farmland management method provided by the present invention, the feature extraction model is trained in the following manner: Extract a preset number of farmland AR image frames from the farmland AR live video stream. Perform data augmentation on the farmland AR image frames and label key regions to obtain sample farmland AR image frames. Construct a sample data set; the sample data set includes sample farmland AR image frames and corresponding true labels; the true labels include any one of true farmland feature data, true device position information, and the true working condition of the external status light. Input the sample farmland AR image frames into a convolutional neural network model for feature extraction to obtain sample feature extraction results. Based on the loss between the sample feature extraction result and the true label, adjust the parameters of the convolutional neural network model to obtain a pre-trained feature extraction model.

[0014] The present invention also provides a farmland management system, including: A cloud server, configured to, when it is determined that operations need to be performed on a target plot, send an operation decision for the target plot to a target operation device; the operation decision is used to instruct the target operation device to execute an operation task corresponding to the operation decision; A local monitoring server, configured to input a farmland AR real-scene video stream into a pre-trained first feature extraction model to obtain first farmland data; the first farmland data includes first operation feature data for characterizing the operation execution situation; and upload the first operation feature data to the cloud server in real time; The cloud server is further configured to compare and analyze the first operation feature data uploaded by the local monitoring server with second operation feature data uploaded by field monitoring devices, determine the operation execution situation of the target plot, and display the operation execution situation of the target plot in the farmland AR real-scene map.

[0015] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, where when the processor executes the computer program, the farmland management method described in any one of the above is implemented.

[0016] The present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the farmland management method described in any one of the above is implemented.

[0017] The present invention also provides a computer program product, including a computer program, and when the computer program is executed by a processor, the farmland management method described in any one of the above is implemented.

[0018] The farmland management method, system, electronic device, storage medium and product provided by the present invention obtain operation feature data by inputting the farmland AR live video stream into the feature extraction model, compare and analyze the operation feature data uploaded by the local monitoring server with the operation feature data uploaded in real time by the field monitoring devices to determine the real-time operation execution situation of the target plot, use the farmland feature data analyzed by AI to verify the validity of the data uploaded by the field monitoring devices, without the need for manual data validity verification, and can avoid the impact of data loss caused by time delay and link damage on data validity when the monitoring data is transmitted through the 4G network, which is beneficial to ensuring the validity of farmland data, thereby improving the accuracy of operation execution situation analysis; by automatically determining the real-time operation execution situation of the target plot after issuing the operation decision to confirm whether the decision has been executed, the workload of manual on-site inspection can be saved, thereby improving the efficiency of farmland management. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0020] Figure 1 is one of the schematic flowcharts of the farmland management method provided by the embodiment of the present invention.

[0021] Figure 2 is a schematic diagram of the farmland data monitoring architecture provided by the embodiment of the present invention.

[0022] Figure 3 is another schematic flowchart of the farmland management method provided by the embodiment of the present invention.

[0023] Figure 4 is yet another schematic flowchart of the farmland management method provided by the embodiment of the present invention.

[0024] Figure 5 is a schematic diagram of the structure of the farmland management system provided by the embodiment of the present invention.

[0025] Figure 6 is a schematic diagram of the architecture of the intelligent farmland management platform deployed on the cloud server provided by the embodiment of the present invention.

[0026] Figure 7 is a schematic diagram of the structure of the electronic device provided by the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0027] To make the objectives, technical solutions and advantages of the present invention clearer, the following will clearly and completely describe the technical solutions in the present invention in conjunction with the accompanying drawings in the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present invention without making creative efforts belong to the scope of protection of the present invention.

[0028] In the description of the embodiments of the present invention, the terms "including", "comprising" or any other variant thereof are intended to cover non-exclusive inclusion, such that a process, method, article or device including a series of elements not only includes those elements but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0029] Figure 1 is one of the flow schematic diagrams of the farmland management method provided by the embodiments of the present invention. Refer to Figure 1 , the embodiments of the present invention provide a farmland management method, and the method may specifically include the following steps: Step 101, when it is determined that operations need to be performed on a target plot, issue an operation decision for the target plot to a target operation device; the operation decision is used to instruct the target operation device to execute the operation task corresponding to the operation decision.

[0030] It should be noted that the execution subject of the farmland management method provided by the embodiments of the present invention may be a cloud server, and a smart farmland management platform may be deployed on the cloud server for farmland management.

[0031] Farmland can be divided into multiple plots. In some embodiments, when it is analyzed based on the farmland data of the target plot that there are adverse situations affecting the growth of crops in the target plot, it can be determined that operations need to be performed on the target plot, and thus decision-making analysis can be performed based on the farmland data of the target plot to determine the operation decision for the target plot and issue it to the target operation device.

[0032] In some embodiments, the operation decision may include an irrigation decision, a plant protection decision, etc.; the operation task corresponding to the irrigation decision is an irrigation task, and the operation task corresponding to the plant protection decision is an irrigation task.

[0033] In some embodiments, the operation device may include an intelligent irrigation control device, a plant protection drone, etc. The intelligent irrigation control device can be used to execute an automatic irrigation task; the plant protection drone can be used to execute an automatic plant protection task.

[0034] Step 102: Compare and analyze the first job feature data uploaded by the local monitoring server with the second job feature data uploaded by the field monitoring device to determine the job execution status of the target plot. The first job feature data is the feature data used to represent the job execution status in the first farmland feature data, and the first farmland feature data is obtained by inputting the farmland AR live video stream into a pre-trained first feature extraction model.

[0035] In the embodiment of the present invention, the local monitoring server can continuously obtain the farmland AR (Augmented Reality) live video stream. During the process of the target job device executing the job task, the local monitoring server can input the image frames in the continuously obtained farmland AR live video stream into the pre-trained first feature extraction model to obtain the first farmland data, and upload the first job feature data used to represent the job execution status in the first farmland data to the cloud server in real time. The cloud server can obtain the first job feature data analyzed by the AI model by the local monitoring server in real time, and compare and analyze it with the second job feature data obtained by the field monitoring device itself and uploaded in real time by the field monitoring device, so as to determine the real-time job execution status of the target plot.

[0036] Aiming at the problem in the prior art that various environmental detection data are usually uploaded regularly, the data volume is limited, and it is difficult to comprehensively reflect the real situation of the farmland environment and crop growth, resulting in low accuracy of data analysis. In the embodiment of the present invention, the job feature data uploaded by the field monitoring device in real time can comprehensively reflect the real situation of the farmland environment and crop growth, which is beneficial to ensuring the data validity, thereby further improving the accuracy of data analysis.

[0037] In the embodiment of the present invention, by using the feature data extracted from the farmland AR live video stream, the data validity of the feature data uploaded by the field monitoring device can be tested. When it is compared and analyzed that the first job feature data is consistent with the second job feature data, the second job feature data can effectively reflect the job situation. At this time, the job execution status of the target plot can be analyzed and determined by combining the first job feature data and the second job feature data. When it is compared and analyzed that the first job feature data is inconsistent with the second job feature data, it indicates that there may be a problem that the field monitoring device fails or the data transmitted by the field monitoring device through the 4G network is lost, that is, the second job feature data cannot effectively reflect the job situation. At this time, the job execution status of the target plot can be directly determined based on the first job feature data, avoiding the problem that relying solely on the feature data sent by the field monitoring device is prone to inaccurate job execution status due to invalid data, thereby improving the accuracy of job execution status analysis.

[0038] In the embodiment of the present invention, by comparing the first operation characteristic data with the second operation characteristic data monitored by the field monitoring device itself, it is not only possible to accurately judge the execution situation after the operation decision is issued, but also there is no need for manual offline confirmation of the operation execution situation, thus saving the workload of manual offline inspection.

[0039] In the embodiment of the present invention, the intelligent farm management platform deployed on the cloud server can make agricultural decisions and issue instructions after phenotypic data AI model comparison and analysis by real-time analyzing the data uploaded by the local monitoring server and the environmental data uploaded by each field monitoring device. After the relevant modules respond, it can confirm that the decision has been executed through AI video stream data comparison, so as to realize remote automatic confirmation of field linkage.

[0040] In the case of executing different decision tasks, the operation characteristic data for comparison and analysis can be different. For example, in the case of executing an irrigation task, the operation characteristic data for comparison and analysis can be humidity data; in the case of executing a plant protection task, the operation characteristic data for comparison and analysis can be the drug residue amount on the surface of the crops.

[0041] In the embodiment of the present invention, by performing AI analysis on the farm AR panoramic video stream on the local monitoring server, only the extracted farmland characteristic data needs to be reported to the cloud server, and there is no need to report the video stream to the cloud server in real time, which is beneficial to reducing the data transmission volume and improving the farmland management efficiency.

[0042] A large amount of data formed by video stream extraction and preprocessing in the embodiment of the present invention can be used for training and learning of the agricultural operation model after secondary extraction in combination with the sensing data (i.e., the data monitored and uploaded by the field monitoring device), so as to improve the model accuracy, thereby improving the accuracy of agricultural decisions, and realizing the secondary cleaning and classification function of the sensing data through the video stream.

[0043] In the embodiment of the present invention, after a large amount of data streams are generated based on the AR real scene for AI comparison training, the execution situation after the decision is issued can be accurately judged, and the full-automatic receipt closed-loop of the decision task completion of the intelligent platform is realized.

[0044] Step 103, display the operation execution situation of the target plot in the farm AR real scene map.

[0045] In some embodiments, after the irrigation decision is manually verified and the information is correct, the irrigation operation can be remotely executed according to the requirements of the irrigation decision, and the operation instructions are synchronously uploaded and marked in the farmland AR real scene. After the operation is completed, the execution effect of the operation plot can be viewed through the farmland AR real scene to complete the task closed-loop; after the plant protection decision is manually verified and the information is correct, the UAV plant protection operation can be remotely executed according to the requirements of the plant protection decision, and the operation instructions are synchronously uploaded and marked in the farmland AR real scene. After the operation is completed, the execution effect of the operation plot can be viewed through the AR real scene to complete the task closed-loop.

[0046] Figure 2 is a schematic diagram of the farmland data monitoring architecture provided by the embodiments of the present invention. Refer to Figure 2 , in some embodiments, the farmland real scene AR presentation can be performed through an aerial camera, and field intelligent modules such as a soil moisture module, a meteorological monitoring module, a disaster monitoring module, a seedling condition monitoring module, and an intelligent irrigation module are embedded in the real scene AR map, so that phenotypic data can be observed in real time through images, and the data is transmitted back to the cloud server. After the cloud server algorithm model analyzes the environmental data, management decisions are made, and after the linkage is executed, the execution effect is confirmed through AI video stream comparison.

[0047] By deploying a smart farm management platform on the cloud server, it is possible to collect, manage, analyze, and make decisions in modules, display the real-time working conditions of each field monitoring device in the form of real scene AR, and perform analysis and comparison based on the farmland AR real scene video stream to assist in intelligent management work such as confirmation of decision results, real-time capture and push of early warning information, etc.

[0048] The local monitoring server can be used as a data storage and computing unit for big data storage, transmission, and computing. It performs AI analysis on the farmland AR real scene video stream, adds labels to the relevant eigenvalue data according to the model algorithm, and then pushes it to the cloud server for the next decision-making judgment.

[0049] In some embodiments, the field monitoring devices may include soil moisture monitoring devices, meteorological monitoring devices, seedling condition monitoring devices, pest condition monitoring devices, disaster monitoring devices, and field green prevention and control devices, etc.

[0050] The soil moisture monitoring devices can be distributed to independent irrigation areas according to the farmland section conditions, monitor the temperature, humidity, PH value, conductivity, nitrogen, phosphorus, and potassium data of multiple layers of soil in this section, and regularly upload them to the cloud server.

[0051] The meteorological monitoring devices can provide data such as air temperature, air humidity, wind speed, wind direction, rainfall, illuminance, and atmospheric pressure in the monitored area and regularly upload them to the cloud server.

[0052] The crop condition monitoring device can monitor the growth cycle of crops in a specific area through multispectral or video, and regularly upload spectral information or phenotypic picture data to the cloud server.

[0053] The pest condition monitoring device can attract and collect pests through the pest condition monitoring device, and regularly upload pest damage data to the cloud server.

[0054] The disaster condition monitoring device can monitor disasters and diseases of different crops, and regularly upload disease data to the cloud server.

[0055] The field green prevention and control device can use the phototaxis of pests or the negative pressure air suction method to physically kill pests, and regularly report the number of pests to the cloud server.

[0056] The intelligent irrigation control device can be used to issue operation instructions to the target plot to execute the automatic irrigation task after the analysis and decision-making on the intelligent farmland management platform.

[0057] The plant protection UAV can be used to issue operation instructions to the target plot to execute the automatic plant protection task after the analysis and decision-making on the intelligent farmland management platform.

[0058] The existing intelligent farmland management solution can also be based on digital twin technology. After mapping agricultural sensing physical objects to the digital twin system, relevant management is carried out. However, this solution has the following disadvantages: (1) The modeling cycle is long. A large amount of data needs to be collected on-site for model establishment, which takes a long time; (2) The operation results of physical object mapping cannot be judged. After remotely controlling field devices through the digital twin system, it is impossible to determine whether the site has executed according to the instructions. In the embodiments of the present invention, by using AR technology, farmland data can be displayed in the farmland AR real scene map and relevant management can be carried out. The farmland AR real scene map is obtained by stitching multiple farmland images, without the need to collect a large amount of data on-site, saving a lot of time; and after issuing the operation decision for the target plot to the target operation device, the real-time operation situation of the target plot can be displayed in real time in the farmland AR real scene map, so as to facilitate judging whether the operation task corresponding to the operation decision has been executed on the target plot.

[0059] In the embodiment of the present invention, by inputting the farmland AR real - scene video stream into the feature extraction model to obtain operation feature data, comparing and analyzing the operation feature data uploaded by the local monitoring server with the operation feature data uploaded in real - time by the field monitoring devices to determine the real - time operation execution situation of the target plot, using the farmland feature data analyzed by AI to verify the validity of the data uploaded by the field monitoring devices, eliminating the need for manual data validity verification, and avoiding the impact of data loss caused by time delay and link damage on data validity when the monitoring data is transmitted through the 4G network, which is beneficial to ensuring the validity of farmland data, thereby improving the accuracy of operation execution situation analysis; by automatically determining the real - time operation execution situation of the target plot after issuing the operation decision to confirm whether the decision has been executed, the workload of manual on - site inspection can be saved, thus improving the efficiency of farmland management.

[0060] In an alternative embodiment, before issuing the operation decision for the target plot to the target operation device when it is determined that operations need to be performed on the target plot, it further includes: obtaining the first farmland feature data uploaded by the local monitoring server and the second farmland feature data uploaded by the field monitoring devices; inputting the first farmland feature data and the second farmland feature data into the farmland data analysis model to obtain a warning message; the warning message is data representing adverse situations affecting crop growth; determining the target plot corresponding to the warning message on the farmland AR real - scene map.

[0061] In some embodiments, the farmland feature data may include soil data, meteorological data, pest and disease data, etc. The local monitoring server can input the farmland AR real - scene video stream into the feature extraction model to extract the first farmland feature data such as soil data, meteorological data, pest and disease data, etc., and upload the first farmland feature data to the cloud server. At the same time, each field monitoring device also uploads the second farmland feature data monitored by itself to the cloud server.

[0062] The cloud server can monitor and analyze through the soil moisture, meteorology, and pest and disease large - scale models, timely screen and warn of the adverse situations affecting crop growth in the farmland feature data, push the warning message to the management personnel, and at the same time push it to the corresponding IoT coordinate points on the farmland AR real - scene map, determine the target plot corresponding to the warning message on the farmland AR real - scene map, so as to call the corresponding warning plot information through AR real - scene for manual verification.

[0063] In the embodiment of the present invention, by combining the farmland feature data transmitted in real - time back by each field monitoring device with the farmland feature data extracted from the farmland AR real - scene video stream for model analysis, it is possible to timely and accurately warn of the adverse situations affecting crop growth, eliminating the need for manual on - site confirmation of the farmland situation, thereby improving the efficiency of farmland management.

[0064] In some embodiments, by combining the data real-time transmitted back by the field monitoring device with the AR live video stream, a large amount of video stream data and sensing data can be extracted to train and learn the irrigation, disaster, disease, and pest models. By continuously optimizing the models, the accuracy of agricultural decision-making can be improved.

[0065] In an alternative embodiment, the method further includes: when the field monitoring device is powered on, obtaining the first device location information uploaded by the local monitoring server and the second device location information uploaded by the field monitoring device; wherein, the first device location information is obtained by inputting the farmland AR live video stream into a pre-trained second feature extraction model; if the first device location information is consistent with the second device location information, the field monitoring device is marked on the farmland AR real scene map.

[0066] In the embodiments of the present invention, the field monitoring device can be installed and deployed according to the actual situation of the farmland plot. After being powered on, the field monitoring device can perform device positioning based on the Beidou positioning system or GPS (Global Positioning System), and upload the device location information to the cloud server. The cloud server can receive the first device location information reported by various field monitoring devices. At the same time, the local monitoring server can input the farmland AR live video stream into a pre-trained second feature extraction model, obtain the second device location information of each field monitoring device, and report it to the cloud server.

[0067] The cloud server can match the position coordinate information of each field monitoring device with the AR real scene coordinate library, that is, compare the first device location information with the first device location information uploaded by the device itself. If the comparison is consistent, it indicates that the first device location information uploaded by the device itself is correct, and then the field monitoring device can be automatically marked on the farmland AR real scene map, realizing the automatic identification of the field monitoring device and its inclusion in the farmland AR real scene map.

[0068] In some embodiments, the field monitoring device can also collect the farmland geographical basic information and upload it to the cloud server. After receiving the farmland geographical basic information, the cloud server can complete the calibration of the farmland geographical basic information, map the physical plot marking information to the real scene AR image, and correspond and bind the two-dimensional plane geographical information with the real scene AR three-dimensional geographical information for display.

[0069] In an alternative embodiment, the method further includes: when performing inspections on each field monitoring device, obtaining device anomaly information uploaded by a target field monitoring device and the operating condition of the external status light of the target field monitoring device uploaded by a local monitoring server; wherein, the operating condition of the external status light is obtained by inputting a farmland AR real-scene video stream into a pre-trained third feature extraction model; comparing and analyzing the device anomaly information with the operating condition of the external status light to determine whether there is an anomaly in the target field monitoring device; if there is an anomaly in the target field monitoring device, issuing a device anomaly alarm; the device anomaly alarm is used to prompt the operation and maintenance personnel that there is a device anomaly situation.

[0070] In an embodiment of the present invention, the entire set of field monitoring devices can be regularly inspected and linked with the AR real scene. When a device is abnormal, the cloud server can detect the abnormality of the device based on the data reported by the field monitoring device. At this time, the device at the corresponding position can be remotely inspected by calling up the AR real scene in the farmland. The local monitoring server inputs the AR real-scene data stream into the third feature extraction model to extract the operating condition of the external status light of the device, and uploads the abnormal condition of the external status light of the device to the cloud server for comparison and analysis to determine whether there is an abnormality in the target field monitoring device, so as to inform the operation and maintenance personnel of the specific situation of the device abnormality.

[0071] In some embodiments, when the external status light shows red, it can indicate that the external status light is in an abnormal operating condition; when the external status light shows green, it can indicate that the external status light is in a normal operating condition.

[0072] In an embodiment of the present invention, by using the AR data stream comparison to identify the operating condition of the external status light of the device, and after the cloud server performs comparison and analysis and classifies the inspection results, an early warning is automatically pushed to relevant personnel, which can timely and accurately determine whether there is an abnormality in the field monitoring device, without the need for manual offline confirmation of whether the device is abnormal, thereby improving the efficiency of farmland management.

[0073] Figure 3 is the second flowchart of the farmland management method provided by an embodiment of the present invention. Refer to Figure 3 , an embodiment of the present invention provides a farmland management method, and the method may specifically include the following steps: Step 301, obtain a farmland AR real-scene video stream; Step 302: Input the farmland AR real-scene video stream into a pre-trained feature extraction model to obtain a feature extraction result. Among them, the feature extraction model includes a first feature extraction model. The step of inputting the farmland AR real-scene video stream into the pre-trained feature extraction model to obtain a feature extraction result includes: when it is determined that operations need to be performed on a target plot, input the farmland AR real-scene video stream into the pre-trained first feature extraction model to obtain first farmland feature data. The first farmland feature data includes first operation feature data used to characterize the operation execution situation. Upload the first operation feature data to the cloud server. The first operation feature data is used to instruct the cloud server to compare and analyze the first operation feature data with second operation feature data uploaded by field monitoring devices, determine the operation execution situation of the target plot, and display the operation execution situation of the target plot in the farmland AR real-scene map.

[0074] It should be noted that the execution subject of the farmland management method provided in the embodiments of the present invention can be a local monitoring server, which is communicatively connected to the cloud server. A smart farmland management platform can be deployed on the cloud server for farmland management.

[0075] The farmland can be divided into multiple plots. In some embodiments, when it is analyzed based on the farmland data of the target plot that there are adverse situations affecting the growth of crops in the target plot, it can be determined that operations need to be performed on the target plot. Thus, decision-making analysis can be carried out based on the farmland data of the target plot to determine the operation decision for the target plot and send it to the target operation device.

[0076] In some embodiments, the operation decision can include an irrigation decision, a plant protection decision, etc.; the operation task corresponding to the irrigation decision is an irrigation task, and the operation task corresponding to the plant protection decision is a plant protection task.

[0077] In some embodiments, the operation device can include an intelligent irrigation control device, a plant protection unmanned aerial vehicle, etc. The intelligent irrigation control device can execute an automatic irrigation task on the target plot after the smart farmland management platform analyzes the decision and sends the operation instruction; the plant protection unmanned aerial vehicle can execute an automatic plant protection task on the target plot after the smart farmland management platform analyzes the decision and sends the operation instruction.

[0078] In an embodiment of the present invention, the local monitoring server can continuously obtain the farmland AR real - scene video stream. During the process of the target operation device executing the operation task, the local monitoring server can input the image frames in the real - time obtained farmland AR real - scene video stream into a pre - trained first feature extraction model to obtain first farmland feature data, and upload the first operation feature data used to characterize the operation execution situation in the first farmland feature data to the cloud server in real time. The cloud server can obtain in real time the first operation feature data analyzed by the local monitoring server through the AI model, and compare and analyze it with the second operation feature data obtained by the field monitoring device itself and uploaded by the field monitoring device in real time, so as to determine the real - time operation execution situation of the target plot.

[0079] In the embodiment of the present invention, by inputting the farmland AR real - scene video stream into the feature extraction model to obtain the operation feature data, comparing and analyzing the operation feature data uploaded by the local monitoring server with the operation feature data uploaded by the field monitoring device in real time to determine the real - time operation execution situation of the target plot, the validity of the data uploaded by the field monitoring device is verified by using the farmland feature data analyzed by AI, without manual data validity verification, and the influence of data loss caused by time delay and link damage on data validity when the monitoring data is transmitted through the 4G network can be avoided, which is beneficial to ensuring the validity of farmland data, thereby improving the accuracy of operation execution situation analysis; by automatically determining the real - time operation execution situation of the target plot after issuing the operation decision to confirm whether the decision has been executed, the workload of manual on - site inspection can be saved, thereby improving the efficiency of farmland management.

[0080] In an alternative embodiment, the feature extraction model further includes a second feature extraction model and a third feature extraction model; the step of inputting the farmland AR real - scene video stream into the pre - trained feature extraction model to obtain the feature extraction result further includes: inputting the farmland AR real - scene video stream into the pre - trained second feature extraction model to obtain the first device position information of the field monitoring device; inputting the farmland AR real - scene video stream into the pre - trained third feature extraction model to obtain the working condition of the external status lights of the field monitoring device.

[0081] In an embodiment of the present invention, field monitoring equipment can be installed and deployed according to the actual situation of the farmland plot. After the field monitoring equipment is positioned based on the Beidou positioning system or GPS (Global Positioning System), the device location information can be uploaded to the cloud server. The cloud server can receive the first device location information reported by various types of field monitoring equipment. At the same time, the local monitoring server can input the farmland AR real-scene video stream into the pre-trained second feature extraction model to obtain the second device location information of each field monitoring device and report it to the cloud server. The cloud server can match the location coordinate information of each field monitoring device with the AR real-scene coordinate library, that is, compare the first device location information with the first device location information uploaded by the device itself. If the comparison is consistent, it means that the first device location information uploaded by the device itself is correct, then the field monitoring equipment can be automatically marked on the farmland AR real-scene map, thereby realizing the automatic identification of the field monitoring equipment and incorporating it into the farmland AR real-scene map.

[0082] In an embodiment of the present invention, the entire set of field monitoring equipment can be regularly inspected and linked with the AR real scene. When the equipment is abnormal, the cloud server can detect the abnormality of the equipment based on the data reported by the field monitoring equipment. At this time, the corresponding position equipment can be called up in the farmland AR real scene for remote inspection. The local monitoring server inputs the AR real scene data stream into the third feature extraction model to extract the working condition of the external status light of the equipment, and uploads the abnormal condition of the external status light of the equipment to the cloud server for comparison and analysis to determine whether the target field monitoring equipment is abnormal, thereby informing the operation and maintenance personnel of the specific situation of the equipment abnormality.

[0083] In an optional embodiment, the method of obtaining an AR real-scene video stream of a farmland includes: obtaining multiple farmland images taken by a farmland AR panoramic acquisition device at the same location at various times; splicing the multiple farmland images to obtain an initial farmland AR panoramic image; post-processing the initial farmland AR panoramic image to obtain an AR real-scene image of the farmland at various times; and arranging the AR real-scene images of the farmland at various times in chronological order to obtain an AR real-scene video stream of the farmland.

[0084] In some embodiments, information label marking can be performed after 360° live-action AR viewfinder shooting by an aerial camera. Specifically, a dedicated panoramic camera or a multi-camera array can be used to capture multiple farmland images in all directions from the same position, and a complete environmental perspective can also be obtained to complete the shooting and collection of farmland images. Then, these photos can be stitched into a 360-degree panoramic image by a monitoring server, and post-processing such as color adjustment, exposure balance, and seam trace elimination can be performed to ensure the smoothness and realism of the final image, thus completing the image stitching and processing. Information labels can be added to the panoramic image, enabling the administrator to not only view the panoramic image but also interact with the scene through operations such as clicking and dragging. Finally, the live-action AR image data can be arranged in chronological order to obtain a farmland AR live-action video stream, which is then pushed to the cloud server in real time to assist in decision-making, display, and information confirmation.

[0085] In an alternative embodiment, the feature extraction model is trained in the following manner: Extract a preset number of farmland AR image frames from the farmland AR live-action video stream; perform data augmentation on the farmland AR image frames and label the key regions to obtain sample farmland AR image frames; construct a sample data set; the sample data set includes sample farmland AR image frames and corresponding true labels; the true labels include any one of true farmland feature data, true device location information, and the true working conditions of external status lights; input the sample farmland AR image frames into a convolutional neural network model for feature extraction to obtain sample feature extraction results; based on the loss between the sample feature extraction results and the true labels, adjust the parameters of the convolutional neural network model to obtain a pre-trained feature extraction model.

[0086] A feature extraction model can be deployed on the local monitoring server. In some embodiments, a pre-trained CNN model can be used to extract visual features in video frames, and the filters automatically learned through the convolutional layer can capture low-level features (such as edges and textures) and higher-level abstract features (such as shapes and patterns) in the image, thus completing feature extraction.

[0087] A video is composed of a series of consecutive frames. Therefore, in addition to spatial information, the relationships in the temporal dimension also need to be considered. In some embodiments, 3D CNN (Convolutional Neural Network) can be used to directly operate in both the temporal and spatial dimensions simultaneously, or combined with recurrent neural networks such as RNN (Recurrent Neural Network), LSTM (Long Short-Term Memory), and GRU (Gated Recurrent Unit) to capture temporal information. Multiple consecutive frames are stacked into a volume as the input of 3D CNN to retain temporal information, thereby completing temporal modeling.

[0088] In some embodiments, the original farmland AR live video stream can be sampled. A certain number of image frames are evenly extracted from the AR live pictures according to the locations marked by the field monitoring devices, and the farmland AR live images are processed through data augmentation techniques such as flipping, rotation, and scaling, so as to increase the generalization ability of the model. Finally, the extracted picture frames can be annotated with key points or regions to complete the preparation of sample data and obtain sample farmland AR image frames.

[0089] In some embodiments, the EfficientNet network architecture can be selected to build a model, and supervised learning is carried out using a large-scale dataset, and the network parameters are optimized through the backpropagation algorithm. The model is pre-trained on the dataset of one acquisition module, and then fine-tuned for the data of different acquisition modules (after fine-tuning, the first feature extraction model, the second feature extraction model, and the third feature extraction model are obtained respectively), thereby completing the construction and training of the model architecture.

[0090] In some embodiments, during performance evaluation and tuning, a dataset can be formed after preprocessing the video stream collected by a single data acquisition module. The model.evaluate() function can be used to obtain the loss value and accuracy, and more detailed metrics such as precision, recall, and F1-score can be calculated through the sklearn.metrics library. The model.predict() can be used to generate prediction results, and these results (i.e., the sample feature extraction results) can be compared with the true labels to adjust the model parameters and obtain a pre-trained feature extraction model. A confusion matrix can be constructed to show the prediction situation for each category, which helps to identify which categories are prone to confusion. The ROC (Receiver Operating Characteristic Curve) curve can be plotted to show the relationship between the true positive rate and the false positive rate at different thresholds. The classification_report can be used to generate a report containing the precision, recall, and F1-score for each category from sklearn.metrics. More data augmentation techniques, such as rotation, scaling, flipping, brightness change, etc., can be applied to improve the generalization ability of the model. Training can be stopped when the performance on the validation set no longer improves to avoid overfitting. If the model performs well on the training set but poorly on the validation set, it may be overfitting; if both performances are poor, it may be underfitting. For overfitting, a Dropout layer and regularization can be introduced, and for underfitting, the model complexity can be increased or more training data can be provided for tuning.

[0091] In some embodiments, in the convolution operation, it can be assumed that the input layer of the model is a color picture with a size of N×N pixels and 3 channels. After passing through a convolutional layer with 4 filters, 4 feature maps are finally output, and the size is the same as the input layer. At this time, there are 4 filters in the convolutional layer, each filter contains 3 kernels, and the size of each kernel is 3x3. Therefore, the number of parameters in the convolutional layer is: Paramsconv = 3×3×3×4.

[0092] The formula for calculating the number of parameters in each convolutional layer can be: params = C o × (k w × k h × C i + 1). Where C o represents the number of input image channels, k w and k h represent the size of the convolutional kernel, C oDenotes the number of output image channels (i.e., Filters in the convolutional kernel), and +1 denotes the bias. Note that when using Batch Normalization, the bias is not required, and the +1 term in the calculation formula is removed. The input image size is generally square, k w =k h =k, so the above formula can be simplified to: params = C o ×k 2 ×C i .

[0093] The computational volume formula for each convolutional layer can be: FLOPs = C i ×k 2 ×C o ×W × H. Among them, C i denotes the number of input image channels, C o denotes the size of the output channel convolutional kernel, and W and H denote the size of the output feature map.

[0094] Aiming at the problems in the prior art that the production cycle of the agricultural scenario is long, the number of sensing data is limited, and there is a lack of large-scale real and effective data for agricultural model training, the embodiments of the present invention construct a sample data set for model training by acquiring the farmland AR real-scene video stream and extracting a preset number of farmland AR image frames from it, which can ensure the authenticity and effectiveness of a large amount of sample data, thereby improving the effectiveness of the model output data, and further improving the accuracy of data analysis.

[0095] Figure 4 is the third flowchart of the farmland management method provided by the embodiments of the present invention. Refer to Figure 4 , in some embodiments, the intelligent farmland management platform deployed on the cloud server may include a farmland basic geographic information collection module, and the intelligent farmland management platform can be docked with the agricultural Internet of Things device platform, so as to locate the field monitoring devices in the farmland, and check whether the data uploaded by the field monitoring devices is correct through AR real-scene tags. If the comparison is incorrect, re-acquire information for farmland device positioning; if the comparison is correct, generate a farmland AR real-scene map, and upload the feature data obtained by AI analysis of the farmland AR real-scene map to the intelligent farmland management platform.

[0096] The intelligent farmland management platform can obtain the data uploaded by field monitoring devices such as soil moisture monitoring devices, meteorological monitoring devices, crop disease monitoring devices, and pest monitoring devices, and combine the AR real-scene tags uploaded by the local monitoring server (including feature information such as farmland feature data, device location information, and the real working conditions of external status lights), and through soil moisture, meteorology, and pest and disease big model monitoring and analysis, timely screen and warn of the adverse situations affecting crop growth.

[0097] In the case of soil moisture warning, through the agricultural data analysis and decision-making unit, the irrigation area corresponding to the warning information can be matched, an irrigation task can be established, and the irrigation task for the to-be-irrigated area can be executed through irrigation equipment. Meanwhile, through the AI analysis of the farmland AR live video stream, the coordinates of the plot where the irrigation task is being executed are extracted, and they are matched with the coordinates of the to-be-irrigated area to determine whether the coordinates are consistent. If the coordinates are consistent, the task execution is confirmed; if the coordinates are inconsistent, then the data uploaded by the field monitoring equipment is rechecked through the AR live tags to determine whether it is correct.

[0098] In the case of crop disease warning, through the agricultural data analysis and decision-making unit, the fertilization area corresponding to the warning information can be matched, a fertilization task can be established, and the fertilization task for the to-be-fertilized area can be executed through fertilization equipment (such as a plant protection UAV). Meanwhile, through the AI analysis of the farmland AR live video stream, the coordinates of the plot where the fertilization task is being executed are extracted, and they are matched with the coordinates of the to-be-fertilized area to determine whether the coordinates are consistent. If the coordinates are consistent, the task execution is confirmed; if the coordinates are inconsistent, then the data uploaded by the field monitoring equipment is rechecked through the AR live tags to determine whether it is correct.

[0099] In the case of pest warning, through the agricultural data analysis and decision-making unit, the pesticide spraying area corresponding to the warning information can be matched, a pesticide spraying task can be established, and the pesticide spraying task for the to-be-pesticide-sprayed area can be executed through pesticide spraying equipment (such as a plant protection UAV). Meanwhile, through the AI analysis of the farmland AR live video stream, the coordinates of the plot where the pesticide spraying task is being executed are extracted, and they are matched with the coordinates of the to-be-pesticide-sprayed area to determine whether the coordinates are consistent. If the coordinates are consistent, the task execution is confirmed; if the coordinates are inconsistent, then the data uploaded by the field monitoring equipment is rechecked through the AR live tags to determine whether it is correct.

[0100] The embodiments of the present invention have the following advantages: Based on the AI data stream comparison, the embodiments of the present invention automatically incorporate field monitoring equipment into the intelligent farm management platform for management. Meanwhile, through the real-time data stream capture and marking of different acquisition modules, a large amount of data can be provided for model training.

[0101] Through the comparison with AR live scenes, the AR live scene modeling period of the embodiments of the present invention is short, and it is updated in real time according to the physical space scene. One high-altitude camera can cover an area of about 10,000 mu in radius, and there is no need to deploy cameras at multiple points, thus greatly reducing the construction cost.

[0102] Through the AI data stream comparison, the embodiments of the present invention can automatically verify the linkage execution results of platform systems, saving the workload of manual on-site inspection.

[0103] Figure 5 It is a schematic structural diagram of the farmland management system provided by an embodiment of the present invention. Refer to Figure 5 , an embodiment of the present invention provides a farmland management system, including: A cloud server 510, configured to, when it is determined that operations need to be performed on a target plot, send an operation decision for the target plot to a target operation device; the operation decision is used to instruct the target operation device to execute an operation task corresponding to the operation decision; A local monitoring server 520, configured to input a farmland AR real-scene video stream into a pre-trained first feature extraction model to obtain first farmland data; the first farmland data includes first operation feature data for characterizing the operation execution situation; and upload the first operation feature data to the cloud server in real time; The cloud server 510 is further configured to compare and analyze the first operation feature data uploaded by the local monitoring server with second operation feature data uploaded by field monitoring devices, determine the operation execution situation of the target plot, and display the operation execution situation of the target plot in the farmland AR real-scene map.

[0104] In the embodiment of the present invention, the farmland AR real-scene video stream is input into the feature extraction model to obtain operation feature data, and the operation feature data uploaded by the local monitoring server is compared and analyzed with the operation feature data uploaded by the field monitoring devices in real time to determine the real-time operation execution situation of the target plot. The farmland feature data analyzed by AI is used to verify the validity of the data uploaded by the field monitoring devices, eliminating the need for manual data validity verification. Moreover, it can avoid the impact of data loss caused by time delay and link damage during data transmission through the 4G network on data validity, which is conducive to ensuring the validity of farmland data and thus improving the accuracy of operation execution situation analysis; by automatically determining the real-time operation execution situation of the target plot after sending the operation decision to confirm whether the decision has been executed, it can save the workload of manual on-site inspection, thereby improving the efficiency of farmland management.

[0105] Figure 6 It is a schematic architecture diagram of the intelligent farmland management platform deployed on the cloud server provided by an embodiment of the present invention. Refer to Figure 6 , the intelligent farmland management platform deployed on the cloud server may include a farmland basic geographic information collection module, a farmland environment and agricultural situation information collection module, a meteorological monitoring, forecasting and early warning collection module, an agricultural plant protection monitoring and early warning collection module, an agricultural data analysis and decision-making unit, a field green prevention and control control unit, a smart irrigation control and execution unit, a plant protection drone control and execution unit, etc.

[0106] The farmland basic geographic information collection module can be used to complete the calibration of the basic geographic information of the farmland, map and connect the physical plot marking information to the real - scene AR image, and correspond and bind the two - dimensional plane geographic information with the real - scene AR three - dimensional geographic information.

[0107] The farmland environment and agricultural situation information collection module can be used to classify the data after collecting soil moisture information, crop growth data, and pest and disease disaster data, and transmit it to the AR real - scene for display through the intelligent farmland management platform.

[0108] The meteorological monitoring, forecasting and early warning collection module can be used to collect meteorological environment data and upload it to the platform, and transmit it to the AR real - scene for display through the intelligent farmland management platform.

[0109] The agricultural plant protection monitoring and early warning collection module can be used to upload soil, meteorological, and pest and disease data to the platform. After monitoring and analyzing through the platform's soil moisture, meteorological, and pest and disease large models, it can screen and warn in time about the adverse situations affecting crop growth. The warning information is pushed to the management personnel and at the same time pushed to the corresponding IoT coordinate points on the AR real - scene map. The corresponding warning plot information can be called through the AR real - scene for manual verification.

[0110] The agricultural data analysis and decision - making unit can be used to analyze the data after manual verification according to the plant protection monitoring and early warning situation, and issue corresponding decision prescriptions through the algorithms in the agricultural model library for relevant management personnel to confirm.

[0111] The field green prevention and control unit can be used to remotely turn on the field green prevention and control equipment according to the reported pest situation, regularly upload the number of pests killed, provide the overall trend of pest control feedback to the agricultural data analysis and decision - making unit for data analysis, and assist in completing the formulation of the next step of decision - making.

[0112] The intelligent irrigation control and execution unit can be used to remotely execute the irrigation operation according to the prescription issued by the decision - making unit. After the information is manually verified to be correct, the operation instructions are synchronously uploaded and marked in the AR real - scene. After the operation is completed, the execution effect of the operation plot can be viewed through the AR real - scene to complete the task closed - loop.

[0113] The plant protection UAV control and execution unit can be used to remotely execute the UAV plant protection operation according to the prescription issued by the decision - making unit. After the information is manually verified to be correct, the operation instructions are synchronously uploaded and marked in the AR real - scene. After the operation is completed, the execution effect of the operation plot can be viewed through the AR real - scene to complete the task closed - loop.

[0114] Figure 7 An entity structure diagram of an electronic device is illustrated, such as Figure 7As shown, the electronic device may include: a processor 710, a communications interface 720, a memory 730, and a communication bus 740. Among them, the processor 710, the communications interface 720, and the memory 730 complete mutual communication through the communication bus 740. The processor 710 may call logical instructions in the memory 730 to execute a farmland management method, which includes: when it is determined that operations need to be performed on a target plot, sending an operation decision for the target plot to a target operation device; the operation decision is used to instruct the target operation device to execute an operation task corresponding to the operation decision; comparing and analyzing first operation feature data uploaded by a local monitoring server with second operation feature data uploaded by a field monitoring device to determine the operation execution situation of the target plot; the first operation feature data is feature data in first farmland feature data used to characterize the operation execution situation, and the first farmland feature data is obtained by inputting a farmland AR live video stream into a pre-trained first feature extraction model; displaying the operation execution situation of the target plot in a farmland AR live map.

[0115] In addition, when the logical instructions in the above-mentioned memory 730 can be implemented in the form of software functional units and sold or used as an independent product, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this 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 for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs, and other various media that can store program codes.

[0116] 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 farmland management method provided by the above-mentioned various methods. The method includes: when it is determined that operations need to be performed on a target plot, sending an operation decision for the target plot to a target operation device; the operation decision is used to instruct the target operation device to execute the operation task corresponding to the operation decision; comparing and analyzing the first operation characteristic data uploaded by a local monitoring server with the second operation characteristic data uploaded by a field monitoring device to determine the operation execution situation of the target plot; the first operation characteristic data is the characteristic data in the first farmland characteristic data used to characterize the operation execution situation, and the first farmland characteristic data is obtained by inputting a farmland AR real-scene video stream into a pre-trained first feature extraction model; displaying the operation execution situation of the target plot in the farmland AR real-scene map.

[0117] In another aspect, the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it is implemented to execute the farmland management method provided by the above-mentioned various methods. The method includes: when it is determined that operations need to be performed on a target plot, sending an operation decision for the target plot to a target operation device; the operation decision is used to instruct the target operation device to execute the operation task corresponding to the operation decision; comparing and analyzing the first operation characteristic data uploaded by a local monitoring server with the second operation characteristic data uploaded by a field monitoring device to determine the operation execution situation of the target plot; the first operation characteristic data is the characteristic data in the first farmland characteristic data used to characterize the operation execution situation, and the first farmland characteristic data is obtained by inputting a farmland AR real-scene video stream into a pre-trained first feature extraction model; displaying the operation execution situation of the target plot in the farmland AR real-scene map.

[0118] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0119] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the essence of the above technical solution, 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 for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0120] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A farmland management method, characterized in that: Applied to cloud servers, including: When it is determined that an operation needs to be performed on the target plot, an operation decision for the target plot is sent to the target operation equipment; the operation decision is used to instruct the target operation equipment to perform the operation task corresponding to the operation decision; Compare and analyze the first operation feature data uploaded by the local monitoring server with the second operation feature data uploaded by the field monitoring device to determine the operation execution status of the target plot; the first operation feature data is feature data in the first farmland feature data used to characterize the operation execution status, and the first farmland feature data is obtained by inputting the farmland AR real-scene video stream into the pre-trained first feature extraction model; The operation execution status of the target plot is displayed in the AR real-life picture of the farmland.

2. The farmland management method according to claim 1, characterized in that: In the case where it is determined that an operation needs to be performed on the target plot, before sending the operation decision for the target plot to the target operation equipment, the method further includes: Acquire first farmland characteristic data uploaded by a local monitoring server and second farmland characteristic data uploaded by a field monitoring device; Inputting the first farmland characteristic data and the second farmland characteristic data into a farmland data analysis model to obtain early warning information; the early warning information is data characterizing an adverse situation affecting the growth of crops; The target plot corresponding to the early warning information is determined on the AR real-life picture of the farmland.

3. The farmland management method according to claim 1, characterized in that: The method further comprises: When the field monitoring device is powered on, first device location information uploaded by the local monitoring server and second device location information uploaded by the field monitoring device are obtained; wherein the first device location information is obtained by inputting the farmland AR real-scene video stream into a pre-trained second feature extraction model; If the first device location information is consistent with the second device location information, the field monitoring equipment is marked on the farmland AR real scene map.

4. The farmland management method according to claim 1, characterized in that: The method further comprises: When inspecting each field monitoring device, obtain the device abnormality information uploaded by the target field monitoring device and the working condition of the external status light of the target field monitoring device uploaded by the local monitoring server; wherein the working condition of the external status light is obtained by inputting the farmland AR real-scene video stream into the pre-trained third feature extraction model; Compare and analyze the equipment abnormality information with the working condition of the external status light to determine whether there is an abnormality in the target field monitoring equipment; If the target field monitoring equipment has an abnormality, an equipment abnormality alarm is issued; the equipment abnormality alarm is used to remind the operation and maintenance personnel that there is an equipment abnormality.

5. A farmland management method, characterized in that: Applied to local monitoring servers, including: Get the farmland AR real-scene video stream; Inputting the farmland AR real-scene video stream into a pre-trained feature extraction model to obtain a feature extraction result; The feature extraction model includes a first feature extraction model; the inputting the farmland AR real scene video stream into the pre-trained feature extraction model to obtain a feature extraction result includes: When it is determined that the target plot needs to be operated, the farmland AR real-scene video stream is input into a pre-trained first feature extraction model to obtain first farmland feature data; the first farmland feature data includes first operation feature data for characterizing the operation execution status; The first operation characteristic data is uploaded to the cloud server; the first operation characteristic data is used to instruct the cloud server to compare and analyze the first operation characteristic data with the second operation characteristic data uploaded by the field monitoring equipment, determine the operation execution status of the target plot, and display the operation execution status of the target plot in the farmland AR real scene map.

6. The farmland management method according to claim 5, characterized in that: The feature extraction model further includes a second feature extraction model and a third feature extraction model; the step of inputting the farmland AR real scene video stream into the pre-trained feature extraction model to obtain a feature extraction result also includes: Inputting the farmland AR real-scene video stream into a pre-trained second feature extraction model to obtain first device position information of the field monitoring device; The farmland AR real-scene video stream is input into the pre-trained third feature extraction model to obtain the working status of the external status light of the field monitoring equipment.

7. The farmland management method according to claim 5, characterized in that: The obtaining of the farmland AR real-scene video stream includes: Acquire multiple farmland images taken by the farmland AR panoramic acquisition device at the same location at various times; The multiple farmland images are stitched together to obtain an initial farmland AR panoramic image; Post-processing the initial farmland AR panoramic image to obtain AR real-scene images of the farmland at various moments; The AR real scene images of the farmland at each moment are arranged in time sequence to obtain an AR real scene video stream of the farmland.

8. The farmland management method according to claim 5, characterized in that: The feature extraction model is trained in the following way: Extract a preset number of farmland AR image frames from the farmland AR real-scene video stream; Performing data enhancement on the farmland AR image frame and marking key areas to obtain a sample farmland AR image frame; Constructing a sample data set; the sample data set includes sample farmland AR image frames and corresponding real labels; The real label includes any one of real farmland feature data, real equipment location information and real working conditions of an external status light; Inputting the sample farmland AR image frame into a convolutional neural network model for feature extraction to obtain a sample feature extraction result; Based on the loss between the sample feature extraction result and the true label, the parameters of the convolutional neural network model are adjusted to obtain a pre-trained feature extraction model.

9. A farmland management system, characterized in that: include: The cloud server is used to send the operation decision for the target plot to the target operation equipment when it is determined that the target plot needs to be operated; The operation decision is used to instruct the target operation equipment to perform the operation task corresponding to the operation decision; A local monitoring server is used to input the farmland AR real-scene video stream into a pre-trained first feature extraction model to obtain first farmland data; the first farmland data includes first operation feature data for characterizing the operation execution status; and the first operation feature data is uploaded to a cloud server in real time; The cloud server is also used to compare and analyze the first operation characteristic data uploaded by the local monitoring server with the second operation characteristic data uploaded by the field monitoring equipment, determine the operation execution status of the target plot, and display the operation execution status of the target plot in the farmland AR real-life map.

10. 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, the farmland management method as described in any one of claims 1-4 or 5-8 is implemented.

11. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the farmland management method as described in any one of claims 1-4 or 5-8 is implemented.

12. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the farmland management method as described in any one of claims 1-4 or 5-8 is implemented.

Images