An agricultural planting carbon emission monitoring and management platform

Abstract

The application discloses an agricultural planting carbon emission monitoring management platform, and belongs to the technical field of carbon emission management. The platform comprises a multi-source data integration subsystem, a data processing subsystem, a data classification subsystem, a model construction subsystem and a management subsystem. The multi-source data integration subsystem is used for acquiring agricultural production data through distributed sensors and simultaneously acquiring image data of an agricultural planting target region. The data processing subsystem is used for pre-processing data to obtain a data set. The data classification subsystem is used for dividing an agricultural planting region into areas, generating a grid network of the planting region, and feeding the processed basic data set and the processed image data set into the grid network based on a spatial vector to perform feature-level fusion and generate a fusion data set. The model construction subsystem is used for constructing a carbon emission prediction model based on a deep feedforward neural network. The management subsystem is used for monitoring and managing agricultural planting carbon emission based on the carbon emission monitoring model.

Description

Technical Field

[0001] This invention belongs to the field of carbon emission management technology, and in particular relates to an agricultural planting carbon emission monitoring and management platform. Background Technology

[0002] As the impacts of global climate change intensify, the accurate monitoring and management of carbon emissions from agricultural activities, a significant source of carbon emissions, has become particularly crucial. Existing agricultural carbon emission monitoring technologies primarily rely on traditional data collection methods, such as manual field surveys and simple computational models. These methods are not only time-consuming and labor-intensive but also have limited accuracy. Furthermore, current technologies rarely apply advanced data processing techniques and machine learning algorithms to the real-time monitoring and prediction of carbon emissions, resulting in insufficient responsiveness and adaptability of agricultural carbon management strategies. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention proposes an agricultural planting carbon emission monitoring and management platform to resolve the issues present in the existing technologies.

[0004] To achieve the above objectives, the present invention provides an agricultural planting carbon emission monitoring and management platform, comprising:

[0005] The multi-source data integration subsystem is used to acquire agricultural production data through distributed sensors, and at the same time acquire image data of the target agricultural planting area;

[0006] A data processing subsystem, connected to the multi-source data integration subsystem, is used to preprocess the agricultural production data and the image data to obtain a basic processing dataset and a processed image dataset.

[0007] A data classification subsystem, connected to the data processing subsystem, is used to divide agricultural planting areas into regions, generate a grid network of planting areas, and feed the processing base dataset and the processed image dataset back to the internal and external factors of the grid network for feature-level fusion based on spatial vectors to generate a fused dataset.

[0008] The model building subsystem is connected to the data classification subsystem and is used to build a carbon emission prediction model based on a deep feedforward neural network.

[0009] The management subsystem is connected to the model building subsystem and the data classification subsystem respectively. It is used to input the fused dataset into the carbon emission prediction model for prediction, obtain carbon emission prediction results, feed the carbon emission prediction results back to the grid network of the planting area for labeling, generate a carbon emission monitoring model, and monitor and manage agricultural planting carbon emissions based on the carbon emission monitoring model.

[0010] Preferably, the multi-source data integration subsystem includes:

[0011] The sensor module is used to deploy distributed sensors based on the agricultural planting environment, acquire agricultural production data through the distributed sensors, and simultaneously feed back the spatial distribution of the distributed sensors to the agricultural production data.

[0012] The image acquisition module is used to acquire image data of the target area for agricultural planting based on the UAV, and at the same time, it feeds back the spatial distribution of the acquired images to the image data based on the UAV path planning route.

[0013] Preferably, the image acquisition module includes:

[0014] The information acquisition unit is used to acquire environmental information of the agricultural planting environment, parameter information of the depth camera on the UAV, weather information, flight strip overlap and flight strip width;

[0015] The flight path construction unit is used to construct the UAV flight path based on the environmental information of the agricultural planting environment, the parameter information of the depth camera on the UAV, the weather information, the flight path overlap and the flight path width.

[0016] The aerial photography unit is used to launch the drone and take aerial photos of the agricultural planting environment along the drone's flight path to obtain image data of the agricultural planting environment.

[0017] The feedback unit is used to spatially label the image data of the agricultural planting environment based on the flight path of the UAV.

[0018] Preferably, the data processing subsystem includes:

[0019] The data anomaly processing module is used to filter and delete outliers in the agricultural production data and the image data to obtain a filtered dataset.

[0020] The feature extraction module is used to label the features in the filtered dataset to obtain a labeled dataset;

[0021] The normalization module is used to normalize the data in the labeled dataset to obtain the processing base dataset and the processing image dataset.

[0022] Preferably, the data classification subsystem includes:

[0023] The spatial acquisition module is used to divide the agricultural planting area into regions based on the spatial distribution of the agricultural planting area and obtain a three-dimensional grid network of the planting area.

[0024] The cropping module is used to crop the processed image dataset into small-scale tiles to generate a tile dataset;

[0025] The association calculation module is used to analyze the association degree between the processing base dataset and the tile dataset, and generate the association between the data and the tiles;

[0026] The spatial calculation module is used to calculate the distribution of data in a spatial state based on the association between the data and the tiles, and to generate multiple data views;

[0027] The fusion module is used to acquire the internal and external factors of the grid network, and to fuse the multiple data views with the tile dataset based on the influence of the internal and external factors to obtain the fused dataset.

[0028] Preferably, the process by which the fusion module obtains the internal and external factors of the grid network includes: obtaining internal land factors based on the attributes of the land itself and the current land use status; obtaining external land factors based on the macro environment in which the land is located; quantifying the internal and external land factors to obtain quantitative indicators; determining the weights of the quantitative indicators based on the Delphi method; and performing multi-factor decision analysis on the quantitative indicators based on the weights to obtain the influence of internal and external factors.

[0029] Preferably, the model building subsystem includes:

[0030] The unit construction module is used to construct combined attention modules based on channel attention units and spatial attention units;

[0031] Network building blocks, used to construct deep feedforward neural networks;

[0032] An import unit is used to import the combined attention unit into the deep feedforward neural network to generate an improved feedforward neural network;

[0033] The training unit is used to train the improved feedforward neural network using a carbon emission sample set to obtain the carbon emission prediction model.

[0034] Preferably, the training unit uses the SaveReLU activation function to activate the improved feedforward neural network;

[0035] The calculation formula for the SaveReLU activation function is as follows:

[0036]

[0037] In the formula, e is a constant with a value range of [0, 0.5].

[0038] Preferably, the management subsystem includes:

[0039] The calculation and prediction module is used to input the fused dataset into the carbon emission prediction model for prediction and obtain carbon emission prediction results.

[0040] The labeling module is used to feed the carbon emission prediction results back to the grid network of the planting area for labeling, thereby generating a carbon emission monitoring model.

[0041] The monitoring module is used to monitor and manage carbon emissions from agricultural planting based on the carbon emission monitoring model.

[0042] Preferably, the monitoring module includes a data storage architecture; the data storage architecture is an XML architecture, in which the data acquisition time is arranged in a progressive order, and the acquisition time is used as a constraint condition, and the plant growth state parameters are used as a detailed description under the time condition. At the same time, the prediction results and monitoring results with the same acquisition time are stored accordingly according to the constraint condition.

[0043] Compared with the prior art, the present invention has the following advantages and technical effects:

[0044] This invention utilizes current and historical data, employing deep feedforward neural networks to predict crop growth and carbon emissions over a future period. This includes predicting carbon uptake and respiration during crop growth stages. Leveraging the predictive capabilities of neural networks, the platform can provide agricultural producers with optimized decision support. For example, based on future weather forecasts and soil nutrient status, it can recommend adjustments to irrigation strategies, fertilization programs, or planting cycles to maximize crop growth while simultaneously reducing carbon emissions. Attached Figure Description

[0045] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0046] Figure 1 This is a schematic diagram of the agricultural planting carbon emission monitoring and management platform software according to an embodiment of the present invention;

[0047] Figure 2 This is a hardware schematic diagram of an agricultural planting carbon emission monitoring and management platform according to an embodiment of the present invention;

[0048] Figure 3 This is a schematic diagram of a deep feedforward neural network according to an embodiment of the present invention. Detailed Implementation

[0049] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0050] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0051] Example 1

[0052] like Figure 1-2 As shown, this embodiment provides an agricultural planting carbon emission monitoring and management platform, including:

[0053] The multi-source data integration subsystem is used to acquire agricultural production data through distributed sensors, and at the same time acquire image data of the target agricultural planting area;

[0054] A data processing subsystem, connected to the multi-source data integration subsystem, is used to preprocess the agricultural production data and the image data to obtain a basic processing dataset and a processed image dataset.

[0055] A data classification subsystem, connected to the data processing subsystem, is used to divide agricultural planting areas into regions, generate a grid network of planting areas, and feed the processing base dataset and the processed image dataset back to the internal and external factors of the grid network for feature-level fusion based on spatial vectors to generate a fused dataset.

[0056] The model building subsystem is connected to the data classification subsystem and is used to build a carbon emission prediction model based on a deep feedforward neural network.

[0057] The management subsystem is connected to the model building subsystem and the data classification subsystem respectively. It is used to input the fused dataset into the carbon emission prediction model for prediction, obtain carbon emission prediction results, feed the carbon emission prediction results back to the grid network of the planting area for labeling, generate a carbon emission monitoring model, and monitor and manage agricultural planting carbon emissions based on the carbon emission monitoring model.

[0058] Furthermore, the multi-source data integration subsystem includes:

[0059] A sensor module is used to deploy distributed sensors based on the agricultural planting environment, acquire agricultural production data through the distributed sensors, and simultaneously feed back the spatial distribution of the distributed sensors to the agricultural production data. The sensor module specifically includes:

[0060] Soil moisture sensor: Measures soil moisture content or soil humidity to help determine irrigation needs and water management strategies. Temperature sensor: Detects soil and air temperature; this data is crucial for assessing the temperature sensitivity of crops during growth and development. Light sensor: Measures light intensity to help assess photosynthetic potential and light energy utilization efficiency, as well as determine the optimal planting location for plants. Air humidity sensor: Monitors air humidity levels to help predict the probability of pests and diseases and optimize irrigation and ventilation strategies. Wind speed and direction sensor: Measures wind speed and direction; this data is important for assessing the impact of cold winds or storms on crops and also helps optimize spraying and pesticide application. Soil pH sensor: Measures soil pH to help adjust soil acidity and alkalinity, thereby optimizing plant nutrient absorption and the growth environment. Electrical conductivity sensor: Detects electrical conductivity in the soil to assess soil salinity or nutrient concentration, aiding in precision fertilization. Atmospheric pressure sensor: Measures changes in atmospheric pressure, helping to predict weather changes, especially important for early warning and response to agricultural disasters. CO2 concentration sensor: Monitors the concentration of carbon dioxide in the air to help assess the rate of photosynthesis and growth status of plants.

[0061] The image acquisition module is used to acquire image data of the target area for agricultural planting based on the UAV, and at the same time, it feeds back the spatial distribution of the acquired images to the image data based on the UAV path planning route.

[0062] Furthermore, the image acquisition module includes:

[0063] The information acquisition unit is used to acquire environmental information of the agricultural planting environment, parameter information of the depth camera on the UAV, weather information, flight strip overlap and flight strip width;

[0064] The flight path construction unit is used to construct the UAV flight path based on the environmental information of the agricultural planting environment, the parameter information of the depth camera on the UAV, the weather information, the flight path overlap and the flight path width.

[0065] The aerial photography unit is used to launch a drone and take aerial photos of the agricultural planting environment along the drone's flight path to obtain image data of the agricultural planting environment; data from satellite images, drones or other aerial images, which can provide information about land cover, vegetation health status, etc.

[0066] The feedback unit is used to spatially label the image data of the agricultural planting environment based on the flight path of the UAV.

[0067] It should be noted that the atmospheric environmental data includes wind speed, wind direction, atmospheric stability, etc., and the preset standard refers to the numerical range corresponding to multiple carbon emission levels. This numerical range is used to evaluate the product of the first carbon emission assessment index and the second carbon emission assessment index. The preset standard is set by the user. Furthermore, the higher the carbon emission level, the higher the corresponding carbon diffusion rate.

[0068] Furthermore, the data processing subsystem includes:

[0069] The data anomaly processing module is used to filter and delete outliers in the agricultural production data and the image data to obtain a filtered dataset.

[0070] The feature extraction module is used to label the features in the filtered dataset to obtain a labeled dataset;

[0071] The normalization module is used to normalize the data in the labeled dataset to obtain the processing base dataset and the processing image dataset.

[0072] Furthermore, the data classification subsystem includes:

[0073] The spatial acquisition module is used to divide the agricultural planting area into regions based on the spatial distribution of the agricultural planting area and obtain a three-dimensional grid network of the planting area.

[0074] The cropping module is used to crop the processed image dataset into small-scale tiles to generate a tile dataset;

[0075] The association calculation module is used to analyze the association degree between the processing base dataset and the tile dataset, and generate the association between the data and the tiles;

[0076] The spatial calculation module is used to calculate the distribution of data in a spatial state based on the association between the data and the tiles, and to generate multiple data views;

[0077] The fusion module is used to acquire the internal and external factors of the grid network, and to fuse the multiple data views with the tile dataset based on the influence of the internal and external factors to obtain the fused dataset.

[0078] Furthermore, the process by which the fusion module obtains the internal and external factors of the grid network includes: obtaining internal land factors based on the attributes of the land itself and the current land use status; obtaining external land factors based on the macro environment in which the land is located; quantifying the internal and external land factors to obtain quantitative indicators; determining the weights of the quantitative indicators based on the Delphi method; and performing multi-factor decision analysis on the quantitative indicators based on the weights to obtain the influence of internal and external factors.

[0079] The influence of internal and external factors includes:

[0080] Internal Factors: Land Use and Soil Management: Crop Type and Planting Density: Different crop types have different growth cycles, root structures, and carbon uptake capacities, affecting their impact on soil organic matter. Fertilizer and Pesticide Use: Different fertilization and pesticide strategies have a direct impact on soil carbon storage and greenhouse gas emissions. Tillage Practices: Tillage depth, tillage frequency, and sowing methods affect soil carbon quality and emissions. Soil Properties: Organic Matter Content and Quality: Soil organic matter has a significant impact on carbon uptake, storage, and release. Soil Aeration and Water Management: These affect soil microbial activity and the rate of organic matter decomposition, thus influencing the carbon cycle. Crop Growth Stages: Growth Cycle and Seasonal Variation: Different stages of crop growth result in different rates of carbon uptake and release, requiring consideration of carbon cycling within the growth cycle.

[0081] External factors: Climate conditions: Precipitation and temperature: Affect crop growth and soil moisture, thus influencing the rate of soil carbon decomposition and carbon release. Sunshine duration and light intensity: Affect crop photosynthetic rate and carbon absorption capacity. Topography and geographical location: Elevation and slope: Affect hydrological processes and soil erosion, thereby influencing soil carbon distribution and loss. Geographical location: Different regional climate conditions and ecological environments have a significant impact on crop growth and soil carbon cycling. Human management measures: Irrigation and drainage management: Have a significant impact on the distribution and stability of soil moisture and organic matter. Land use change: Human activities such as agricultural practices, land cover change, and land reclamation alter the processes of soil carbon storage and release.

[0082] This method extracts all duplicate data from real-time agricultural carbon emission data to ensure data uniqueness. This helps improve the accuracy of data analysis and processing, avoiding errors caused by duplicate data. The technical solution considers not only adjacent duplicate data but also duplicate data spaced one unit apart, thus providing a more comprehensive assessment of the data's authenticity and validity. By setting upper and lower limits for the reference time difference and comparing the time difference calculated based on the timestamp with these reference values, it is possible to determine which duplicate data are genuine and meaningful, and which may be anomalies or errors. The use of a limit adjustment coefficient λ allows the upper and lower limits of the reference time difference to be adaptively adjusted according to actual conditions, improving the algorithm's flexibility and adaptability. When the time difference between two adjacent duplicate data exceeds the target reference time difference, it means that these two data points are likely meaningful and are therefore retained. When the time difference is within the target reference time difference, only the data with the later timestamp is retained, which helps ensure the data's timeliness and accuracy. By filtering out duplicate and invalid data, subsequent data processing and analysis become more efficient, reducing unnecessary computation and resource consumption.

[0083] Association rule mining can be used to analyze the impact of different environmental conditions, crop types, and management practices on carbon emissions. By mining large-scale historical data, potential correlations can be discovered, such as how specific weather patterns affect carbon uptake or crop growth rates.

[0084] like Figure 3 As shown, the model building subsystem further includes:

[0085] The unit construction module is used to construct combined attention modules based on channel attention units and spatial attention units;

[0086] Network building blocks, used to construct deep feedforward neural networks;

[0087] An import unit is used to import the combined attention unit into the deep feedforward neural network to generate an improved feedforward neural network;

[0088] The training unit is used to train the improved feedforward neural network using a carbon emission sample set to obtain the carbon emission prediction model.

[0089] Furthermore, the training unit uses the SaveReLU activation function to activate the improved feedforward neural network;

[0090] The calculation formula for the SaveReLU activation function is as follows:

[0091]

[0092] In the formula, e is a constant with a value range of [0, 0.5].

[0093] Furthermore, the management subsystem includes:

[0094] The calculation and prediction module is used to input the fused dataset into the carbon emission prediction model for prediction and obtain carbon emission prediction results.

[0095] The labeling module is used to feed the carbon emission prediction results back to the grid network of the planting area for labeling, thereby generating a carbon emission monitoring model.

[0096] The monitoring module is used to monitor and manage carbon emissions from agricultural planting based on the carbon emission monitoring model.

[0097] Furthermore, the monitoring module includes a data storage architecture; the data storage architecture is an XML architecture, in which the data acquisition time is arranged in a progressive order, and the acquisition time is used as a constraint condition, and the plant growth state parameters are used as a detailed description under the time. At the same time, the prediction results and monitoring results with the same acquisition time are stored accordingly based on the constraint condition.

[0098] Example 2

[0099] This embodiment provides an agricultural planting carbon emission monitoring and management platform, including:

[0100] The multi-source data integration subsystem is used to acquire agricultural production data through distributed sensors, and at the same time acquire image data of the target agricultural planting area;

[0101] A data processing subsystem, connected to the multi-source data integration subsystem, is used to preprocess the agricultural production data and the image data to obtain a basic processing dataset and a processed image dataset.

[0102] A data classification subsystem, connected to the data processing subsystem, is used to divide agricultural planting areas into regions, generate a grid network of planting areas, and feed the processing base dataset and the processed image dataset back to the internal and external factors of the grid network for feature-level fusion based on spatial vectors to generate a fused dataset.

[0103] The model building subsystem is connected to the data classification subsystem and is used to build a carbon emission prediction model based on a deep feedforward neural network.

[0104] The management subsystem is connected to the model building subsystem and the data classification subsystem respectively. It is used to input the fused dataset into the carbon emission prediction model for prediction, obtain carbon emission prediction results, feed the carbon emission prediction results back to the grid network of the planting area for labeling, generate a carbon emission monitoring model, and monitor and manage agricultural planting carbon emissions based on the carbon emission monitoring model.

[0105] Specifically:

[0106] This embodiment constructs a multi-source data integration subsystem, which can effectively acquire agricultural production data from distributed sensors, including various key indicators such as temperature, humidity, soil fertility, and light intensity. Simultaneously, the system can also acquire image data of the target agricultural planting area, providing a comprehensive information source for subsequent data analysis and processing.

[0107] Next, this data is passed to the data processing subsystem, which is primarily responsible for preprocessing the agricultural production data and image data. The preprocessing process includes data cleaning, format conversion, and standardization, aiming to eliminate noise and outliers in the data and improve its quality and reliability. After preprocessing, this embodiment obtains a basic processing dataset and a processed image dataset, which provide a solid foundation for subsequent data analysis and model building.

[0108] Subsequently, the data classification subsystem divides the agricultural planting area into regions, generating a raster network of these regions. This network divides the planting area into several small raster units, each containing specific agricultural information and environmental features. Based on spatial vectors, the processing base dataset and the processing image dataset are fed back into the internal and external factors of the raster network for feature-level fusion. This fusion method fully utilizes the complementarity between different data sources, improving the comprehensive utilization rate and accuracy of the data. Through feature-level fusion, this embodiment generates a fused dataset containing rich agricultural information and environmental features, providing strong support for subsequent model construction.

[0109] Next, the model building subsystem will construct a carbon emission prediction model based on a deep feedforward neural network. A deep feedforward neural network is a powerful machine learning model capable of automatically learning and extracting complex features from data and making effective predictions. Through training and optimizing the model, this embodiment can achieve accurate prediction of carbon emissions from agricultural planting. This prediction model not only helps this embodiment understand the current carbon emission situation but also predicts future carbon emission trends, providing important decision-making basis for the sustainable development of agricultural production.

[0110] Finally, the management subsystem inputs the fused dataset into the carbon emission prediction model for prediction, obtaining the carbon emission prediction results. These results are then fed back into the grid network of the planting area for labeling, generating a carbon emission monitoring model. Based on this monitoring model, this embodiment can monitor and manage agricultural planting carbon emissions in real time. When carbon emissions exceed a preset threshold, the system will promptly issue an alarm, reminding this embodiment to take corresponding intervention and adjustment measures. In this way, this embodiment can achieve effective control and management of agricultural planting carbon emissions, promoting the sustainable development of agricultural production.

[0111] The sensor module is used to deploy distributed sensors based on the agricultural planting environment, acquire agricultural production data through the distributed sensors, and simultaneously feed back the spatial distribution of the distributed sensors to the agricultural production data.

[0112] The image acquisition module is used to acquire image data of the target area for agricultural planting based on the UAV, and at the same time, it feeds back the spatial distribution of the acquired images to the image data based on the UAV path planning route.

[0113] Sensor modules are increasingly used in agricultural planting environments, and their key role is to enable the deployment of distributed sensors to accurately acquire agricultural production data. These sensors can be placed in various corners of farmland, providing strong data support for agricultural production by monitoring key parameters such as soil moisture, temperature, and light intensity in real time.

[0114] Specifically, the sensor module uses wireless communication technology to transmit data collected by various distributed sensors to the data processing center in real time. Simultaneously, these sensors can adjust their acquisition frequency and accuracy according to actual needs, adapting to different planting environments and crop growth stages. Furthermore, the sensor module can feed back the spatial distribution information of the distributed sensors to agricultural production data, enabling agricultural producers to more intuitively understand the growth status of different areas within the farmland, thereby developing more precise planting strategies.

[0115] Besides the sensor module, the image acquisition module also plays a crucial role in agricultural planting. This module is primarily based on drone technology, using high-definition cameras and other image acquisition equipment to acquire image data of the target agricultural planting area. The drone can flexibly fly over farmland, acquiring information such as crop growth status and pest and disease conditions from different angles and altitudes. During the image acquisition process, the image acquisition module can also plan routes based on the drone's path and feed back the spatial distribution information of the acquired images to the image data.

[0116] Firstly, the main responsibility of the data anomaly processing module in this embodiment is to rigorously review agricultural production data and image data, filtering out and deleting outliers. In the agricultural field, data often exhibits anomalies due to various factors such as the natural environment, equipment malfunctions, and human error. For example, temperature or soil moisture data may deviate from the normal range due to sensor malfunctions, while image data may appear blurry or distorted due to factors such as shooting angle, lighting conditions, or equipment performance. Through the data anomaly processing module, this embodiment can effectively remove these non-compliant data, ensuring the accuracy and reliability of subsequent analysis results.

[0117] The function of the feature extraction module is to label the features in the selected dataset for subsequent data processing and analysis. In agricultural production, various factors such as crop growth, soil conditions, and climate conditions affect yield and quality. Through the feature extraction module, this embodiment can transform these factors into specific numerical values ​​or indicators, thereby facilitating subsequent data analysis. Simultaneously, the feature extraction module can also perform in-depth mining and correlation analysis of specific features according to actual needs, revealing the potential patterns and value behind the data.

[0118] The main function of the normalization module is to normalize the data in the labeled dataset to eliminate differences in units between different data points, making the data more comparable and easier to process. In agricultural production, different data may have different units and ranges, such as temperature, humidity, and light intensity. These differences can lead to biases in data analysis and comparison. This embodiment uses the normalization module to transform these data to the same scale.

[0119] Furthermore, the data classification subsystem includes:

[0120] First, the spatial acquisition module, based on the spatial distribution characteristics of agricultural planting areas, uses Geographic Information System (GIS) technology to finely divide the planting areas and generate a three-dimensional raster network. Through this step, this embodiment can clearly understand the geographical location, shape, and area of ​​the planting areas.

[0121] The cropping module's function is to crop the processed image dataset into smaller-scale patches, generating a patch dataset. This module uses image processing techniques to segment and crop the original image to better extract and analyze detailed information about the planting areas.

[0122] The association calculation module is the core of the data classification subsystem. This module analyzes the association between the base dataset and the tile dataset to generate the association relationships between the data and the tiles.

[0123] In the spatial computing module, this embodiment further calculates the distribution characteristics of the data in spatial states based on the association between data and map tiles, specifically including:

[0124] Interpolation is performed on data between discrete sampling points to generate a continuous spatial surface. Commonly used interpolation methods include Kriging interpolation, inverse distance weighted interpolation (IDW), and spline interpolation. Interpolation can help fill in the gaps between sampling points and better understand the data distribution characteristics of the entire planting area.

[0125] Statistical methods (such as analysis of variance and spatial autocorrelation analysis) are used to assess the differences and correlations in data between different locations. These analyses can reveal the trends and degrees of difference in soil moisture, temperature, and other factors at different spatial locations.

[0126] The fusion module is responsible for combining multiple data views with the tile dataset to obtain a fused dataset. This module comprehensively considers the influence of both internal and external factors of the raster network, integrating data from different sources and formats through data fusion technology to form a comprehensive and accurate dataset. The fused dataset not only contains basic information about the planting area but also reflects the interactions and influences between various factors.

[0127] The process by which the fusion module obtains the internal and external factors of the grid network includes: obtaining internal land factors based on the land's inherent attributes and current land use status; obtaining external land factors based on the macro-environment in which the land is located; quantifying the internal and external land factors to obtain quantitative indicators; determining the weights of the quantitative indicators based on the Delphi method; and performing multi-factor decision analysis on the quantitative indicators based on the weights to obtain the influence of internal and external factors.

[0128] The model construction subsystem includes:

[0129] First, the unit building module is the core of the entire subsystem. It is primarily responsible for constructing a combined attention module based on channel attention units and spatial attention units. Channel attention units focus on the information interaction between different channels (such as different sensors or data dimensions), enhancing information transmission in key channels and suppressing the influence of secondary channels to improve model performance. Spatial attention units, on the other hand, focus on spatial information in images or feature maps, further improving model accuracy by focusing on key regions or features.

[0130] Channel Attention Unit: The input is a feature map, typically the output of a convolutional layer. Global pooling (such as average pooling or max pooling) is used to perform global statistics on each channel. Fully connected layers or similar mechanisms are used to learn the weights between channels, generating a channel attention weight vector. The features of each channel are multiplied by the corresponding attention weight to enhance useful channel feature responses. Spatial Attention Unit: The input is a feature map. A spatial attention weight map is typically generated through convolutional operations. The spatial attention weight map can be two-dimensional, with each value representing the importance of that location. The feature map is multiplied by the spatial attention weight map to enhance the feature responses at important spatial locations. The enhanced feature maps calculated by the channel attention and spatial attention units are then multiplied element-wise or summed element-wise.

[0131] The network building unit is responsible for constructing deep feedforward neural networks. A deep feedforward neural network is a multi-layered neural network that extracts useful feature representations from raw data by passing and transforming information layer by layer. In this subsystem, the network building unit selects appropriate network structures and parameters based on specific task requirements and data characteristics to construct a high-performance deep feedforward neural network.

[0132] The import unit introduces the combined attention unit into the deep feedforward neural network, generating an improved feedforward neural network. This step is crucial; by integrating the combined attention module into the network, the network can focus more on key information during training, thereby improving the model's predictive performance.

[0133] The training unit uses a carbon emission sample set to train an improved feedforward neural network, resulting in the final carbon emission prediction model. During training, the model continuously learns the patterns and regularities in the sample data, minimizing prediction errors by optimizing network parameters. Through numerous training iterations, the model gradually improves its prediction accuracy, ultimately forming a reliable carbon emission prediction model.

[0134] Furthermore, the training unit uses the SaveReLU activation function to activate the improved feedforward neural network;

[0135] The calculation formula for the SaveReLU activation function is as follows:

[0136]

[0137] In the formula, e is a constant with a value range of [0, 0.5].

[0138] Furthermore, the management subsystem includes:

[0139] The computational prediction module is the core engine of the entire management subsystem. Its primary task is to input the fused dataset into the carbon emission prediction model for prediction. This fused dataset consists of carefully collected, organized, and integrated data from various sources, including but not limited to soil quality, climate conditions, crop types, and planting methods in the planting area. This data is input into the carbon emission prediction model, and through the model's complex algorithms and computational processes, carbon emission prediction results are derived. These prediction results not only provide a quantitative estimate of carbon emissions from agricultural planting but also provide crucial information for subsequent labeling and monitoring efforts.

[0140] The labeling module plays a crucial role in translating prediction results into practical operational guidance. It is responsible for feeding back the carbon emission prediction results output by the calculation and prediction module into the grid network of the planting area for labeling. The key to this step is transforming abstract prediction data into intuitive and actionable labeling information. Through the processing of the labeling module, each grid cell is assigned a corresponding carbon emission level or label, thereby generating a detailed carbon emission monitoring model. This model clearly displays the carbon emission status of different areas within the planting area, providing strong support for subsequent monitoring and management work.

[0141] The monitoring module is the executor of the entire management subsystem. Based on a carbon emission monitoring model, it monitors and manages agricultural carbon emissions in real time. Through this module, managers can understand the carbon emission status of the planting area in real time, promptly identify potential carbon emission risks, and take corresponding intervention and adjustment measures. Furthermore, the monitoring module can continuously analyze and mine carbon emission data, providing a scientific basis for optimizing and improving agricultural planting.

[0142] Furthermore, the monitoring module includes a data storage architecture; the data storage architecture is an XML architecture, in which the data acquisition time is arranged in a progressive order, and the acquisition time is used as a constraint condition, and the plant growth state parameters are used as a detailed description under the time. At the same time, the prediction results and monitoring results with the same acquisition time are stored accordingly based on the constraint condition.

[0143] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An agricultural planting carbon emission monitoring and management platform, characterized in that, include: The multi-source data integration subsystem is used to acquire agricultural production data through distributed sensors, and at the same time acquire image data of the target agricultural planting area; The data processing subsystem is used to preprocess the agricultural production data and the image data to obtain a basic processing dataset and a processed image dataset. The data classification subsystem is used to divide agricultural planting areas into regions, generate a grid network of planting areas, and feed the processing base dataset and the processing image dataset back to the internal and external factors of the grid network based on spatial vectors for feature-level fusion to generate a fused dataset. The model building subsystem is used to build carbon emission prediction models based on deep feedforward neural networks. The management subsystem is used to input the fused dataset into the carbon emission prediction model for prediction, obtain carbon emission prediction results, feed the carbon emission prediction results back to the grid network of the planting area for labeling, generate a carbon emission monitoring model, and monitor and manage agricultural planting carbon emissions based on the carbon emission monitoring model.

2. The agricultural planting carbon emission monitoring and management platform according to claim 1, characterized in that, The multi-source data integration subsystem includes: The sensor module is used to deploy distributed sensors based on the agricultural planting environment, acquire agricultural production data through the distributed sensors, and simultaneously feed back the spatial distribution of the distributed sensors to the agricultural production data. The image acquisition module is used to acquire image data of the target area for agricultural planting based on the UAV, and at the same time, it feeds back the spatial distribution of the acquired images to the image data based on the UAV path planning route.

3. The agricultural planting carbon emission monitoring and management platform according to claim 1, characterized in that, The image acquisition module includes: The information acquisition unit is used to acquire environmental information of the agricultural planting environment, parameter information of the depth camera on the UAV, weather information, flight strip overlap and flight strip width; The flight path construction unit is used to construct the UAV flight path based on the environmental information of the agricultural planting environment, the parameter information of the depth camera on the UAV, the weather information, the flight path overlap and the flight path width. The aerial photography unit is used to launch the drone and take aerial photos of the agricultural planting environment along the drone's flight path to obtain image data of the agricultural planting environment. The feedback unit is used to spatially label the image data of the agricultural planting environment based on the flight path of the UAV.

4. The agricultural planting carbon emission monitoring and management platform according to claim 1, characterized in that, The data processing subsystem includes: The data anomaly processing module is used to filter and delete outliers in the agricultural production data and the image data to obtain a filtered dataset. The feature extraction module is used to label the features in the filtered dataset to obtain a labeled dataset; The normalization module is used to normalize the data in the labeled dataset to obtain the processing base dataset and the processing image dataset.

5. The agricultural planting carbon emission monitoring and management platform according to claim 1, characterized in that, The data classification subsystem includes: The spatial acquisition module is used to divide the agricultural planting area into regions based on the spatial distribution of the agricultural planting area and obtain a three-dimensional grid network of the planting area. The cropping module is used to crop the processed image dataset into small-scale tiles to generate a tile dataset; The association calculation module is used to analyze the association degree between the processing base dataset and the tile dataset, and generate the association between the data and the tiles; The spatial calculation module is used to calculate the distribution of data in a spatial state based on the association between the data and the tiles, and to generate multiple data views; The fusion module is used to acquire the internal and external factors of the grid network, and to fuse the multiple data views with the tile dataset based on the influence of the internal and external factors to obtain the fused dataset.

6. The agricultural planting carbon emission monitoring and management platform according to claim 1, characterized in that, The process by which the fusion module obtains the internal and external factors of the grid network includes: obtaining internal land factors based on the land's inherent attributes and current land use status; obtaining external land factors based on the macro-environment in which the land is located; quantifying the internal and external land factors to obtain quantitative indicators; determining the weights of the quantitative indicators based on the Delphi method; and performing multi-factor decision analysis on the quantitative indicators based on the weights to obtain the influence of internal and external factors.

7. The agricultural planting carbon emission monitoring and management platform according to claim 1, characterized in that, The model building subsystem includes: The unit construction module is used to construct combined attention modules based on channel attention units and spatial attention units; Network building blocks, used to construct deep feedforward neural networks; An import unit is used to import the combined attention unit into the deep feedforward neural network to generate an improved feedforward neural network; The training unit is used to train the improved feedforward neural network using a carbon emission sample set to obtain the carbon emission prediction model.

8. The agricultural planting carbon emission monitoring and management platform according to claim 7, characterized in that, The training unit uses the SaveReLU activation function to activate the improved feedforward neural network; The calculation formula for the SaveReLU activation function is as follows: In the formula, e is a constant with a value range of [0, 0.5].

9. The agricultural planting carbon emission monitoring and management platform according to claim 1, characterized in that, The management subsystem includes: The calculation and prediction module is used to input the fused dataset into the carbon emission prediction model for prediction and obtain carbon emission prediction results. The labeling module is used to feed the carbon emission prediction results back to the grid network of the planting area for labeling, thereby generating a carbon emission monitoring model. The monitoring module is used to monitor and manage carbon emissions from agricultural planting based on the carbon emission monitoring model.

10. The agricultural planting carbon emission monitoring and management platform according to claim 1, characterized in that, The monitoring module includes a data storage architecture; the data storage architecture is an XML architecture, in which data is stored in a progressive order of acquisition time, with acquisition time as a constraint, and plant growth state parameters as a detailed description under time, and prediction results and monitoring results with the same acquisition time are stored accordingly based on the constraints.

Images