Rice growth image real-time acquisition method

Abstract

The application discloses a rice growth image real-time acquisition method, comprising the following steps: real-time data acquisition; data processing; data visualization and analysis; the application extracts key features of rice growth through multi-dimensional data acquisition and fusion, combined with image analysis and machine learning algorithm; real-time acquisition of field environment data and rice plant height information, based on a long short-term memory network model, extraction of time sequence characteristics of environment parameters and plant height parameters, capture of long-term dependence relationship, realization of accurate prediction of rice plant height; the prediction result can be visualized and analyzed through a heat map and the like, and the rice growth condition is intuitively displayed; through a large number of trained long short-term memory network models, accurate prediction of rice plant height can be realized, and in subsequent implementation and acquisition of rice growth images, artificial observation or regular shooting is not needed, and real-time prediction of rice growth is improved.

Description

Technical Field

[0001] This invention relates to the field of rice growth technology, specifically a method for real-time acquisition of rice growth images. Background Technology

[0002] Rice is one of the world's most important food crops, with more than half of the global population relying on it as their primary food source. Beyond being a staple food, rice has evolved over a long history and has played many economic roles contributing to humankind. One such role is rice bran, the outer layer of rice grains, which can be pressed for oil. Rice bran can also be pickled or even served as a standalone dish called stir-fried rice bran. According to scientific research, rice bran is also highly nutritious, containing 64% of the nutrients found in rice and over 90% of the nutrients essential for the human body. Besides rice bran oil, rice bran protein and its nutrients are also major areas of research for scientists.

[0003] In recent years, with the continuous increase in global food demand, precision agriculture management has played an important role in improving agricultural production efficiency and stability. Rice, as one of the world's major food crops, is significantly affected by environmental factors and management measures during its growth process. To achieve precision agriculture management, rapidly obtaining information on rice growth status is crucial. Traditional rice growth monitoring methods rely on manual observation or periodic photography, which suffers from insufficient time and accuracy.

[0004] Therefore, the present invention provides a method for real-time acquisition of rice growth images. Summary of the Invention

[0005] (a) Technical problems to be solved

[0006] To address the shortcomings of existing technologies, this invention provides a method for real-time acquisition of rice growth images. By introducing an LSTM model, the accuracy of rice plant height prediction can be significantly improved, precision agricultural management strategies can be optimized, long-term dependencies in time series can be captured, and multimodal data fusion, self-attention mechanisms and other technologies can be combined to ensure the reliability and practicality of the model, providing a scientific basis and decision support for agricultural production, thereby solving the technical problems recorded in the background art.

[0007] (II) Technical Solution

[0008] To achieve the above objectives, the present invention provides the following technical solution:

[0009] A method for real-time acquisition of rice growth images includes the following steps:

[0010] Real-time data acquisition: Environmental data and image data of rice plant height are collected through sensors, as well as management data are obtained;

[0011] Data processing: The collected image data is transmitted to the cloud or local server via wireless communication for image analysis and feature extraction, as well as preprocessing of environmental and management data;

[0012] Data visualization and analysis: Rice growth analysis, and then using data visualization tools to combine and analyze image data, management data and environmental data to obtain rice growth images.

[0013] Further, the steps for acquiring the environmental data and image data are as follows:

[0014] Divide the paddy field into areas, and divide each area into a square with a length and width of five meters.

[0015] Sensor installation: Install sensors within the divided square area, including environmental sensors such as light sensors, temperature sensors, humidity sensors, soil sensors, and image sensors.

[0016] Data acquisition: Light intensity, temperature, humidity and soil data in the field are recorded using light sensors, temperature sensors, humidity sensors and soil sensors. The height of rice plants in a square area is measured using an image sensor.

[0017] Sensor locations: Light, temperature, and humidity sensors are set up as single sensors for field measurements, while soil and image sensors are set up in each square area, or a drone carrying an image sensor is used to periodically collect data on the plant height of rice in the square area.

[0018] Furthermore, the image analysis and feature extraction steps are as follows:

[0019] Image preprocessing: Correcting the image, including brightness adjustment, color equalization and geometric correction, removing image noise using Gaussian filters or median filters, and unifying the image size to a fixed size;

[0020] Object detection: Use object detection algorithms to identify rice plants in images and analyze the outline of rice plants through edge detection;

[0021] Feature extraction: The height characteristics of rice are analyzed using shape feature extraction algorithms. For tilted rice, geometric correction angles are used to calculate the actual height.

[0022] Feature fusion: The average height of all the rice plants obtained is calculated, and the average rice plant height within the square area is taken.

[0023]

[0024] in, Let H represent the average rice plant height within the t-th square region, which represents the average rice plant height within that square region. N represents the total number of samples taken. i Let i represent the height of the rice plant in sequence i.

[0025] Furthermore, the preprocessing steps for the environmental data are as follows:

[0026] Data cleaning: Identify and remove outliers from environmental data, and then fill in missing and outliers through interpolation;

[0027] Data normalization: Standardize and normalize environmental data to scale the data range to between [0,1].

[0028] Data transformation: Convert environmental data into time series data and perform spatial transformation on rice plant height to facilitate the subsequent generation of heat maps or distribution maps.

[0029] Furthermore, the data fusion for the rice growth analysis is as follows:

[0030] The processed and normalized environmental data are fused together, and the rice plant height from the image data is combined to generate an input feature vector:

[0031] X t =[S t T t W t R t G t H t-1 ] T ;

[0032] Among them, X t Represented as a data feature vector, S t Represented as illumination data, T t Represented as temperature data, W t Represented as humidity data, R t Represented as soil data, G t Represented as management data, H t-1 T represents the growth height of the rice plants within the square area at the previous moment, and T represents time.

[0033] Soil data includes nitrogen content, phosphorus content, potassium content, and pH value;

[0034] The management data comes from farmers' production logs and includes information on water storage and fertilizer application.

[0035] Furthermore, the soil data is processed as follows:

[0036] The collected data R tThe weights are then used to merge the components:

[0037]

[0038] Among them, w j R represents the weight values ​​of soil elements in sequence j. j Let j represent the soil element in sequence j, and n represent the types of soil elements. This is represented by the bias caused by the microbial population in the soil;

[0039]

[0040] Where β0 represents the intercept term, β1, β2, β3, β4, β5, and β6 all represent regression coefficients, κ represents the error term, R1, R2, R3, and R4 represent nitrogen content data, phosphorus content data, potassium content data, and pH value, respectively, v1 represents the number of microbial species in the soil, and v2 represents the total number of microorganisms in the soil.

[0041] The effects of soil elements on microbial populations were analyzed using a linear regression model, and the influence of changes in soil elements on rice plant height was measured and controlled through microbial populations.

[0042] Furthermore, the management data is calculated as follows:

[0043]

[0044] Where φ0 represents the intercept term, φ1, φ2, φ3, and φ4 all represent regression coefficients, μ represents the error term, and D1 and D2 represent the water storage duration and fertilizer application amount, respectively.

[0045] Nonlinear regression is used to calculate and manage data on water storage duration and fertilizer application.

[0046] Furthermore, the rice growth analysis calculations are as follows:

[0047] H t =LSTMCell(S t ,T t W t ,R t G t H t-1 )+θ;

[0048] Where θ represents the intercept term, which is used to adjust the model's predicted values, similar to the intercept term in linear regression. H t The predicted height of the rice plant is represented by (S). t ,T t W t ,R t Gt H t-1 The input feature vector is represented as X. t =[S t T t W t R t G t H t-1 ] T ;

[0049] LSTMCell(W i U i V i );

[0050] Among them, W i U i V i These represent the input, hidden state, and output weight matrices, respectively.

[0051] Furthermore, the weight matrix W of the input gate i as follows:

[0052] i t =σ(W i ·X t +U i ·h t-1 +b i );

[0053] Among them, X t This represents the input feature vector, h. t-1 This indicates the hidden state from the previous step, b i This represents the bias term of the input gate, i t Let σ represent the output of the input gate, and let σ represent the activation function of the gate.

[0054] The input gate weight matrix is ​​used to calculate the linear combination between the input features and the hidden state, and determines the degree of influence of the input information on the hidden state.

[0055] Hidden gate weight matrix U i as follows:

[0056] h t =σ(W h ·X t +U i ·h t-1 +b h );

[0057] Among them, W h Represented as the weight matrix of the hidden gate, U determines the degree of influence of the input features on the hidden state. iRepresented as a weight matrix hidden to the hidden state, b captures the interaction between the input features and the hidden state. h Represented as the bias term for the hidden gate, h provides the initial linear combination result. t This is represented as the output of the hidden door;

[0058] The hidden gate weight matrix is ​​used to update the hidden state, determine the linear combination between the current input and the hidden state, and capture long-term dependencies.

[0059] Update hidden status:

[0060] h t =h t-1 ·i t +(1-i t )·h t

[0061] Output weight matrix V i as follows:

[0062] o t =σ(W o ·X t +U i ·h t-1 +b o );

[0063] Among them, W o Represented as the weight matrix of the output gate, U determines the degree of influence of the input features on the output value. i Represented as the weight matrix from the output to the hidden state, it captures the interaction between the input features and the hidden state, b o Represented as the bias term of the output gate, it provides the initial linear combination result, o t This is represented as the output value of the output gate;

[0064] The output gate weight matrix is ​​used to map the hidden state to the output features and determines the degree of influence of the hidden state on the output.

[0065] The final prediction output is as follows:

[0066] The LSTM model maps the hidden state to the actual predicted rice plant height using the probability value calculated by the output gate.

[0067]

[0068] in, This is expressed as a predicted value for rice plant height;

[0069] Compare the predicted rice plant height with the average rice plant height. By comparing and judging, the LSTM model can adjust its prediction of rice plant height, thereby improving the accuracy of rice plant height prediction.

[0070]

[0071] Where L(θ) represents the loss function, and N represents the total number of samples. Let be the average rice plant height within the t-th square region. This is expressed as a predicted value for rice plant height;

[0072] The model is tuned using a loss function:

[0073] θ←θ-η▽L(θ)

[0074] Where η represents the learning rate, θ represents the model parameters, i.e., the intercept term, which is used to adjust the model's predicted values, and L(θ) represents the loss function.

[0075] Furthermore, the prediction results of the LSTM model are visualized using a heatmap:

[0076] Data preparation: The predicted rice plant height output by the LSTM model is matched according to square regions to determine the coordinate information of the field and to define the color depth corresponding to different rice plant heights.

[0077] Data visualization tools: Use GIS tools or data visualization libraries to load the forecast results and geographic coordinate data into the visualization tool;

[0078] Heatmap generation: Draw a distribution map of fields based on geographic coordinates, add the predicted rice plant height as the third-dimensional data layer of the map, and fill the heatmap with color gradients based on the predicted rice plant height, with high-value areas displayed in red and low-value areas displayed in green. Add light data, temperature data, humidity data, soil data, management data, rice plant height, units of each parameter, time range, and labels for square areas to the heatmap;

[0079] Dynamic heatmap: Displays the change of rice plant height over time, aggregates and visualizes the prediction results by time dimension, sets up the dynamic update function of the heatmap, and shows the distribution of rice plant height at different time points.

[0080] (III) Beneficial Effects

[0081] This invention provides a method for real-time acquisition of rice growth images, which has the following beneficial effects:

[0082] In the early stages, by combining image sensors and multi-sensor data acquisition, real-time image acquisition and analysis of rice growth process was achieved, providing technical support for precision agricultural management. Through the acquisition and fusion of multi-dimensional data, combined with image analysis and machine learning algorithms, key features of rice growth were extracted.

[0083] First, the paddy field is divided into square areas of fixed size. Light sensors, temperature sensors, humidity sensors, and image sensors are deployed in each area to collect field environmental data and rice plant height information in real time. The collected data is transmitted to the cloud or local server via wireless communication technology for image preprocessing, target detection, and feature extraction. After standardization and normalization, the environmental data and image data are combined with management data, such as irrigation water volume and fertilizer application, to form an input feature vector for model training.

[0084] Based on a long short-term memory network model, the time-series features of environmental parameters and plant height parameters are extracted to capture long-term dependencies and achieve accurate prediction of rice plant height. The model dynamically updates the hidden state and outputs the predicted value through input gate, hidden gate and output gate mechanism. The prediction results can be visualized and analyzed through heat maps and other methods to intuitively show the rice growth status.

[0085] By using a large number of trained long short-term memory network models, accurate prediction of rice plant height can be achieved. When acquiring images of rice growth in subsequent implementations, it is not necessary to rely on manual observation or regular shooting, thus improving the real-time predictability of rice growth. Attached Figure Description

[0086] Figure 1 This is a schematic diagram of the steps of a method for real-time acquisition of rice growth images according to the present invention;

[0087] Figure 2 This is a schematic diagram of the heatmap generation steps in a method for real-time acquisition of rice growth images according to the present invention. Detailed Implementation

[0088] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0089] Please see Figure 1-2 This invention provides a method for real-time acquisition of rice growth images, comprising the following steps:

[0090] Real-time data acquisition: Environmental data and image data of rice plant height are collected through sensors, as well as management data are obtained;

[0091] Data processing: The collected image data is transmitted to the cloud or local server via wireless communication for image analysis and feature extraction, as well as preprocessing of environmental and management data;

[0092] Data visualization and analysis: Rice growth analysis, and then using data visualization tools to combine and analyze image data, management data and environmental data to obtain rice growth images.

[0093] In this embodiment, the preferred steps for acquiring environmental data and image data are as follows:

[0094] Divide the paddy field into areas, and divide each area into a square with a length and width of five meters.

[0095] Sensor installation: Install sensors within the divided square area, including environmental sensors such as light sensors, temperature sensors, humidity sensors, soil sensors, and image sensors.

[0096] Data acquisition: Light intensity, temperature, humidity and soil data in the field are recorded using light sensors, temperature sensors, humidity sensors and soil sensors. The height of rice plants in a square area is measured using an image sensor.

[0097] Sensor locations: Light sensors, temperature sensors, and humidity sensors are set up as single sensors for field measurements, while soil sensors and image sensors are set up in each square area, or a drone carrying an image sensor is used to collect rice plant height data in the square area at regular intervals.

[0098] It should be noted that by dividing the paddy fields, precise management can be achieved, and the divided paddy fields facilitate the installation of sensors and other equipment, enabling accurate data collection.

[0099] In this embodiment, preferably, the image analysis and feature extraction steps are as follows:

[0100] Image preprocessing: Correcting the image, including brightness adjustment, color equalization and geometric correction, removing image noise using Gaussian filters or median filters, and unifying the image size to a fixed size;

[0101] The formula for calculating a Gaussian filter is as follows:

[0102]

[0103] Where I′(x, y) is the pixel value at coordinate (x, y) in the filtered image data, I(x+i, y+j) is the pixel value at coordinate (x+i, y+j) in the original image data, G(i, j) is the weight value of the Gaussian kernel at (i, j), a and b are the radii of the Gaussian kernel in the x and y directions respectively, and a = b, and k is the error coefficient generated by the Gaussian kernel during calculation;

[0104] The formula for calculating the Gaussian kernel G(i,j) is as follows:

[0105]

[0106] The formula for calculating k is as follows:

[0107] Where i and j are the distances from the center of the Gaussian kernel, and σ is the standard deviation of the Gaussian kernel;

[0108] Object detection: Use object detection algorithms to identify rice plants in images and analyze the outline of rice plants through edge detection;

[0109] Feature extraction: The height characteristics of rice are analyzed using shape feature extraction algorithms. For tilted rice, geometric correction angles are used to calculate the actual height.

[0110] Feature fusion: The average height of all the rice plants obtained is calculated, and the average rice plant height within the square area is taken.

[0111]

[0112] in, Let H represent the average rice plant height within the t-th square region, which represents the average rice plant height within that square region. N represents the total number of samples taken. i Represented as the plant height of rice in sequence i;

[0113] It should be noted that by processing the image information of rice plant height, we can improve the acquisition of rice plant height data, and obtain the average value of rice plant height in the divided area, which facilitates rapid calculation and analysis.

[0114] In this embodiment, preferably, the preprocessing steps for the environmental data are as follows:

[0115] Data cleaning: Identify and remove outliers from environmental data, and then fill in missing and outliers through interpolation;

[0116] The calculation formula for the cleaning module is as follows:

[0117]

[0118] Where F is the outlier factor, D is the collected lithium battery data, and |z o | represents the number of samples in the o-th cluster, d(p,z) o ) represents the distance between sample p and the o-th cluster center, and z represents the distance between sample p and the o-th cluster center. o Let it be the o-th cluster center;

[0119] For distances far from the cluster center, the data is averaged from the center, and the results of the clustering analysis algorithm are used for further calculations.

[0120] That is Among them, z o Data information for cluster centers,

[0121] If the distance of the data information is greater than 3d, the data signal is determined to be abnormal data information;

[0122] Data normalization: Standardize and normalize environmental data to scale the data range to between [0,1].

[0123] Data transformation: Convert environmental data into time series data and perform spatial transformation on rice plant height to facilitate the subsequent generation of heat maps or distribution maps;

[0124] It should be noted that by cleaning, normalizing, and transforming the data, the information becomes more accurate, and the dimensionality of the data is reduced, making it easier to perform rapid calculations and analysis, and reducing computational burden.

[0125] In this embodiment, preferably, the data fusion for the rice growth analysis is as follows:

[0126] The processed and normalized environmental data are fused together, and the rice plant height from the image data is combined to generate an input feature vector:

[0127] X t =[S t T t W t R t G t H t-1 ] T ;

[0128] Among them, X t Represented as a data feature vector, S t Represented as illumination data, T t Represented as temperature data, W t Represented as humidity data, R t Represented as soil data, G t Represented as management data, Ht-1 T represents the growth height of the rice plants within the square area at the previous moment, and T represents time.

[0129] Soil data includes nitrogen content, phosphorus content, potassium content, and pH value;

[0130] The management data comes from farmers' production logs and includes water storage information and fertilizer application information.

[0131] It should be noted that fusing the processed environmental data makes it easier for the environmental data to be input into the LSTM model for analysis and processing in subsequent calculations.

[0132] In this embodiment, preferably, the soil data is processed as follows:

[0133] The collected data R t The weights are then used to merge the components:

[0134]

[0135] Among them, w j R represents the weight values ​​of soil elements in sequence j. j Let j represent the soil element in sequence j, and n represent the types of soil elements. This is represented by the bias caused by the microbial population in the soil;

[0136]

[0137] Where β0 represents the intercept term, β1, β2, β3, β4, β5, and β6 all represent regression coefficients, κ represents the error term, R1, R2, R3, and R4 represent nitrogen content data, phosphorus content data, potassium content data, and pH value, respectively, v1 represents the number of microbial species in the soil, and v2 represents the total number of microorganisms in the soil.

[0138] The effects of soil elements on microbial populations were analyzed using a linear regression model, and the influence of changes in soil elements on rice plant height was measured and controlled through microbial populations.

[0139] It should be noted that, in order to analyze the elements in the soil and combine the influence of elements on microorganisms, it is convenient to control and regulate the effect of soil elements on rice growth.

[0140] In this embodiment, preferably, the management data is calculated as follows:

[0141]

[0142] Where φ0 represents the intercept term, φ1, φ2, φ3, and φ4 all represent regression coefficients, μ represents the error term, and D1 and D2 represent the water storage duration and fertilizer application amount, respectively.

[0143] Nonlinear regression is used to calculate and manage data on water storage duration and fertilizer application.

[0144] It should be noted that by analyzing and processing the management data, it is easier to analyze the impact of water storage duration and fertilizer application on rice growth, and by integrating the management data, the subsequent computational burden can be reduced.

[0145] In this embodiment, preferably, the rice growth analysis calculation is as follows:

[0146] H t =LSTMCell(S t ,T t W t ,R t G t H t-1 )+θ;

[0147] Where θ represents the intercept term, which is used to adjust the model's predicted values, similar to the intercept term in linear regression. H t The predicted height of the rice plant is represented by (S). t ,T t W t ,R t G t H t-1 The input feature vector is represented as X. t =[S t T t W t R t G t H t-1 ] T ;

[0148] LSTMCell(W i U i V i );

[0149] Among them, W i U i V i These represent the input, hidden state, and output weight matrices, respectively.

[0150] It should be noted that using LSTM models to analyze the impact of environmental and management data on rice growth facilitates training the accuracy of LSTM models and can effectively improve the prediction and judgment of rice growth.

[0151] In this embodiment, preferably, the weight matrix W of the input gate... i as follows:

[0152] i t =σ(W i ·X t +U i ·h t-1 +b i );

[0153] Among them, X t This represents the input feature vector, h. t-1 This indicates the hidden state from the previous step, b i This represents the bias term of the input gate, i t Let σ represent the output of the input gate, and let σ represent the activation function of the gate.

[0154] The input gate weight matrix is ​​used to calculate the linear combination between the input features and the hidden state, and determines the degree of influence of the input information on the hidden state.

[0155] Hidden gate weight matrix U i as follows:

[0156] h t =σ(W h ·X t +U i ·h t-1 +b h );

[0157] Among them, W h Represented as the weight matrix of the hidden gate, U determines the degree of influence of the input features on the hidden state. i Represented as a weight matrix hidden to the hidden state, b captures the interaction between the input features and the hidden state. h Represented as the bias term for the hidden gate, h provides the initial linear combination result. t This is represented as the output of the hidden door;

[0158] The hidden gate weight matrix is ​​used to update the hidden state, determine the linear combination between the current input and the hidden state, and capture long-term dependencies.

[0159] Update hidden status:

[0160] h t =h t-1 ·i t +(1-i t )·h t

[0161] Output weight matrix V i as follows:

[0162] o t =σ(W o ·X t +U i ·h t-1 +b o );

[0163] Among them, W o Represented as the weight matrix of the output gate, U determines the degree of influence of the input features on the output value. i Represented as the weight matrix from the output to the hidden state, it captures the interaction between the input features and the hidden state, b o Represented as the bias term of the output gate, it provides the initial linear combination result, o t This is represented as the output value of the output gate;

[0164] The output gate weight matrix is ​​used to map the hidden state to the output features and determines the degree of influence of the hidden state on the output.

[0165] The final prediction output is as follows:

[0166] The LSTM model maps the hidden state to the actual predicted rice plant height using the probability value calculated by the output gate.

[0167]

[0168] in, This is expressed as a predicted value for rice plant height;

[0169] Compare the predicted rice plant height with the average rice plant height. By comparing and judging, the LSTM model can adjust its prediction of rice plant height, thereby improving the accuracy of rice plant height prediction.

[0170]

[0171] Where L(θ) represents the loss function, and N represents the total number of samples. Let be the average rice plant height within the t-th square region. This is expressed as a predicted value for rice plant height;

[0172] The model is tuned using a loss function:

[0173]

[0174] Where η represents the learning rate, θ represents the model parameters, i.e., the intercept term, which is used to adjust the model's predicted values, and L(θ) represents the loss function;

[0175] It should be noted that, through the above calculation and analysis of the input gate, hidden gate, and output gate, and combined with the actual rice growth height and the predicted rice growth height, the LSTM model is adjusted to improve the prediction accuracy of the LSTM model.

[0176] In this embodiment, preferably, the prediction results of the LSTM model are visualized using a heatmap:

[0177] Data preparation: The predicted rice plant height output by the LSTM model is matched according to square regions to determine the coordinate information of the field and to define the color depth corresponding to different rice plant heights.

[0178] Data visualization tools: Use GIS tools or data visualization libraries to load the forecast results and geographic coordinate data into the visualization tool;

[0179] Heatmap generation: Draw a distribution map of fields based on geographic coordinates, add the predicted rice plant height as the third-dimensional data layer of the map, and fill the heatmap with color gradients based on the predicted rice plant height, with high-value areas displayed in red and low-value areas displayed in green. Add light data, temperature data, humidity data, soil data, management data, rice plant height, units of each parameter, time range, and labels for square areas to the heatmap;

[0180] Dynamic heatmap: Displays the change of rice plant height over time, aggregates and visualizes the prediction results by time dimension, sets up the dynamic update function of the heatmap, and shows the distribution of rice plant height at different time points;

[0181] It should be noted that the visualization of heat maps makes it easy to intuitively show the growth status of rice.

[0182] Comparison Table of Parameters for Real-time Acquisition Methods of Rice Growth Images

[0183]

[0184]

[0185]

[0186] Comparison and summary:

[0187] Sensor parameters: More comprehensive in terms of the number and types of sensors, especially with the addition of soil microbial population detection function;

[0188] Data fusion parameters: For the first time, the fusion of multiple elemental parameters (nitrogen, phosphorus, potassium, pH value) was achieved, and combined with microbial population analysis, the utilization of soil information was enhanced;

[0189] Heatmap parameters: The color gradient of the heatmap is more detailed, the dynamic update frequency is higher, and the real-time monitoring is smoother;

[0190] Loss function and optimization: The model can be adjusted and controlled through the loss function. The loss function is more complex and supports more refined optimization.

[0191] The above comparison clearly demonstrates the significant advantages of my system in terms of multi-element parameter fusion, model complexity, and real-time monitoring.

[0192] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0193] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0194] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0195] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0196] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0197] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0198] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0199] The above description is merely a specific embodiment 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.

[0200] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes 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.

Claims

1. A method for real-time acquisition of rice growth images, characterized in that, Includes the following steps: Real-time data acquisition: Environmental data and image data of rice plant height are collected through sensors, and management data is obtained. Data processing: The collected image data is transmitted to the cloud or local server via wireless communication for image analysis and feature extraction, as well as preprocessing of environmental and management data; Data visualization and analysis: Rice growth analysis, and then using data visualization tools to combine and analyze image data, management data and environmental data to obtain rice growth images; The data fusion for the rice growth analysis is as follows: The processed and normalized environmental data are fused together, and the rice plant height from the image data is combined to generate an input feature vector: ； in, Represented as a data feature vector, Represented as illumination data, Represented as temperature data, Represented as humidity data, Represented as soil data, Represented as management data, This represents the growth height of the rice plants within the square area at the previous moment. Represented as time; Soil data includes nitrogen content, phosphorus content, potassium content, and pH value; The management data comes from farmers' production logs and includes water storage information and fertilizer application information. The soil data was processed as follows: Collected data The weights are then used to merge the components: ； in, Represented as a sequence The weight values ​​of soil elements, Represented as a sequence Soil elements, Represented as the types of soil elements. This is represented by the bias caused by the microbial population in the soil; ； in, Represented as the intercept term, All are expressed as regression coefficients. Represented as an error term, These are represented as nitrogen content data, phosphorus content data, potassium content data, and pH value, respectively. This represents the number of microbial species in the soil. This is expressed as the total number of microorganisms in the soil; The effects of soil elements on microbial populations were analyzed using a linear regression model, and the influence of changes in soil elements on rice plant height was measured and controlled through microbial populations.

2. The method for real-time acquisition of rice growth images according to claim 1, characterized in that: The steps for acquiring environmental and image data are as follows: Divide the paddy field into areas, and divide each area into a square with a length and width of five meters. Sensor installation: Install sensors within the divided square area, including environmental sensors such as light sensors, temperature sensors, humidity sensors, soil sensors, and image sensors. Data acquisition: Light intensity, temperature, humidity and soil data in the field are recorded using light sensors, temperature sensors, humidity sensors and soil sensors. The height of rice plants in a square area is measured using an image sensor. Sensor locations: Light, temperature, and humidity sensors are set up as single sensors for field measurements, while soil and image sensors are set up in each square area, or a drone carrying an image sensor is used to periodically collect data on the plant height of rice in the square area.

3. The method for real-time acquisition of rice growth images according to claim 1, characterized in that: The image analysis and feature extraction steps are as follows: Image preprocessing: Correcting the image, including brightness adjustment, color equalization and geometric correction, removing image noise using Gaussian filters or median filters, and unifying the image size to a fixed size; Object detection: Use object detection algorithms to identify rice plants in images and analyze the outline of rice plants through edge detection; Feature extraction: The height characteristics of rice are analyzed using shape feature extraction algorithms. For tilted rice, geometric correction angles are used to calculate the actual height. Feature fusion: The average height of all the rice plants obtained is calculated, and the average rice plant height within the square area is taken. ； in, Represented as the first The average height of rice plants within each square area represents the total height of rice plants within that square area. This represents the total amount sampled. Represented as a sequence The height of the rice plants.

4. The method for real-time acquisition of rice growth images according to claim 1, characterized in that: The preprocessing steps for the environmental data are as follows: Data cleaning: Identify and remove outliers from environmental data, and then fill in missing and outliers through interpolation; Data normalization: Standardize and normalize environmental data to scale the data range to between [0,1]. Data transformation: Convert environmental data into time series data and perform spatial transformation on rice plant height to facilitate the subsequent generation of heat maps or distribution maps.

5. The method for real-time acquisition of rice growth images according to claim 1, characterized in that: The management data is calculated as follows: ； in, Represented as the intercept term, All are expressed as regression coefficients. Represented as an error term, These are respectively expressed as water storage duration and fertilizer application amount; Nonlinear regression is used to calculate and manage data on water storage duration and fertilizer application.

6. The method for real-time acquisition of rice growth images according to claim 1, characterized in that: The rice growth analysis and calculations are as follows: ; in, This is represented as the intercept term, which is used to adjust the model's predicted values, similar to the intercept term in linear regression. This is represented as the predicted height of the rice plants. The feature vector of the input is represented as, i.e., as ; ； in, These represent the input, hidden state, and output weight matrices, respectively.

7. The method for real-time acquisition of rice growth images according to claim 6, characterized in that: Input gate weight matrix as follows: ； in, This indicates that it is the input feature vector. This indicates that it is the hidden state from the previous step. This indicates the bias term of the input gate. This is represented as the output of the input gate. This is represented as the activation function for gating; The input gate weight matrix is ​​used to calculate the linear combination between the input features and the hidden state, and determines the degree of influence of the input information on the hidden state. Hidden gate weight matrix as follows: ； in, Represented as the weight matrix of the hidden gate, it determines the degree of influence of the input features on the hidden state. Represented as a weight matrix hidden to the hidden state, it captures the interaction between the input features and the hidden state. This is represented as the bias term for the hidden gate, providing the initial linear combination result. This is represented as the output of the hidden door; The hidden gate weight matrix is ​​used to update the hidden state, determine the linear combination between the current input and the hidden state, and capture long-term dependencies. Update hidden status: ； Output weight matrix as follows: ； in, Represented as the weight matrix of the output gate, it determines the degree of influence of the input features on the output value. This is represented as a weight matrix from the output to the hidden state, capturing the interaction between the input features and the hidden state. This is represented as the bias term of the output gate, providing the initial linear combination result. This is represented as the output value of the output gate; The output gate weight matrix is ​​used to map the hidden state to the output features and determines the degree of influence of the hidden state on the output. The final prediction output is as follows: The LSTM model maps the hidden state to the actual predicted rice plant height using the probability value calculated by the output gate. ； in, This is expressed as a predicted value for rice plant height; Compare the predicted rice plant height with the average rice plant height. By comparing and judging, the LSTM model can adjust its prediction of rice plant height, thereby improving the accuracy of rice plant height prediction. ； in, Represented as a loss function, This represents the total amount sampled. Represented as the first The average height of rice plants within the square area This is expressed as a predicted value for rice plant height; The model is tuned using a loss function: ； in, This represents the learning rate. This is represented as a model parameter, specifically the intercept term. The intercept term is used to adjust the model's predicted values. This is represented as a loss function.

8. The method for real-time acquisition of rice growth images according to claim 7, characterized in that: The prediction results of the LSTM model are visualized using a heatmap: Data preparation: The predicted rice plant height output by the LSTM model is matched according to square regions to determine the coordinate information of the field and to define the color depth corresponding to different rice plant heights. Data visualization tools: Use GIS tools or data visualization libraries to load the forecast results and geographic coordinate data into the visualization tool; Heatmap generation: Draw a distribution map of fields based on geographic coordinates, add the predicted rice plant height as the third-dimensional data layer of the map, and fill the heatmap with color gradients based on the predicted rice plant height, with high-value areas displayed in red and low-value areas displayed in green. Add light data, temperature data, humidity data, soil data, management data, rice plant height, units of each parameter, time range, and labels for square areas to the heatmap; Dynamic heatmap: Displays the change of rice plant height over time, aggregates and visualizes the prediction results by time dimension, sets up the dynamic update function of the heatmap, and shows the distribution of rice plant height at different time points.

Images