Intelligent garden-oriented multi-element data perception terminal device system and processing method

Abstract

The application discloses a smart garden-oriented multi-element data sensing terminal device system and processing method. The system is composed of a central server and a group of terminal devices, and the terminal device contains four functional modules, namely, a data capture and image acquisition module, a central control and operation module, a wireless communication module and an independent power supply module. The method collects characteristic data such as soil temperature and humidity locally through sensors. Among them, the soil oxidation-reduction potential sensor adopts a molybdenum electrode to avoid the production of an oxidation film on the electrode surface and the adsorption of other impurities. The central server decides to initiate a federal learning task. Model training invitations are sent to each sensor. The sensor receives the model training information from the central server and decides whether to accept and feedback. After receiving the feedback, the central server distributes the training model to the sensors that agree to participate in the training, and performs the federal learning training process. The application significantly improves the accuracy of soil oxidation-reduction potential measurement and the durability of the electrode.

Description

Technical Field

[0001] This invention belongs to the field of intelligent sensing and relates to a multi-data sensing terminal device system and processing method that uses a single sensor to simultaneously capture characteristic data of soil temperature and humidity, nitrogen, phosphorus and potassium, pH value and soil redox potential, thereby enabling soil testing and planting suggestions. Background Technology

[0002] Sensor technology has been widely applied in data capture, image acquisition, and intelligent sensing. Among these, soil property sensors collect data such as temperature, humidity, electrical conductivity, nitrogen, phosphorus, and potassium by inserting probes into the soil, displaying the measured values ​​on a screen. However, currently, no soil sensor can collect data on the soil's redox potential. Furthermore, commercially available soil sensors only capture information and cannot provide early warnings about plant care based on the measured soil data. However, with the rapid development of IoT technology and the increasing prevalence of smart sensors, smart soil sensors can collect and store large amounts of soil property data locally, analyze various soil properties, and provide early warnings about whether plants need irrigation. However, if smart soil sensors only use locally trained models for prediction, this will affect the model's accuracy. Federated learning processes can utilize local data collected by multiple sensors for joint training, improving the accuracy of the prediction model. Therefore, how to use intelligent soil sensors to measure the redox potential of soil, apply federated learning processes to intelligent soil sensor systems, and design a multi-data perception and processing method and terminal equipment for smart gardens are urgent problems to be solved. Summary of the Invention

[0003] To overcome the shortcomings of existing technologies, this invention provides a multi-data sensing terminal device system and processing method for smart gardens, which can achieve high-precision measurement of soil redox potential, improve the accuracy of prediction models, and provide maintenance suggestions based on the actual soil conditions and corresponding plants.

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

[0005] This invention proposes a multi-data sensing terminal device system for smart gardens. The system consists of a central server and a set of terminal devices, which can be represented as N = {1,...,N}. The terminal devices include four functional modules: a data acquisition and image gathering module, a central control and operation module, a wireless communication module, and an independent power supply module. The data acquisition and image gathering module includes redox potential sensors, pH sensors, temperature sensors, humidity sensors, vision sensors, and nitrogen, phosphorus, and potassium sensors. Each of these sensors is connected to the external environment via electrode probes to capture and accurately calibrate various elements such as redox potential and moisture content from the soil. The central control and operation module mainly includes the system's main CPU, RAM, ROM memory, and sensor controllers. It processes and stores the data collected by the data acquisition and image gathering module. The wireless communication module enables data transmission between the central control and operation module and the data acquisition and image gathering module, such as a Bluetooth or WiFi module. The independent power supply module supplies power to the above three modules and includes solar panels, rechargeable batteries, and a power sensing and control chip.

[0006] Furthermore, in the aforementioned multi-data sensing terminal device system for smart gardens, the visual sensor employs a miniature camera to capture high-definition plant images. After receiving the images, the system's main CPU processor performs preliminary processing and compares them with image data stored in the image database in the ROM memory to determine the species of plant to be cultivated.

[0007] Furthermore, in the aforementioned multi-data sensing terminal equipment system for smart gardens, an independent power supply module provides energy to the system, comprising solar panels, rechargeable batteries, and a power control chip. The solar panels offer advantages such as reusability and small size. The power control chip uses an internal photoresistor to promptly acquire information about the surrounding sunlight conditions and controls the operating status and power management of the solar panels and rechargeable batteries. When sunlight is sufficient, the solar panels convert solar energy into electrical energy; a portion of this energy is used to support the normal operation of the system's terminal equipment, obtaining accurate data; the remaining portion is stored in the rechargeable battery for continuous power supply at night or on cloudy or rainy days, ensuring the sensors operate efficiently around the clock.

[0008] Furthermore, in the aforementioned multi-data sensing terminal equipment system for smart gardens, the soil redox potential sensor mainly consists of a three-electrode system, a voltage and current measurement chip, and an analog-to-digital converter. The three electrodes include: a molybdenum electrode, a saturated calomel electrode, and a silver-silver chloride electrode. The molybdenum electrode is used to avoid the formation of an oxide film on the electrode surface and the adsorption of other impurities.

[0009] The present invention also provides a method for performing multi-source data sensing and processing using the above-mentioned terminal device system, comprising the following steps:

[0010] S1: The sensor in the terminal device collects information on soil temperature, humidity, nitrogen, phosphorus, potassium, pH value and plant species, and then adjusts the molybdenum electrode depth and the spacing between the molybdenum electrode and the saturated calomel electrode of the redox potential measurement sensor according to the soil conditions.

[0011] S2: Soil redox potential Eh was determined using a depolarization method;

[0012] S3: The system's main CPU processor corrects the directly measured soil redox potential Eh based on the soil pH value, the soil temperature at the time of measurement, and the daily average soil temperature.

[0013] S4: The system's main CPU processor processes and stores the data collected by the sensors;

[0014] S5: The central server decides to initiate a federated learning task and sends model training invitations to each terminal device. Each terminal device receives the model training information from the central server, decides whether to accept it and provides feedback. After receiving the feedback, the central server distributes the training model to the terminal devices that agree to participate in the training and carries out the federated learning training process.

[0015] S6: After the federated learning process is completed, the central server will distribute the trained model to each terminal device, and the terminal devices will provide maintenance suggestions for the cultivated plants based on the trained model.

[0016] Furthermore, in step S1 of the above processing method, the value range of the molybdenum electrode depth b is (0.2≤b≤0.4m), and the value range of the distance a between the molybdenum electrode and the saturated calomel electrode is (0.2≤b≤0.4m). <a≤0.5m)。

[0017] Furthermore, in step S2 above, the determination of the soil redox potential Eh using the depolarization method includes: firstly, the voltage-current measurement chip acquires the current I of the molybdenum electrode and the voltage ΔV between the molybdenum electrode and the saturated calomel electrode, which are then transmitted to the CPU processor via an analog-to-digital converter. Then, the system's main CPU processor calculates the soil resistivity using the formula... To obtain the minimum resistivity of the soil, the electrode system is used to measure the soil redox potential under the condition of minimum soil resistivity, where x = b / a. c = 0.9193 - 0.6122b + 0.8464b 2Finally, the system's main CPU processor sends the optimal values ​​of a and b to the controller, which then adjusts the specific positions of the measuring electrodes. Further, in step S3 above, the correction of the directly measured soil redox potential Eh based on soil pH, soil temperature at the time of measurement, and daily average soil temperature specifically includes: first, correcting Eh based on soil pH... Where R = 8.314471 J / (mol·K), F = 96485.3383 C / mol, and T is the Kelvin temperature of the soil at the time of measurement. Then, Eh... pH7 Based on soil temperature at the time of measurement and MST correction for daily average soil temperature, Eh corrected =Eh pH7 +[MST-(T-273.15)]×-7.58, to obtain the soil redox potential Eh for subsequent treatment. corrected .

[0018] In step S1, the soil temperature, humidity, nitrogen, phosphorus, potassium, pH value, soil redox potential, and plant species information collected by the sensors are processed by the system's main CPU and stored in the form of a one-dimensional vector x. The presence or absence of a significant change in soil characteristics detected by the sensors in the terminal device within a short period (i.e., irrigation or other maintenance measures taken by staff) is defined as y, where y is either 0 or 1. Therefore, the i-th local sample data of terminal device n is ultimately represented as... Stored locally on the terminal device in the form of [data / format], It is considered data. The labels correspond to the data. Further, in step S6, the terminal device calculates the parameter values ​​by combining the newly collected local data with the trained federated learning model. This indicates whether the plants need sprinkler irrigation under the current soil conditions.

[0019] The beneficial effects of this invention are:

[0020] 1. This method uses a molybdenum electrode to avoid the formation of an oxide film and the adsorption of other impurities on the electrode surface, thereby enhancing the electrode's durability and eliminating the need for manual electrode cleaning.

[0021] 2. Setting up the electrode system in the region of lowest soil resistivity improves the accuracy of subsequent measurements. The directly measured soil redox potential (Eh) is corrected based on soil pH, soil temperature during measurement, and daily average soil temperature. The corrected results better reflect the actual maintenance requirements of soil-grown plants.

[0022] 3. Federated learning processes can utilize local data collected from multiple sensors for joint training, effectively improving the accuracy of prediction models. Attached Figure Description

[0023] Figure 1 This is a diagram of the internal architecture of a terminal device based on federated learning.

[0024] Figure 2 A system diagram for federated learning.

[0025] Figure 3 This is a flowchart of a multi-data perception and processing method for smart gardens. Detailed Implementation

[0026] The present invention will now be described in detail with reference to the accompanying drawings.

[0027] This invention provides a multi-data sensing terminal device system and processing method for smart gardens, which enables high-precision measurement of soil redox potential and improves the accuracy of prediction models. Based on the actual soil conditions and corresponding plants, maintenance suggestions are provided.

[0028] Specifically, the aforementioned multi-data sensing terminal equipment system and processing method for smart gardens, such as... Figure 2 As shown, the system consists of a central server and a set of terminal devices, which can be represented as N = {1,...,N}. Each terminal device contains four functional modules, such as... Figure 1 As shown: This includes a data acquisition and image gathering module, a central control and operation module, a wireless communication module, and an independent power supply module. The data acquisition and image gathering module comprises a soil redox potential sensor, a pH sensor, a temperature sensor, a humidity sensor, a vision sensor, and a nitrogen, phosphorus, and potassium sensor. Each sensor is connected to the external environment via an electrode probe to capture relevant elements from the soil and perform precise calibration. The central control and operation module includes the system's main CPU, RAM, ROM memory, and a sensor controller, which processes and stores the data collected by the data acquisition and image gathering module. The wireless communication module enables data transmission between the central control and operation module and the data acquisition and image gathering module. The independent power supply module provides power to all three modules.

[0029] Specifically, the multi-data sensing terminal equipment system and processing method for smart gardens uses a miniature camera as the visual sensor to capture high-definition plant images. After receiving the high-definition plant images, the system's main CPU processor performs preliminary processing and compares and analyzes them with the image data stored in the ROM memory to determine the types of plants to be cultivated.

[0030] Specifically, the multi-data sensing terminal equipment system and processing method for smart gardens includes an independent power supply module comprising a solar cell or solar panel, a rechargeable battery, and a power sensing and control chip. The power sensing and control chip can obtain the light conditions around the soil in a timely manner through an internal photoresistor and control the working status and power management of the solar cell or solar panel and the rechargeable battery. When there is sufficient sunlight, the solar cell or solar panel converts solar energy into electrical energy. Part of this energy is used to support the normal operation of the terminal equipment and obtain accurate data, while the other part is stored in the rechargeable battery for continuous power supply at night or on cloudy or rainy days, so as to ensure the high-efficiency operation of the sensor around the clock.

[0031] Specifically, the aforementioned multi-data sensing terminal equipment system and processing method for smart gardens mainly includes the acquisition of soil characteristic information, adjustment of soil redox potential measurement sensor electrodes, measurement and correction of soil redox potential, and federated learning training process. The processing flow is as follows: Figure 3 As shown.

[0032] Specifically, the multi-data sensing terminal equipment system and processing method for smart gardens includes a soil redox potential sensor mainly composed of a three-electrode system, a CS5460A voltage and current measurement chip, and an analog-to-digital converter. The three electrodes include a molybdenum electrode, a saturated calomel electrode, and a silver-silver chloride electrode. The molybdenum electrode is used to avoid the formation of an oxide film on the electrode surface and the adsorption of other impurities. The specific measurement steps include:

[0033] Step 1: Adjust the position of the soil redox potential (Eh) measurement sensor electrodes;

[0034] Step 2: Determine the soil redox potential Eh using the depolarization method;

[0035] Step 3: Correct the directly measured soil redox potential Eh based on soil pH, soil temperature at the time of measurement, and daily average soil temperature.

[0036] Specifically, in the specific steps of soil redox potential measurement, in step 1, the intelligent sensor adjusts the molybdenum electrode depth b (0.2≤b≤0.4m) and the distance a between the molybdenum electrode and the saturated calomel electrode according to the soil conditions. <a≤0.5m)。

[0037] First, the CS5460A voltage and current measurement chip acquires the current I of the molybdenum electrode and the voltage ΔV between the molybdenum electrode and the saturated calomel electrode, and transmits them to the system's main CPU via an analog-to-digital converter.

[0038] Then, the system's main CPU uses the soil resistivity formula... The minimum resistivity of the soil is obtained, allowing the electrode system to measure the soil redox potential under conditions of minimum soil resistivity. Where x = b / a, c = 0.9193 - 0.6122b + 0.8464b 2 Finally, the system's main CPU sends the optimal values ​​of a and b to the controller, which then adjusts the specific positions of the measuring electrodes.

[0039] Specifically, in the specific steps of the soil redox potential measurement, in step 2, firstly, the controller adjusts the polarization voltage to 675mV, the silver-silver chloride electrode is used as the auxiliary electrode, and the molybdenum electrode is connected to the positive terminal of the power supply for a duration of t. 阳极 =Anodic polarization is performed for 10 seconds; then the polarization power supply is cut off to depolarize. The CS5460A chip records the molybdenum electrode potential E every 30 seconds for 6 minutes. 阳极 Registers stored inside the CPU.

[0040] Then, the saturated calomel electrode was subjected to cathodic polarization using this method, with a polarization time t. 阴极 =10s, similarly, record the depolarization E 阴极 Electrically located in registers inside the CPU.

[0041] Finally, the system's main CPU determines the polarization time t. 阳极 t 阴极 and E 阳极 E 阴极 Solve for E 阳极 =y1+z1logt 阳极 and E 阴极 =y2+z2logt 阴极 The soil redox potential is obtained by finding the intersection of the two linear equations.

[0042] Specifically, the specific steps for measuring soil redox potential (Eh) include step 3, where the depolarization method is used to determine the soil redox potential Eh. The CPU processor corrects the directly measured soil redox potential Eh based on the soil pH, soil temperature at the time of measurement, and daily average soil temperature. First, Eh is corrected based on the soil pH. Where R = 8.314471 J / (mol·K), F = 96485.3383 C / mol, and T is the Kelvin temperature of the soil at the time of measurement. Then, for Eh pH7 Based on soil temperature at the time of measurement and MST correction for daily average soil temperature, Eh corrected =Eh pH7 +[MST-(T-273.15)]×-7.58, to obtain the soil redox potential Eh for subsequent treatment. corrected .

[0043] Specifically, the federated learning process of the multi-data sensing terminal device system and processing method for smart gardens includes the following steps:

[0044] Step 1: The terminal device collects a large amount of characteristic data on soil temperature, humidity, nitrogen, phosphorus, potassium, pH value, and soil redox potential locally.

[0045] Step 2: The central server decides to initiate the federated learning task and sends model training invitations to each terminal device. The terminal devices receive the model training information from the central server, decide whether to accept, and provide feedback. Upon receiving feedback, the central server distributes the training model to the terminal devices that have agreed to participate in the training, initiating the federated learning training process.

[0046] Step 3: After the federated learning process is completed, the central server will distribute the trained model to each terminal device, which can then provide maintenance suggestions for the cultivated plants based on the trained model.

[0047] Specifically, in the federated learning process of the terminal device system, in step one, the soil temperature and humidity, nitrogen, phosphorus, potassium, pH value, soil redox potential, and plant species information collected by the sensors are processed by the system's main CPU and stored in the form of a one-dimensional vector x. Whether the terminal device senses that workers have performed irrigation or other maintenance measures (the sensing method is: a significant change in soil characteristics within a short period of time) is defined as y (y takes the value 0 or 1). Therefore, the i-th local sample data of terminal device n is ultimately... It is stored locally on the terminal device in the form of [format]. It is considered data. These are the labels corresponding to the data. In step three, the terminal device calculates the parameter values ​​by combining the newly collected local data with the trained federated learning model. This indicates whether the plants need sprinkler irrigation under the current soil conditions.

[0048] Specifically, the multi-data sensing terminal device system and processing method for smart gardens, in the federated learning process of the terminal device system, step two of the federated learning process for terminal device n specifically includes five steps, including:

[0049] Step a) The central server sends the model parameters of the task to the terminal device n;

[0050] Step b) Terminal device n trains the model locally. Terminal device n updates its local model using soil characteristic data collected locally. After multiple local iterations, terminal device n uploads the local model to the central server. The local iterations by terminal device n employ a mini-batch gradient descent method, and the local update process is as follows:

[0051]

[0052] in, The local model parameters are obtained by terminal device n in the (s-1)th local iteration during the t-th global iteration. For F(w) in The gradient value at point B is λ, where λ is the learning rate and B is the batch size selected in the mini-batch gradient descent algorithm.

[0053] Step c) The central server receives local model parameters uploaded from terminal devices within the system, performs global aggregation, and updates the global model. The process of updating the global model is as follows:

[0054]

[0055] Step d) The central server distributes the new global model to the terminal device n for data training in the next round of local iteration.

[0056] Step e) Repeat steps b), c), and d) until the trained global model converges, thus completing the entire federated learning training process. The goal of federated learning is to train a global model parameter vector w that minimizes the global loss function.

[0057]

[0058] in f(w,x) represents the sum of training data from all participating terminal devices in the system. nl ,y nl Given a model w, in the training data {x} nl ,y nl The loss function F on} n (w) is the local loss function of terminal device n.

[0059] The present invention and its embodiments have been described above illustratively, but this description is not restrictive. The figures shown are only one embodiment of the present invention, and the actual structure is not limited thereto. Therefore, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the present invention, such designs should fall within the protection scope of the present invention.

Claims

1. A method for multi-source data sensing and processing, the method utilizing a multi-source data sensing terminal device system for smart gardens, characterized in that, The system consists of a central server and a set of terminal devices, which can be represented as N = {1,...,N}. Each terminal device contains four functional modules: a data acquisition and image processing module, a central control and operation module, a wireless communication module, and an independent power supply module. The data acquisition and image processing module includes a soil redox potential sensor, a pH sensor, a temperature sensor, a humidity sensor, a vision sensor, and a nitrogen, phosphorus, and potassium sensor. Each sensor is connected to the external environment via an electrode probe to capture relevant elements from the soil and perform precise calibration. The soil redox potential sensor consists of a three-electrode system, a voltage and current measurement chip, and an analog-to-digital converter. The electrodes include: a molybdenum electrode, a saturated calomel electrode, and a silver-silver chloride electrode; the central control and operation module includes the system's main CPU, RAM, ROM memory, and sensor controller, which processes and stores the data acquired by the data capture and image acquisition module; the wireless communication module enables data transmission between the central control and operation module and the data capture and image acquisition module; an independent power supply module supplies power to the above three modules; the visual sensor uses a miniature camera to capture high-definition plant images. After receiving the high-definition plant images, the system's main CPU processor performs preliminary processing and compares and analyzes them with the image data stored in the ROM memory to determine the type of plant to be cultivated; The method includes the following steps: S1: The sensor in the terminal device collects information on soil temperature, humidity, nitrogen, phosphorus, potassium, pH value and plant species, and then adjusts the molybdenum electrode depth and the spacing between the molybdenum electrode and the saturated calomel electrode of the redox potential sensor according to the soil conditions. S2: Soil redox potential Eh was determined using a depolarization method; S3: The system's main CPU processor corrects the measured soil redox potential Eh based on the soil pH value, the soil temperature at the time of measurement, and the daily average soil temperature, obtaining the soil redox potential Eh for subsequent treatment. corrected ; S4: Soil temperature, humidity, nitrogen, phosphorus, potassium, pH value, and soil redox potential (Eh) collected by the sensor. corrected The plant species information is processed by the system's main CPU and stored as a one-dimensional vector x. Whether the sensors in the terminal device detect a significant change in soil characteristics within a short period (i.e., irrigation or other maintenance measures taken by staff) is defined as y, where y is either 0 or 1. Therefore, the i-th local sample data of terminal device n is ultimately represented as... Stored locally on the terminal device in the form of [data / format]. It is considered data. The labels corresponding to the data; S5: The central server decides to initiate a federated learning task and sends model training invitations to each terminal device. Each terminal device receives the model training information from the central server, decides whether to accept it and provides feedback. After receiving the feedback, the central server distributes the training model to the terminal devices that agree to participate in the training and carries out the federated learning training process. S6: After the federated learning process is completed, the central server distributes the trained model to each terminal device. The terminal devices then use the newly collected local data and the trained federated learning model to calculate the parameter values. This indicates whether the plants need sprinkler irrigation under the current soil conditions.

2. The method for multi-source data perception and processing according to claim 1, characterized in that, In step S1, the depth b of the molybdenum electrode ranges from 0.2 to 0.4 m, and the distance a between the molybdenum electrode and the saturated calomel electrode ranges from 0 to 1. <a≤0.5m。 3. The method for multi-source data sensing and processing according to claim 2, characterized in that, The method for determining the soil redox potential Eh using depolarization as described in S2 includes: first, a voltage-current measurement chip acquires the current I of the molybdenum electrode and the voltage ΔV between the molybdenum electrode and the saturated calomel electrode, which are then transmitted to the CPU processor via an analog-to-digital converter. Then, the system's main CPU processor calculates the soil resistivity using the formula... To obtain the minimum resistivity of the soil, the electrode system is used to measure the soil redox potential under the condition of minimum soil resistivity, where x = b / a. c = 0.9193 - 0.6122b + 0.8464b 2 Finally, the system's main CPU processor sends the optimal values ​​of a and b to the controller, which then adjusts the specific positions of the measuring electrodes.

4. The method for multi-source data sensing and processing according to claim 1, characterized in that, Step S3, which involves correcting the directly measured soil redox potential Eh based on soil pH, soil temperature at the time of measurement, and daily average soil temperature, specifically includes: first, correcting Eh based on soil pH. Where R = 8.314471 J / (mol·K), F = 96485.3383 C / mol, and T is the Kelvin temperature of the soil at the time of measurement. Then, Eh... pH7 Based on soil temperature at the time of measurement and MST correction for daily average soil temperature, Eh corrected =Eh pH7 +[MST-(T-273.15)]×-7.58, to obtain the soil redox potential Eh for subsequent treatment. corrected .

5. The method for multi-source data perception and processing according to claim 1, characterized in that, The independent power supply module includes a solar cell or solar panel, a rechargeable battery, and a power sensing and control chip. The power sensing and control chip can obtain the light conditions around the soil in a timely manner through an internal photoresistor, and control the working status and power management of the solar cell or solar panel and the rechargeable battery. When there is sufficient sunlight, the solar cell or solar panel converts solar energy into electrical energy. Part of the electrical energy is used to support the normal operation of the terminal equipment and obtain accurate data, while the other part is stored in the rechargeable battery for continuous power supply at night or on cloudy or rainy days, so as to ensure the sensor's high-efficiency operation around the clock.

Images