Soil moisture content prediction system

Abstract

The invention discloses a soil moisture content prediction system, which relates to the technical field of soil monitoring and comprises a small-core processor module, a large-core processor module and a sensor module, the large-core processor module is deployed with a large language model and a visual model; the sensor module collects environmental parameter data of a target soil area in real time; the small-core processor module stores the environment parameter data collected in real time to a local cache; the large-core processor module is used for synchronizing locally cached environment parameter data to a memory of the large-core processor module when the large-core processor module is powered on, the sensor module is controlled to collect an environment image, and the visual model analyzes the environment image to obtain image analysis text content; the large language model predicts the soil volumetric water content in a future set time period according to the input information; the input information comprises the image analysis text content and the environment parameter data at the corresponding moment. According to the invention, system power consumption can be reduced, and high-computing-power data analysis and timely prediction of soil volumetric water content can be realized.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of soil monitoring, in particular to a soil moisture condition prediction system. BACKGROUND

[0002] Soil moisture is surface water, ground water and underground water. In agricultural production, hydrological observation, meteorological observation stations and soil and water conservation monitoring stations, surface water is of great concern. Although there are methods for measuring soil volume water content, there is a lack of prediction of soil volume water content. With the increasing maturity of artificial intelligence (AI) technology, advantages have been highlighted in many application scenarios, making intelligent monitoring possible. At present, artificial intelligence is combined with hydrology, but artificial intelligence is applied to the server side, and some language models are used for human-computer interaction, which has problems such as "data lag, low analysis efficiency and passive decision-making".

[0003] In addition, from the perspective of the technical field, the existing hydrological monitoring technology usually adopts a technical scheme of a remote telemetry terminal (RTU) + a data communication unit (DTU) + a sensor. The sensor is, for example, a soil moisture meter or a rain gauge. Most RTUs use 8-bit / 16-bit or 32-bit single-chip microcomputers, some use bare machines, and some use micro-embedded systems such as RT-thread\rtos\ucos-ii. The single-chip microcomputer has the characteristics of single core and low power consumption. The scheduling method used by the operating system is multitasking, which allocates the priority of tasks through the method of interruption and time slice, and schedules the running time of each task through the task queue, rather than multi-process and multi-thread. The memory and Flash of the single-chip microcomputer of the RTU itself are relatively small, with a memory of 128 KB and a Flash of 512 KB. The storage of data needs to rely on external storage chips or TF cards. Therefore, when processing large data, it is limited, for example, a 2M picture cannot be processed due to the small memory. SUMMARY

[0004] The purpose of the application is to provide a soil moisture condition prediction system, which can reduce system power consumption, realize high-performance data analysis and timely prediction of soil volume water content.

[0005] To achieve the above purpose, the application provides the following solutions: The application provides a soil moisture condition prediction system, which comprises a small-core processor module, a large-core processor module and a sensor module; the large-core processor module is deployed with a large language model and a visual model; The sensor module is used for real-time collection of environmental parameter data of a target soil area; The small-core processor module is used for storing the real-time collected environmental parameter data to a local cache. The large core processor module is configured to synchronize locally cached environmental parameter data to the memory of the large core processor module when powered on, control the sensor module to collect an environmental image of a target soil region, control the visual model to analyze the environmental image to obtain image analysis text content, and control the large language model to predict the soil volume water content of the target soil region in a future set period based on input information, wherein the input information includes the image analysis text content and environmental parameter data at a corresponding time.

[0006] Optionally, the sensor module includes an air temperature and humidity sensor, an illumination intensity sensor, a wind speed and direction sensor, a camera, a rain gauge, and a TDR signal collector, wherein the TDR signal collector is configured to measure a time domain reflectance trace of the target soil region, and the time domain reflectance trace is used to determine the soil volume water content. The environmental parameter data includes air temperature, air humidity, illumination intensity, wind speed, wind direction, an environmental image, rainfall, and soil volume water content.

[0007] Optionally, the soil moisture prediction system further includes a human-computer interaction module, wherein the human-computer interaction module is configured to display the time domain reflectance trace and the prediction result of the large language model, and is further configured to input soil parameters of the target soil region, wherein the soil parameters include soil texture, dry bulk density, and field capacity, and the input information of the large language model further includes the soil parameters.

[0008] Optionally, the soil moisture prediction system further includes a communication module, wherein the communication module includes a USB communication interface and a UART communication interface, the small core processor module is connected to the communication module through the UART communication interface, and the large core processor module is connected to the communication module through the USB communication interface; when the communication module uses the UART communication interface, the USB communication interface is closed; and when the communication module uses the USB communication interface, the UART communication interface is closed.

[0009] Optionally, the large language model is a DeepSeek offline version. The small core processor module adopts an MCU processor, and the large core processor module adopts an ARM architecture processor.

[0010] Optionally, the large core processor module is further configured to determine a green color proportion in the environmental image, and the input information of the large language model further includes the green color proportion.

[0011] Optionally, the determination of the green color proportion in the environmental image specifically includes: converting the environmental image into an environmental image in HSV format; The pixels in the preset green pixel interval in the environment image in HSV format are converted to white, and the pixels outside the preset green pixel interval are converted to black, to obtain a black and white image; The proportion of white pixels to the overall pixels in the black and white image is taken as the green proportion.

[0012] Optionally, the soil moisture prediction system further comprises a dynamic power management module, the dynamic power management module is configured to dynamically turn on or turn off the power supply of each functional module according to a preset strategy, and the functional modules include a small core processor module, a large core processor module, a communication module and a sensor module.

[0013] Optionally, the soil moisture prediction system further comprises a remote monitoring platform, the remote monitoring platform is connected with the large core processor module, and the remote monitoring platform is configured to display the prediction result of the large language model, and perform software upgrading on the large language model and the visual model, remotely call a program running process log of the large core processor module, remotely modify a preset parameter in the program running process, and remotely call a time domain measurement trace.

[0014] Optionally, the future setting period is 24 hours, 7 days or 15 days in the future.

[0015] According to the specific embodiments provided in the application, the following technical effects are disclosed: The application provides a soil moisture prediction system, the small core processor module is configured to store the real-time collected environmental parameter data to a local cache; the large core processor module is configured to synchronize the environmental parameter data in the local cache to the memory of the large core processor module when powered on, control the visual model to analyze the environment image after the large core processor module collects a picture, and obtain image analysis text content; control the large language model to predict the soil volume water content of a target soil region in a future setting period according to input information; the input information includes the image analysis text content and the environmental parameter data at the corresponding moment; the small core processor module is configured to perform long-term real-time sensor data collection, the large core processor module is configured to perform soil volume water content prediction when the soil volume water content prediction result is needed, the system power consumption is reduced, high computing power data analysis can be realized, and the visual model and the large language model are used to realize timely prediction of the soil volume water content. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the application or the prior art, the following will briefly introduce the drawings needed in the embodiments. Obviously, the drawings in the following description are only some embodiments of the application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0017] Figure 1 A functional module schematic diagram of a soil moisture content prediction system provided by an embodiment of the present application is provided.

[0018] Figure 2 A comparison effect diagram of a soil volume water content prediction value and a soil volume water content measured value provided by an embodiment of the present application is provided. DETAILED DESCRIPTION

[0019] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0020] The above purposes, features and advantages of the present application can be more obvious and easy to understand. The present application will be further described in detail below with reference to the drawings and specific embodiments.

[0021] In an exemplary embodiment, the present application provides a soil moisture content prediction system, as shown in Figure 1 A soil moisture content prediction system includes a small core processor module, a large core processor module and a sensor module; the large core processor module is deployed with a large language model and a visual model.

[0022] The sensor module is configured to collect environmental parameter data of a target soil region in real time.

[0023] The small core processor module is configured to store the collected environmental parameter data in a local cache.

[0024] The large core processor module is configured to synchronize the environmental parameter data in the local cache to a memory of the large core processor module when powered on, control the sensor module to collect an environmental image of the target soil region, control the visual model to analyze the environmental image to obtain image analysis text content, and control the large language model to predict a soil volume water content of the target soil region in a future set period according to input information, the input information including the image analysis text content and environmental parameter data at a corresponding time.

[0025] The small core processor module and the large core processor module are core processors, the small core processor module adopts an MCU (micro control unit) processor, and the large core processor module adopts an ARM architecture processor. The ARM processor is equipped with a Linux operating system. The visual model adopts a visual model of the Alibaba Group. The small core processor module has low power consumption, and a working current is 5 mA (12 V), and is suitable for a long-term working environment. The large core processor module has strong computing power, a main frequency, a memory, and a hard disk, but has large power consumption, and a working current is 500 mA (12 V), and has large long-term power consumption. Therefore, the small core processor module and the large core processor module are combined to realize high computing power and low power consumption compatibility.

[0026] The small core processor module is a low-power integrated module, and the small core processor module is used for long-term real-time sensor data acquisition. The large core processor module is a high-computing-power processor module, and the small core processor module and the large core processor module work cooperatively, so that system power consumption can be reduced, and high-computing-power data analysis can be realized.

[0027] The application combines the source and destination of soil surface water, and predicts the change trend of soil surface water content through an artificial intelligence model.

[0028] The source of soil moisture includes rainfall, water in the air, artificial irrigation, and underground water from top to bottom. The loss of soil moisture includes evaporation, plant absorption, and water infiltration from top to bottom.

[0029] In combination with the above elements, if the infiltration and evaporation process of soil moisture is associated with parameters such as rainfall, air temperature and humidity, air flow, light intensity, soil texture, soil porosity, and soil underground water, the relationship between the elements can be calculated through synchronous monitoring, and the trend of soil moisture can be calculated, and the soil volume water content can be estimated.

[0030] The application introduces AI into hydrological monitoring, and the core significance lies in solving the pain points such as "data lag, inefficient analysis, inaccurate prediction, and passive decision-making" in traditional hydrological monitoring, improving the perception, analysis, prediction, and regulation ability of the hydrological process through intelligent means, and providing more scientific and efficient support for water resource management, disaster prevention and reduction, and ecological protection.

[0031] Through "perception intelligence, analysis automation, prediction precision, and decision optimization", the hydrological monitoring and management mode is reconstructed, and the mode is changed from "passive response to nature" to "active cognition, utilization, and regulation of nature". This not only can improve the ability of disaster prevention and reduction and protect water resource safety, but also can provide key technical support for coping with climate change and realizing sustainable development, and is an inevitable trend of future development of the hydrological industry.

[0032] In an exemplary embodiment, for the function of the visual model, the AI is asked to identify the type of crop in the picture, which growth stage it is in, and analyze the growth state from the leaves. For example, the environmental image of the target soil area is a sunflower image, and the image analysis text content output by the visual model includes: 1. This picture shows a sunflower, which is in the flowering stage in autumn, with yellow or orange flowers and black seeds. 2. From the picture, the sunflower is in the mature stage, the sunflower flowers have fully opened, and the seeds have also matured. It is ready for harvest. 3. The leaves and flowers of the sunflower look withered, which may be due to drought or lack of water. Sunflowers need adequate water to grow and bloom, and lack of water may cause the leaves and flowers to wither.

[0033] The environmental image of the target soil area is a wheat image, and the image analysis text content output by the visual model includes: 1. Identified as wheat; 2. From the picture, the wheat is in the growth stage; 3. The leaves are long and oval in shape, with jagged edges, and the color of the leaves is usually green or light yellow, and the growth state is good.

[0034] The environmental image of the target soil area is a long-range image, and the image analysis text content output by the visual model includes: This picture shows an orange plantation. According to the orange trees and fruits in the picture, it can be judged that these plants are in the growing season and the growth state is good.

[0035] In an exemplary embodiment, the soil moisture prediction system further comprises a dynamic power management module, which is used to dynamically turn on or turn off the power of each functional module according to a preset strategy, and the functional modules include small core processor module, large core processor module, communication module and sensor module.

[0036] For example, the preset strategy control includes taking 8 points as the benchmark, and controlling the sensor module to measure the environmental image and soil water content of the target soil area once every 6 hours, and monitoring the environmental parameters in real time. If the environmental variable changes exceed the preset threshold, a measurement increase mechanism is triggered, and the measurement density is modified, such as automatically adjusting the measurement of the environmental image and environmental parameter data of the target soil area to once every hour during rainfall, for timely prediction update.

[0037] Power dynamic management strategy: when the device is started for the first time, the default setting is used, for example, taking 8 points as the benchmark every day, and starting the system once every 6 hours. During this period, the MCU always monitors parameters such as rainfall, wind speed and direction, temperature, and light intensity. This part of the sensor is powered on, while the communication module and large core processor module are powered off.

[0038] When the change in rainfall is monitored, the monitoring density is increased, for example, the system is started once an hour, and the data changes on site are observed and reported in a timely manner. When the system is started, the large core processor module is powered on, the communication module is powered on, and all sensors are powered on. After all the measurements are completed, the system automatically shuts down, and the MCU controls the large core processor module and the communication module to power off.

[0039] The application controls the power supply of the communication module, the large core processor module, the sensor module and the peripheral device through the dynamic power management module, uses a low-power MCU processor, is responsible for low-power management and dynamic power management, and ensures long-time operation of the system in unattended conditions.

[0040] The large core processor module is connected with the communication module, the man-machine interaction module, the sensor module, the peripheral interface and the dynamic power management module, is used for realizing data processing and system control, adopts a multi-core ARM processor, has high computing power, high main frequency, large memory and large capacity hard disk, supports external computing power expansion module, and meets complex data processing requirements. The large core processor module comprises a memory, a hard disk and a multi-core ARM architecture processor.

[0041] The dynamic power management module is controlled by the small core processor module, dynamically manages the power supply of each functional module, and realizes energy-saving operation. For example, one picture is collected once an hour, the small core processor turns on the power of the large core processor module and the high-definition camera, the large core processor module is started (powered on), the high-definition camera is connected to collect environmental images, and the high-definition camera can be a 360 ball machine, a panoramic camera, a multiple variable focus lens or a gun machine.

[0042] The dynamic power management module controls the connection of the communication module, the large core processor module, the sensor module and the peripheral interface.

[0043] The soil moisture content prediction system further comprises a storage module, the storage module adopts on-board storage, and is used for storing an operating system, an AI model and collected data.

[0044] In one example embodiment, the sensor module comprises an air temperature and humidity sensor, an illumination intensity sensor, a wind speed and direction sensor, a camera, a rain gauge and a TDR (Time-Domain Reflectometry) signal collector, the TDR signal collector is used for measuring a time-domain reflection trace of a target soil region, and the time-domain reflection trace is used for determining the soil volume water content. Wherein, the time-domain reflection trace is an electromagnetic wave transmission process, the transmission speed of the electromagnetic wave in the target soil region can be obtained through the time-domain reflection trace, the dielectric constant of the soil can be calculated through the transmission speed of the electromagnetic wave, the relationship between the dielectric constant and the soil volume water content is established, and then the soil volume water content of the target soil region is determined.

[0045] The environmental image is a high-definition image. Since the small core processor module may cause memory overflow when collecting the high-definition image in real time, the camera collects the environmental image only after the large core processor module is powered on, thereby avoiding memory overflow.

[0046] The environmental parameter data includes air temperature, air humidity, light intensity, wind speed, wind direction, environmental image, rainfall, and soil volume water content.

[0047] The sensor module is connected with the large core processor module through a special interface and transmits monitoring data in real time.

[0048] The air temperature and humidity sensor monitors the temperature and humidity in the air. The higher the temperature, the faster the soil moisture evaporates, and the humidity also affects the evaporation of soil moisture.

[0049] The light intensity sensor measures the light intensity, which affects the amount of soil moisture evaporation. Through different light, the amount of soil moisture evaporation can be predicted.

[0050] The wind speed and direction sensor monitors the wind speed and direction, evaluates the influence of air circulation on soil moisture, and predicts future water loss through the relationship between water loss in the soil carried away by different wind speeds.

[0051] The high-definition camera collects environmental images for visual model analysis of vegetation coverage and types.

[0052] The rain gauge monitors rainfall to evaluate the supply of rainfall to soil moisture. Through rainfall analysis, the infiltration process and infiltration amount are analyzed.

[0053] The TDR signal collector measures the soil volume water content.

[0054] The soil moisture prediction system further includes a human-computer interaction module; the human-computer interaction module is used to display time domain reflection (TDR) traces and prediction results of the large language model, and is also used to input soil parameters of a target soil area, the soil parameters including soil texture, dry bulk density, and field capacity; input information of the large language model further includes the soil parameters. The time domain reflection trace is information (amplitude, phase) of reflected electromagnetic waves in the process of electromagnetic wave transmission in soil, similar to a ground penetrating radar. The reflected electromagnetic waves can display electromagnetic wave information like an oscilloscope, and amplitude and phase information can be obtained.

[0055] The human-computer interaction module includes an online weather forecast interface, which is used for online weather forecast.

[0056] The human-machine interface module also includes a screen, buttons, and indicator lights for users to set soil parameters and view system status. The module is also used to set configuration parameters, including the communication module's IP address, port number, device code, communication protocol, and measurement frequency.

[0057] The host computer for the human-computer interaction module can be a mobile phone, a personal computer (PC), or other terminal with a screen. It can display TDR traces and analyze the current state of the probe in the soil through the TDR traces. It can also intuitively display the current working status of the device through LED indicators and other means.

[0058] The soil moisture prediction system also includes peripheral interfaces, such as USB, LAN, and RS485 interfaces, which support the connection of various peripheral devices.

[0059] The soil moisture prediction system also includes a computing power expansion module and a storage expansion module, both of which are connected to the system via peripheral interfaces.

[0060] The soil moisture prediction system also includes a communication module, which is a high-speed wireless communication module capable of high-speed data transmission and remote control. The small-core processor module and the large-core processor module share the communication module. The communication module includes a USB communication interface and a UART communication interface. The large-core processor module is connected to the communication module via the USB communication interface, and the small-core processor module is connected to the communication module via the UART communication interface. When the communication module uses the UART communication interface, the USB communication interface is turned off; conversely, when the communication module uses the USB communication interface, the UART communication interface is turned off. The UART communication interface uses MOSFET isolation. The MOSFET is only turned on when the large-core processor module is powered on, allowing the pin voltage of the small-core processor module to be transmitted to the large-core processor module. Otherwise, it is turned off to prevent the small-core processor module from transmitting voltage to the large-core processor module via the UART communication interface when the large-core processor module is powered off.

[0061] The communication module can connect to the network for both small-core and large-core processor modules. Once connected, the measurement results collected by the sensor module can be synchronized to the server. It can also perform remote software upgrades, remote TDR trace retrieval, remote parameter modification, remote diagnostics, and remote weather forecast queries.

[0062] This application incorporates a communication module and uses a USB interface, achieving uplink and downlink speeds of up to 50M / S, allowing a 2MB image to be uploaded in just 1-3 seconds.

[0063] The large language model mentioned is DeepSeek, specifically the offline version of DeepSeek.

[0064] The large-core processor module is also used to determine the proportion of green in the environmental image, and the input information of the large language model also includes the proportion of green.

[0065] In one example embodiment, determining the percentage of green in an environmental image specifically includes: The environmental image is converted into an HSV format environmental image.

[0066] In an HSV format environment image, pixels within a preset green pixel range are converted to white, and pixels outside the preset green pixel range are converted to black, resulting in a black and white image. Since the edges of the black and white image regions may not be ideal, the segmentation result often contains unwanted interference. This is where erosion and dilation come in.

[0067] The ratio of white pixels to the total number of pixels in the black and white image is used as the green ratio. Specifically, this includes: 1) Convolving the structuring element with the black and white image to obtain the center pixel. The structuring element is customizable and can be a rectangle. The black and white image is divided into several squares, with the size of the rectangle equal to the size of one square. The rectangle can only contain 0 or 1. The rectangle is checked sequentially in the squares of the black and white image and marked as 0 or 1. 2) Marking the white outline. 3) Counting the pixels within the white outline and the total number of pixels. 4) Dividing the two pixel counts.

[0068] The acquired images are processed by a visual model deployed on a large-core processor module to analyze the types of vegetation in the captured images, their water requirements for growth, and to calculate the water absorption value of the vegetation from the soil in that environment. High-definition cameras capture photos, and the vegetation cover of a region is calculated by the percentage of green in a fixed area of ​​the image, thereby analyzing the soil water loss value under that type of vegetation in that area.

[0069] In one example embodiment, the soil moisture prediction system further includes a remote monitoring platform connected to the large-core processor module. The remote monitoring platform is used to display the prediction results of the large language model, upgrade the software of the large language model and the visual model, remotely retrieve the program execution log of the large-core processor module, remotely modify preset parameters during program execution, remotely retrieve time-domain measurement traces, and analyze the sensor status in the sensor module.

[0070] Preset parameters refer to the parameters of the equipment during field operation. These parameters include IP address, port number, telemetry station code (user-assigned ID), measurement interval, and operating parameters. Operating parameters include peak search threshold, measurement frequency, image capture time, and parameters used in the calculation process. These preset parameters are stored as variables in the device's Flash memory. Each device is unique, possessing a unique ID. The communication module connects to the server in real-time, and the interaction with the server includes the ID. At the factory, the device defaults to a default IP address and port number; upon installation in the field, these are replaced with the user's server IP address and port number.

[0071] For analyzing sensor status, such as TDR signal acquisition devices, the principle is similar to radar. The probe is inserted into the soil, invisible to the naked eye, but electromagnetic waves are transmitted through the probe and can return information carrying soil data, similar to ground-penetrating radar imaging. The generated traces can be used to analyze the probe's status in the soil, such as whether there are gaps, rodent damage, or the presence of plant roots and stones around the probe.

[0072] In one example embodiment, the future time period is the next 24 hours, the next 7 days, or the next 15 days. This application combines factors such as temperature and humidity, season, crops, and historical trends to establish a data model. Based on the current soil volumetric water content, it predicts the trend over the next 24 hours, 7 days, or even 15 days, with predictive results as follows: Figure 2 As shown.

[0073] In one example embodiment, the soil moisture prediction system of this application implements specific functions through the following steps 1-6.

[0074] Step 1: System initialization and power management.

[0075] The system is initialized, and a dynamic power management module is established. The small core processor module controls the dynamic power management module through GPIO (General-Purpose Input / Output). According to the preset strategy, the power supply of each functional module is dynamically turned on or off, and the power supply of each functional module is turned on when it is needed and turned off when it is not needed, so as to ensure low power consumption operation of the system.

[0076] A connection should be established between the large-core processor module and the small-core processor module to prevent the small-core processor module from losing power when the large-core processor module has not finished processing its work.

[0077] For sensors that require long-term operation, the small-core processor module collects data, while the large-core processor module synchronizes the data after power-on.

[0078] A Linux operating system is built on the large-core processor module and deployed on a multi-core ARM processor. The large language model and vision model depend on the Linux operating system.

[0079] When installing the hardware of a soil moisture prediction system, soil parameters are input through a human-computer interaction module, and the dry bulk density of the soil is obtained by measuring with a ring cutter.

[0080] Step 2: Sensor data acquisition and synchronization.

[0081] After the sensor collects the corresponding environmental parameters, it uploads them to the small core processor module in real time. The small core processor module continuously collects sensor data and stores it in the local cache. After the large core processor module is powered on, it synchronizes the data from the local cache through the communication interface and ensures the integrity and consistency of the data.

[0082] Step 3: Build an AI model and analyze the data collected in Step 2.

[0083] A Linux operating system, an offline version of DeepSeek, and a visual model were deployed on a large-core processor module. Based on existing data, the offline version of DeepSeek and the visual model were trained. The model summarized the patterns between various elements according to historical trends. After training, the collected sensor data was processed to analyze and predict the trends of environmental parameters and soil moisture changes. When training the offline version of DeepSeek, the image-derived text content (obtained through the visual model) and the corresponding environmental parameter data were used as input data, and the predicted soil volumetric water content was used as label data.

[0084] High-definition photos of the target soil area are captured by high-definition cameras. The type of crop and the growth stage of the crop are determined by visual models and large language models. Based on the different growth stages of the crop, the required soil moisture is determined.

[0085] Step 4: Based on the analysis and prediction results obtained in Step 3, further analyze the impact of environmental parameters on soil moisture.

[0086] The sensor module transmits environmental parameter data in real time, while the big data processor module uses AI models to analyze the data and further assess the impact of factors such as rainfall, air temperature and humidity, light intensity, wind speed and direction on soil moisture.

[0087] Step 5: Soil moisture prediction and early warning.

[0088] By combining historical and real-time monitoring data, a data prediction model (trained using the DeepSeek offline version) is established through multi-source data integration via a large-core processor module. This model calculates soil moisture change trends for the next 24 hours, 7 days, or even 15 days and issues early warnings based on preset thresholds. For example, if the original data prediction model assumed that an increase of 100mm in rainfall would lead to a 0.5% increase in moisture content, but later monitoring revealed that the actual increase could be as high as 1%, a new model or algorithm would be developed to correct this value.

[0089] Step 6: Data synchronization and remote control.

[0090] The large-core processor module synchronizes the prediction results and system status to the server through the communication module, supporting remote access and control, and enabling parameter modification and software upgrades.

[0091] This application implements dynamic power management through a small-core processor module, and processes data that cannot be processed by a large-core processor module after power failure, such as parameter modification and real-time online processing.

[0092] The high-performance processor module utilizes a multi-core ARM processor, capable of using eight or more cores with a clock speed exceeding 1GHz, more than 4GB of memory, and more than 64GB of onboard Flash memory. This performance far surpasses that of currently used RTU processors in hydrology. Employing a processor with NPU acceleration, its computing power is greater than or equal to 5 TOPS. External expansion can further enhance its computing power, enabling smooth deployment of an offline version of DeepSeek. Dynamic power-on of the high-performance processor module achieves a balance between high computing power and low power consumption. It can also connect to network cameras, PTZ cameras, bullet cameras, and panoramic cameras, supporting secondary development for focus adjustment and magnification of vegetation details. Each image can be up to 5MB in size, and a large visual model can be used to identify vegetation coverage, species, growth stage, and status within the image. The application's memory can cache more data, such as images, wind speed data, vegetation status, rainfall, air temperature and humidity, historical data, soil volumetric water content, TDR measurement traces, etc. This data can be synchronized to a large data model for in-depth data pattern mining.

[0093] This application will monitor all aspects of water sources and flow directions in hydrological monitoring, deeply explore the relationships between various elements, lay the foundation for soil moisture forecasting, and even provide scientific data support for water resource allocation. It can reduce floods and promote soil and water conservation, which is of great significance in many fields.

[0094] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0095] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A soil moisture prediction system, characterized in that, The soil moisture prediction system includes: a small-core processor module, a large-core processor module, and a sensor module; the large-core processor module is equipped with a large language model and a visual model. The sensor module is used to collect environmental parameter data of the target soil area in real time. The small-core processor module is used to store the real-time collected environmental parameter data into a local cache; The large-core processor module is used to synchronize locally cached environmental parameter data to the memory of the large-core processor module when powered on, control the sensor module to collect environmental images of the target soil area, control the vision model to parse the environmental images to obtain image parsing text content, and control the large language model to predict the soil volume water content of the target soil area in a future set time period based on the input information; the input information includes the image parsing text content and the environmental parameter data at the corresponding time.

2. The soil moisture prediction system according to claim 1, characterized in that, The sensor module includes an air temperature and humidity sensor, a light intensity sensor, a wind speed and direction sensor, a camera, a rain gauge, and a TDR signal acquisition device. The TDR signal acquisition device is used to measure the time-domain reflectometry trace of the target soil area, and the time-domain reflectometry trace is used to determine the soil volumetric water content. Environmental parameter data include air temperature, air humidity, light intensity, wind speed, wind direction, environmental images, rainfall, and soil volumetric water content.

3. The soil moisture prediction system according to claim 2, characterized in that, The soil moisture prediction system also includes a human-computer interaction module; the human-computer interaction module is used to display the temporal reflectance trace and the prediction results of the large language model, and is also used to input soil parameters of the target soil area, including soil texture, dry bulk density and field water holding capacity; the input information of the large language model also includes the soil parameters.

4. The soil moisture prediction system according to claim 3, characterized in that, The soil moisture prediction system also includes a communication module, which includes a USB communication interface and a UART communication interface. The small-core processor module is connected to the communication module through the UART communication interface, and the large-core processor module is connected to the communication module through the USB communication interface. When the communication module uses the UART communication interface, the USB communication interface is turned off; when the communication module uses the USB communication interface, the UART communication interface is turned off.

5. The soil moisture prediction system according to claim 1, characterized in that, The large language model is an offline version of DeepSeek. The small-core processor module uses an MCU processor, and the large-core processor module uses an ARM architecture processor.

6. The soil moisture prediction system according to claim 1, characterized in that, The large-core processor module is also used to determine the proportion of green in the environmental image, and the input information of the large language model also includes the proportion of green.

7. The soil moisture prediction system according to claim 6, characterized in that, Determining the percentage of green in an environmental image specifically includes: Convert the environmental image into an HSV format environmental image; In an HSV format environment image, pixels within a preset green pixel range are converted to white, and pixels outside the preset green pixel range are converted to black, resulting in a black and white image. The proportion of white pixels to all pixels in the black and white image is taken as the proportion of green pixels.

8. The soil moisture prediction system according to claim 1, characterized in that, The soil moisture prediction system also includes a dynamic power management module, which is used to dynamically turn on or off the power of each functional module according to a preset strategy. The functional modules include a small core processor module, a large core processor module, a communication module, and a sensor module.

9. The soil moisture prediction system according to claim 1, characterized in that, The soil moisture prediction system also includes a remote monitoring platform, which is connected to the big core processor module. The remote monitoring platform is used to display the prediction results of the big language model, upgrade the big language model and the visual model, remotely retrieve the program running log of the big core processor module, remotely modify the preset parameters during program running, and remotely retrieve time-domain measurement traces.

10. The soil moisture prediction system according to claim 1, characterized in that, The future time period is defined as the next 24 hours, the next 7 days, or the next 15 days.

Images