Soil volumetric moisture content measurement system based on LoRa-RSSI and UAV

Abstract

This invention discloses a soil volumetric water content (VWC) measurement system based on LoRa-RSSI and a drone. The system comprises several underground LoRa nodes for collecting environmental monitoring data; a host computer embedded in the soil; and an aerial node carried by a small drone. Combining IoUT technology, this invention designs an innovative system for soil VWC measurement based on LoRa received signal strength and a drone. The system utilizes the changes in LoRa-RSSI between the soil's internal transmitter and the drone's aerial receiver during the drone's angular rotation. Combined with a Long Short-Term Memory (LSTM) network, it collects differential LoRa-RSSI values ​​and uses a deep learning (DL) algorithm to calculate soil VWC, achieving relatively accurate soil VWC data. This invention eliminates the need for depth measurement of VWC data, utilizes soft sensors to measure soil VWC, and offers low cost, high efficiency, and small measurement error, providing a novel approach for soil VWC measurement design.

Description

Technical Field

[0001] This invention relates to a soil volumetric moisture content measurement system. Background Technology

[0002] In agricultural planting, the root system of crops mainly depends on the soil environment.

[0003] Research indicates that soil VWC (Volumetric Water Content) has a significant impact on crop growth. Soil VWC is a crucial component of soil, significantly influencing root growth and the life activities of soil microorganisms. When soil VWC is too low, soil moisture content is low, hindering photosynthesis and reducing crop yield and quality. Severe water shortage can even lead to crop wilting and death. Conversely, excessively high soil VWC can worsen soil aeration, preventing normal vegetation growth. Therefore, maintaining an appropriate soil VWC level is essential for normal crop growth and development.

[0004] Traditionally, soil VWC measurement methods include soil moisture sensors, weighing methods, and standing wave ratio (SWR) methods. However, these traditional methods are often inconvenient for agricultural production. While soil moisture meters can greatly facilitate soil VWC measurement, widespread use of soil moisture sensors could impose a burden on farmers' economic interests, and sensor maintenance would also consume time in agricultural operations. Therefore, a convenient and accurate soil VWC measurement method is needed for agricultural production. Summary of the Invention

[0005] The technical problem to be solved by this invention is that: in the past, soil moisture monitoring mostly relied on various soil moisture sensors to obtain soil moisture data. These sensors not only require a certain amount of capital investment, but also are often subject to reliability and durability issues when buried underground for long-term use. In addition, since traditional soil moisture sensor probes need to be inserted into the cultivated layer, this also affects the agricultural farming process.

[0006] To achieve the above objectives, the technical solution of the present invention provides a soil volumetric moisture content measurement system based on LoRa-RSSI and a drone, characterized in that it includes:

[0007] Several underground LoRa nodes are used to collect environmental monitoring data. All environmental monitoring data collected by the underground LoRa nodes are transmitted to the host computer buried in the soil via LoRa-based underground Internet of Things.

[0008] The host computer, buried in the soil, receives environmental monitoring data transmitted from the underground LoRa nodes and sends LoRa data packets to the omnidirectional antenna of the air node via a flat panel antenna buried in the soil. During this process, the main lobe of the flat panel antenna is perpendicular to the horizontal plane, and the main lobe plane of the omnidirectional antenna is also perpendicular to the horizontal plane. The host computer also receives soil-air LoRa-RSSI values ​​sent from the air node, obtains the relationship between environmental monitoring data and RSSI and SNR based on the Friis formula, and uses LSTM to measure soil VWC. The Friis formula is shown below:

[0009] P RX =P TX +G TX +G RX -L UG -L UG-AG -0.5(L AG +L Surface )-L M -10log 10 (x 2 )

[0010] P RX It is soil RSSI, P TX It is the soil emission power output, G TX It is the gain of the transmitting antenna in the corresponding direction, G. RX It is the gain of the receiving antenna in the corresponding direction, L UG The loss is caused by the underground transmission of wireless signals, L UG-AG It is the refraction loss caused by the wireless signal passing through the soil-air interface, L AG L represents the loss caused by the transmission of wireless signals over the ground. Surface The attenuation is due to transverse waves, L M This is due to various losses from various possible sources, as well as miscellaneous losses due to the inability to satisfy the Friis equation, 10log 10 (x 2 This is path loss caused by multipath fading.

[0011] The air node is carried by a small drone and is used to receive LoRa data packets sent by the host computer through a receiving antenna and obtain the soil-air LoRa-RSSI value. Then, it sends this soil-air LoRa-RSSI value back to the host computer. By using the rotation of the drone, the air node is controlled to change the angle between the transmitting antenna in the soil and the receiving antenna of the air node, so that the air node obtains different soil-air LoRa-RSSI values ​​for the same soil VWC. This series of different soil-air LoRa-RSSI values ​​is sent back to the host computer, which uses the feature value sequence composed of soil temperature, drone rotation angle and RSSI at each angle at time T to calculate the soil VWC at that time.

[0012] Preferably, the underground LoRa node is an IoUT sensor node established using a LoRa module for the acquisition and transmission of environmental monitoring data, with a focus on monitoring soil temperature.

[0013] Preferably, the system's measurement target is soil VWC, which is an indirect measurement data.

[0014] Preferably, the host computer's antenna is a directional flat panel antenna, and the air node antenna is an omnidirectional antenna.

[0015] Preferably, the input LSTM data is processed into [[[dataA1],[dataT1],[dataR1]],[[dataA2],[dataT2],[dataR2]],[[dataA3],[dataT3],[dataR3]]…[[dataAn],[dataTn],[dataRn]]], and [dataV1],[dataV2],[dataV3]…[dataVn] data are selected as labels. After normalizing the data, the calculated value of soil VWC is output, where dataAn is the UAV rotation angle, dataTn is the environmental monitoring data, dataRn is the soil-air LoRa-RSSI value, and [dataVn] is the soil VWC.

[0016] Preferably, the root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination (R²) are selected. 2 As an evaluation metric for the LSTM:

[0017]

[0018]

[0019]

[0020] In the formula, a i p represents the true value. iThis represents the calculated value, where N represents the number of sample points. It is the average of the true values.

[0021] This invention combines IoUT technology to design an innovative system for soil volumetric water content (VWC) measurement based on LoRa Received Signal Strength Indication (RSSI) and a drone. The LoRa-based underground Internet of Things (IoT) has seen rapid development in recent years. This invention utilizes the characteristics of LoRa signals to indirectly measure soil VWC, significantly reducing the cost and maintenance of soil VWC measurement sensors and related expenses, achieving zero-cost functional reuse. Traditional RSSI-based VWC measurement methods require multiple ground base stations to obtain data for use by the LSTM method. In this system, a drone is used to rotate while hovering, collecting data from eight angles to obtain a feature value data sequence for use by a Long Short-Term Memory (LSTM) network. Differential LoRa-RSSI values ​​are collected, and the soil VWC is calculated using a deep learning (DL) algorithm, resulting in more accurate soil VWC data. This method not only has better real-time performance than traditional methods but is also simpler, has better error correction, requires less data acquisition time, and eliminates the need for multiple ground base stations, significantly reducing system complexity and engineering costs. Furthermore, this system significantly reduces measurement errors caused by lateral waves and obstacles such as vegetation on the soil surface during the transmission of electromagnetic signals through the soil surface when using ground base stations to collect RSSI radio signals. This invention eliminates the need for depth measurement of VWC data, effectively using soft sensors to indirectly measure soil VWC. This approach is low-cost, highly efficient, and has minimal measurement error, providing a completely new approach to soil VWC measurement design. Attached Figure Description

[0022] Figure 1 This illustrates the system framework of the present invention;

[0023] Figure 2 The system workflow is illustrated;

[0024] Figure 3 The data processing flow is illustrated;

[0025] Figure 4 This illustrates how the LSTM of the present invention is used;

[0026] Figure 5 A schematic diagram of the internal node structure of the soil;

[0027] Figure 6 This is a schematic diagram of the aerial node structure;

[0028] Figure 7This is a schematic diagram of the host computer structure;

[0029] Figure 8 The diagram illustrates the comparison between the calculated values ​​and the actual values ​​of 32 sets of data.

[0030] Figure 9 The absolute values ​​of error from 32 soil VWC measurements are shown. Detailed Implementation

[0031] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Furthermore, it should be understood that after reading the teachings of this invention, those skilled in the art can make various alterations or modifications to the invention, and these equivalent forms also fall within the scope defined by the appended claims.

[0032] This invention is based on the following design concept: Previous studies have shown that when a LoRa data terminal inside the soil communicates with a LoRa data terminal outside the soil, the RSSI value of the receiver can be obtained using the Friis formula, which is:

[0033] P RX =P TX +G TX +G RX -L UG -L UG-AG -0.5(L AG +L Surface )-L M -10log 10 (x 2 )

[0034] P RX It is the receiver RSSI, P TX It is the internal transmission power of the soil terminal, G. TX It is the gain of the transmitting antenna in the corresponding direction, G. RX It is the gain of the receiving antenna in the corresponding direction, L UG The loss is caused by the underground transmission of wireless signals, L UG-AG It is the loss caused by the wireless signal passing through the soil-air interface, L AG L represents the loss caused by the transmission of wireless signals over the ground. Surface The attenuation is due to the side waves, L M This is due to various losses from various possible sources, as well as miscellaneous losses due to the inability to satisfy the Friis equation, 10log 10 (x 2 This is path loss caused by multipath fading.

[0035] Previous studies have shown that LoRa signal loss in underground soil is closely related to volumetric water content (VWC). Therefore, it is hoped that soil VWC can be indirectly calculated from the RSSI value at the receiver. However, as can be seen from the formula, there are a series of interference terms in this transmission process, which may adversely affect the indirect measurement and calculation results. In recent years, deep learning has provided strong error suppression capabilities, which can effectively reduce numerical interference factors in the calculation process.

[0036] Therefore, in this invention, a deep learning method is introduced to optimize this computational process. Soil VWC is a time-varying value; obtaining a reliable sequence of feature values ​​within a short timeframe further improves the accuracy of the deep learning computation results. As can be seen from the formula, the directivity of the antenna gain helps to obtain a series of RSSI data sequences related to directional angle features.

[0037] Furthermore, to further improve the accuracy of indirect measurements, the front-end data acquisition method should be innovated. Previous methods often placed the receiver on the ground, which introduces errors into the RSSI value due to surface obstacles and side waves. However, by changing the structure and using a drone to lift the receiver into the air for reception, ground surface errors can be largely suppressed, further improving the indirect measurement results.

[0038] Based on this idea, this embodiment discloses a soil volumetric moisture content measurement system based on LoRa-RSSI and a drone, which includes:

[0039] like Figure 5As shown, an underground LoRa node is an IoUT (Internet of Underground Things) sensor node established using a LoRa module for the acquisition and transmission of environmental monitoring data. Environmental monitoring data includes soil moisture and temperature; in this embodiment, the sensor node collects soil temperature. The node is designed around an STM32F103ZET6 microcontroller chip. To save space, a TQFP-144 package is recommended. The node's temperature sensor uses a Sensirion STS31-DIS sensor and connects to the STM32F103ZET6 via an I2C data interface. The node's LoRa communication module uses an E22-400T22D module from Chengdu Yibaite, and the antenna is a Qifan 433MHz Little Pepper antenna. The node is powered by an 18650 battery via a TPS73033 at 3.3V. The node is encapsulated in a waterproof 6cm radius PVC pipe and placed vertically underground at a random location around the host computer, with the plane of the flat antenna as the reference plane. The average soil temperature at that moment is obtained by averaging the soil temperature data returned by these nodes.

[0040] Lightweight aerial nodes carried by small drones, such as Figure 6 As shown, the air node consists of an STM32 microcontroller, a LoRa communication module, and a power module. It receives data sent from the host computer and obtains the soil-to-air LoRa-RSSI value, then sends this value back to the host computer. The node is designed around an STM32F103ZET6 microcontroller chip (TQFP-144 package). The LoRa communication module uses an E22-400T22D module from Chengdu Ebate Technology Co., Ltd., and the antenna is a Kirin 433MHz flexible soft rubber rod external omnidirectional antenna. The node is powered by an 18650 battery via a TPS73033 at 3.3V. The air node is housed in a cylindrical waterproof container made of PET material, which is secured to the back of a DJI Marvic 2 drone using clips. The main lobe of the air node's antenna is perpendicular to the ground, and the antenna protrudes from the front of the drone to avoid interference with its LoRa signal transmission.

[0041] Existing research has found a strong correlation between soil VWC and the RSSI of LoRa wireless transmission receivers, and the interference caused by electromagnetic waves during signal propagation underground can be analyzed using the Friis formula, as shown below:

[0042] P RX =P TX +G TX +G RX -LUG -L UG-AG -0.5(L AG +L Surface )-L M -10log 10 (x 2 )

[0043] P RX It is soil RSSI, P TX It is the soil emission power output, G TX It is the gain of the transmitting antenna in the corresponding direction, G. RX It is the gain of the receiving antenna in the corresponding direction, L UG The loss is caused by the underground transmission of wireless signals, L UG-AG This refers to the refraction loss of wireless signals as they pass through the soil-air interface, L AG L represents the loss caused by the transmission of wireless signals over the ground. Surface The attenuation is due to transverse waves, L M This refers to various losses caused by possible sources (e.g., obstacles in the first Fresnel zone, antenna polarization mismatch, etc.), as well as miscellaneous losses due to the failure to satisfy some conditions of the Friis equation. In fact, although in most cases of UG2AG transmission, the distance between antennas can be much greater than the carrier wavelength, the transmitting and receiving antennas may have different polarizations, and they may be misaligned. Furthermore, since the transmitting antenna is buried underground, the unobstructed free space assumption can never be satisfied. Finally, 10log 10 (x 2 This is due to path loss caused by multipath fading.

[0044] like Figure 7 As shown, the recommended host computer is the EPC-S202 fanless handheld embedded industrial computer. The data acquisition software in the host computer is developed based on Qt 6.4. After the data collected by the underground LoRa node is transmitted to the host computer, it sends data to the air node through a flat panel antenna buried in the soil, and receives the soil-air LoRa-RSSI values ​​sent by the air node. The recommended LoRa module for the host computer is the Chengdu Ebate E22-400T22S wireless LoRa module, using two channels: one for communication with the node in the soil, and the other for communication with the UAV air node via the flat panel antenna. Both modules communicate with the host computer via UART to USB. The recommended flat panel antenna is the Qifan 433MHz flat panel antenna. When this antenna is placed, the main lobe should be perpendicular to the horizontal plane. The host computer utilizes the strong correlation between soil VWC and RSSI, such as... Figure 5As shown, an indirect measurement of soil VWC is achieved using a Long Short-Term Memory (LSTM) recurrent neural network. When the air node receives data from the host computer, the controlled rotation of the UAV during hovering causes a change in the direction of the air node's antenna, thus obtaining differential data in the time series (i.e., RSSI values ​​at different angles). In other words, by changing the angle between the transmitting and receiving antennas (in the air) through the rotation of the UAV, different soil-air LoRa-RSSI values ​​of the same soil VWC are obtained. The different soil-air LoRa-RSSI values ​​of the same soil VWC at time T, along with the feature value sequence consisting of the UAV rotation angle and soil temperature corresponding to the different soil-air LoRa-RSSI values, are input into the LSTM, and the LSTM calculates the soil VWC at time T.

[0045] Typically, underground nodes use omnidirectional antennas, resulting in the electromagnetic energy being transmitted omnidirectionally and dispersed. Furthermore, the absorption, reflection, and scattering caused by soil debris, plant roots, and other elements introduce significant measurement errors as the electromagnetic wave penetrates the soil. In this embodiment, the transmitting antenna (underground) uses a flat panel antenna, which provides stronger electromagnetic energy confinement, concentrating the energy on the main lobe (whose main lobe width is much smaller than that of an omnidirectional antenna, and it lacks large side lobes and back lobes). After being transmitted through the flat panel antenna, the electromagnetic wave is vertically transmitted through the soil into the air, avoiding errors caused by significant scattering and reflection from the surrounding environment. Moreover, because the energy is concentrated on the main lobe, the received RSSI signal from the aerial UAV node is also stronger, further reducing errors caused by environmental radio interference.

[0046] The host computer and flat panel antenna are housed in an IP67-standard waterproof box, with the antenna on top and the host computer and communication module below. The host computer is powered by 220V AC via an underground cable, which is converted to 12VDC by an adapter. The LoRa communication module on the host computer is powered via USB.

[0047] The workflow of the system for data acquisition disclosed in this invention is as follows: Figure 2 As shown. First, the host computer needs to be initialized. The drone, carrying sensors, flies to the predetermined altitude to act as an aerial node. Once the host computer software test begins, the system starts working. The success of the data acquisition system setup is determined by whether the host computer can receive data normally and whether the data format is accurate.

[0048] In this invention, the soil-air LoRa-RSSI value is used as a feature value input to LSTM to calculate soil VWC, combined with... Figure 3In the soil VWC calculation task of this invention, the input to the LSTM is usually multi-timestep three-dimensional data, but our data does not have a clear time step. We process the data into [[[dataA1],[dataT1],[dataR1]],[[dataA2],[dataT2],[dataR2]],[[dataA3],[dataT3],[dataR3]]…[[dataAn],[dataTn],[dataRn]]], input the three-dimensional data into the LSTM, select the data [dataV1],[dataV2],[dataV3]…[dataVn] as labels, normalize the data, and output the calculated value VWC, where dataAn is the UAV rotation angle, dataTn is the soil temperature, dataRn is the soil-air LoRa-RSSI value, and [dataVn] is the soil VWC.

[0049] To evaluate the calculation results, we selected the following three evaluation indicators:

[0050] (1) Root Mean Square Error (RMSE)

[0051] RMSE measures the average magnitude of errors and is sensitive to outliers in the calculated values. A smaller RMSE indicates a better calculation result. This is expressed as follows:

[0052]

[0053] Among them, a i p represents the true value. i This represents the calculated value, and N represents the number of sample points.

[0054] (2) Mean Absolute Error (MAE)

[0055] MAE uses the same units as the original data; it can only compare models with errors in the same units. Its magnitude is approximately the same as RMSE, but the error value is relatively smaller. The formula is as follows:

[0056]

[0057] (3) Coefficient of Determination

[0058] Coefficient of determination R 2 This describes the proportion of the dependent variable variance explained by the regression model in the total variance, as shown in the following formula:

[0059]

[0060] In the formula, It is the average of the true values.

[0061] To achieve better learning efficiency, we build an LSTM neural network layer. The LSTM has 270 hidden neurons, but we don't enable all of them; we only enable 20%. We then configure the training parameters. The training parameters mainly include the maximum number of iterations, gradient threshold, initial rate, initial rate descent period, and initial rate descent factor, as shown in the table below.

[0062]

[0063] Based on the above design, corresponding tests were conducted. The system test site was located in Songjiang District, Shanghai, East China. The average annual precipitation is 1103.2 mm, with 137 rainy days. The soil is mainly silty clay loam, with calcium carbonate increasing from top to bottom. The soil color is mainly yellowish-brown, with calcareous nodules. Photinia trees, approximately 2m tall with an average canopy height of about 1.3m, were planted around the test site, with a spacing of about 2-3m between trees. Due to random influences such as sunlight, plant canopy, and rainfall, the soil moisture around the test point will exhibit random variation trends related to the surrounding environment. Compared to the controlled environment of an indoor laboratory, this scenario presents a greater challenge for measurement but also offers higher realism. In the test, soil samples were first taken from the test area, and the VWC (Volatile Water Content) value of the soil in this area was obtained using the standard drying method. Then, a drone was operated to hover at a preset height of 3m, with the projection distance of this hovering point 5m from the flat panel antenna. The drone then rotates 360° at 45° intervals, collecting LoRa-RSSI data at each angle and sending it to the host computer. The host computer software records the soil temperature, processed RSSI value, relative soil VWC, and drone rotation angle into a data file. After completing the LoRa-RSSI data acquisition, Matlab is used to input soil temperature, RSSI value, and drone rotation angle as feature values ​​into an LSTM network, while historical soil VWC is input as a label into the LSTM network, and the current soil VWC value is calculated. Figure 8 By comparing the calculated values ​​with the actual values ​​of 32 sets of data, we found that the measured values ​​were basically similar to the actual values. Therefore, we plotted the soil VWC measurement error for each set. Figure 9 The measurement error was less than 0.35% in 32 measurements, which meets the soil VWC control threshold required for the growth of most crops. The test results show that the system disclosed in this invention can effectively achieve accurate VWC measurement.

Claims

1. A soil volumetric moisture content measurement system based on LoRa-RSSI and a drone, characterized in that, include: Several underground LoRa nodes are used to collect environmental monitoring data. All environmental monitoring data collected by the underground LoRa nodes are transmitted to the host computer buried in the soil via LoRa-based underground Internet of Things. The host computer, buried in the soil, receives environmental monitoring data transmitted from underground LoRa nodes and sends LoRa data packets to the omnidirectional antenna of the air node via a flat panel antenna buried in the soil. During this process, the main lobe of the flat panel antenna is perpendicular to the horizontal plane, and the main lobe plane of the omnidirectional antenna is also perpendicular to the horizontal plane. The host computer also receives soil-air LoRa-RSSI values ​​sent from the air node, obtains the relationship between environmental monitoring data and RSSI and SNR based on the Friis formula, and uses LSTM to measure soil VWC. The Friis formula is shown below: It is soil RSSI. It is the soil emission power output. It is the gain of the transmitting antenna in the corresponding direction. It is the gain of the receiving antenna in the corresponding direction. The loss is caused by the underground transmission of wireless signals. It is the refraction loss caused by the wireless signal passing through the soil-air interface. This represents the loss caused by the transmission of wireless signals over the ground. The attenuation is due to transverse waves. The losses are due to obstacles and antenna polarization mismatch in the first Fresnel zone, as well as miscellaneous losses due to failure to satisfy the Friis equations. This is due to path loss caused by multipath fading. The air node is carried by a small drone and is used to receive LoRa data packets sent by the host computer through a receiving antenna and obtain the soil-air LoRa-RSSI value. Then, it sends this soil-air LoRa-RSSI value back to the host computer. By using the rotation of the drone, the air node is controlled to change the angle between the transmitting antenna in the soil and the receiving antenna of the air node, so that the air node obtains different soil-air LoRa-RSSI values ​​for the same soil VWC. This series of different soil-air LoRa-RSSI values ​​is sent back to the host computer, which uses the feature value sequence composed of soil temperature, drone rotation angle and RSSI at each angle at time T to calculate the soil VWC at that time.

2. The soil volumetric moisture content measurement system based on LoRa-RSSI and UAV as described in claim 1, characterized in that, The underground LoRa node is an IoUT sensor node established using a LoRa module for the acquisition and transmission of environmental monitoring data.

3. The soil volumetric moisture content measurement system based on LoRa-RSSI and UAV as described in claim 1, characterized in that, The input LSTM data is processed into [[[dataA1],[dataT1],[dataR1]],[[dataA2],[dataT2],[dataR2]],[[dataA3],[dataT3],[dataR3]]…[[dataAn],[dataTn],[dataRn]]], and [dataV1],[dataV2],[dataV3]…[dataVn] data are selected as labels. After normalizing the data, the calculated value of soil VWC is output, where dataAn is the UAV rotation angle, dataTn is the environmental monitoring data, dataRn is the soil-air LoRa-RSSI value, and [dataVn] is the soil VWC.

4. The soil volumetric moisture content measurement system based on LoRa-RSSI and UAV as described in claim 1, characterized in that, Selecting root mean square error Mean absolute error and coefficient of determination As an evaluation metric for the LSTM: In the formula, Represents the actual value. This represents the calculated value. Indicates the number of sample points. It is the average of the true values.

Images