Portable complex medium moisture content intelligent detection terminal and system based on terminal side lightweight algorithm

Abstract

The invention relates to the technical field of complex medium moisture content detection, and discloses a portable complex medium moisture content intelligent detection terminal and system based on a terminal side lightweight algorithm, and the system comprises a detection terminal, an integrated multi-type sensor module, a low-power-consumption embedded processing unit and a lightweight algorithm. Electromagnetic scattering noise is eliminated through the microstrip antenna and the anti-interference filter circuit, and the moisture content of a single point can be calculated locally in real time. Each terminal uploads data to a sink node through a LoRa module by adopting a TDMA mechanism, the node calculates the regional moisture content through a space-time calibration and dynamic weighted fusion algorithm, generates a continuous thermodynamic diagram by means of inverse distance weighted interpolation, and finally realizes visualization and alarm functions through a mobile terminal APP. The problems that traditional equipment depends on cloud computing power and is poor in portability and high in detection delay are effectively solved, and high-precision, low-power-consumption and real-time moisture content detection is achieved through terminal-side intelligent computing and miniaturized hardware design.

Description

Technical Field

[0001] This invention relates to the field of complex media detection technology, and in particular to a portable intelligent detection terminal and system for complex media moisture content based on a lightweight terminal-side algorithm. Background Technology

[0002] Traditional equipment in existing complex media moisture content detection technologies is mainly applicable to scenarios where real-time detection requirements are not high, the location is fixed, and the power supply is stable, such as media characteristic analysis in a laboratory environment and batch sampling inspection of media moisture content in a factory workshop. In these scenarios, traditional equipment can detect the moisture content of complex media. By uploading the collected data to a cloud server and using server-side model inference to derive the detection results, it can meet the needs of quantitative analysis of moisture content within a certain range, providing basic data support for related research and production.

[0003] However, for scenarios with strict requirements for real-time monitoring, such as dynamic quality monitoring of production lines, traditional equipment suffers from high latency due to its reliance on cloud computing, making it unable to provide timely feedback on detection results and failing to meet the need for real-time adjustment of production parameters. In mobile scenarios such as field exploration and construction sites, the large size and reliance on mains power of traditional equipment result in poor portability, hindering flexible deployment to complete detection tasks. Furthermore, in heterogeneous and complex media environments, the single sensor used in traditional equipment is susceptible to electromagnetic interference, leading to large detection errors and making it difficult to adapt to the complexity of the medium, thus failing to provide accurate moisture content data. Summary of the Invention

[0004] Therefore, the purpose of this invention is to provide a portable intelligent detection terminal and system for the moisture content of complex media based on a lightweight terminal-side algorithm, which aims to solve the problems of traditional equipment relying on cloud computing power, poor portability, and high detection latency.

[0005] To achieve the aforementioned objectives, the technical solution adopted is as follows: An intelligent electromagnetic detection terminal for moisture content in complex media, comprising: Sensor module: integrates 1-10GHz electromagnetic wave sensor, temperature sensor and humidity auxiliary sensor, used to collect electromagnetic wave reflection coefficient, ambient temperature and humidity data in complex media; Embedded processing module: It adopts a low-power FPGA chip, integrates embedded customized hardware acceleration logic, and carries a hardware implementation unit for lightweight algorithms on the terminal side to perform parallel local processing on the data collected by the sensor module. Storage module: Used to store raw sensor data, algorithm model parameters, and detection results; Output module: Real-time output of moisture content detection values ​​and equipment operating status; Power module: Uses a rechargeable lithium battery, supporting continuous operation for ≥8 hours.

[0006] As a further improvement of the present invention, the lightweight algorithm is an improved random forest algorithm, which achieves lightweighting in the following ways: The number of decision trees is pruned to retain 50-80 core trees (60% of the original model trees); feature importance ranking is used to select the top 3 key features: electromagnetic wave frequency, temperature, and dielectric constant of the medium; a quantization weight matrix is ​​introduced to convert floating-point operations into 8-bit integer operations, and the model size is compressed to ≤500KB.

[0007] As a further improvement of the present invention, the electromagnetic wave sensor adopts a microstrip antenna structure with a size ≤5cm×5cm and a transmission power ≤10mW. The receiving end integrates an anti-interference filter circuit to eliminate electromagnetic scattering noise in complex media.

[0008] As a further improvement of the present invention, the embedded processing module adopts a low-power FPGA chip, supports parallel hardware acceleration logic, and integrates a 32-bit soft-core processor.

[0009] As a further improvement of the present invention, the moisture content calculation process of the lightweight algorithm includes: Reflection coefficient collected by electromagnetic wave sensor Feature extraction is performed to obtain the resonant frequency offset. With amplitude attenuation rate ; Combination temperature ,humidity Parameters, constructing the input feature vector ; Input a lightweight random forest model, output water content The expression is: ; in, The number of decision trees after pruning (50≤ k ≤80), For the first The output of each decision tree.

[0010] A moisture content intelligent detection system based on any of the above-mentioned detection terminals, characterized in that it includes: At least three of the aforementioned detection terminals; Wireless data aggregation node: Receives detection data from various terminals via LoRa technology, performs spatiotemporal consistency calibration, and integrates data from multiple nodes; Mobile App: Supports visualization of detection data, query of historical records, and threshold alarms.

[0011] As a further improvement of the present invention, the multi-node data fusion adopts a weighted average algorithm, and the weight coefficients are dynamically adjusted according to the detection accuracy of each terminal, as expressed in the following expression: ; in, W The moisture content after fusion. For the first The weight of each terminal, For the first The detection values ​​of each terminal, For the first Historical standard deviation of each terminal This represents the average standard deviation of all terminals.

[0012] As a further improvement of the present invention, the wireless data aggregation node supports a maximum of 32 terminal accesses, with a communication distance of ≥500m, adopts a dynamic time slot allocation mechanism to avoid data conflicts, and the data transmission delay is ≤100ms.

[0013] The beneficial effects of this invention are: This invention integrates an anti-interference filter circuit with a microstrip antenna structure at the receiving end to eliminate electromagnetic wave scattering noise in complex media; and it calculates the single-point water content locally in real time using a lightweight improved random forest algorithm embedded on the terminal side. Subsequently, each terminal uploads data to the aggregation node via the LoRa module according to the TDMA mechanism. The node first performs spatiotemporal calibration to remove outliers, and then calculates the regional fused water content using a dynamic weighted fusion algorithm. W It combines inverse distance weighted interpolation to generate continuous regional moisture content heat maps, and finally realizes visualization and anomaly alarm through a mobile APP. For the first time, it has achieved a whole-process innovation from local acquisition of multiple parameters of complex media, real-time intelligent calculation on the terminal side to multi-node collaborative integration. It not only improves the accuracy of single-point detection, but also breaks through the spatial limitations of traditional single-point detection, providing a brand-new solution for regional complex media moisture content detection. Attached Figure Description

[0014] The accompanying drawings, which form part of this application, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 A schematic diagram illustrating a multi-terminal collaborative detection application scenario for farmland soil moisture content. Figure 2 Workflow diagram of intelligent detection terminal and system for farmland soil moisture content; Figure 3 This image shows an example of a multi-terminal collaborative detection application for soil moisture content in farmland. Detailed Implementation

[0015] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0016] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0017] An intelligent electromagnetic detection terminal for moisture content in complex media, comprising: Sensor module: integrates 1-10GHz electromagnetic wave sensor, temperature sensor and humidity auxiliary sensor, used to collect electromagnetic wave reflection coefficient, ambient temperature and humidity data in complex media; Embedded processing module: It adopts a low-power FPGA chip, integrates embedded customized hardware acceleration logic, and carries a hardware implementation unit for lightweight algorithms on the terminal side to perform parallel local processing on the data collected by the sensor module. Storage module: Used to store raw sensor data, algorithm model parameters, and detection results; Output module: Real-time output of moisture content detection values ​​and equipment operating status; Power module: Uses a rechargeable lithium battery, supporting continuous operation for ≥8 hours.

[0018] The lightweight algorithm is an improved random forest algorithm, which achieves lightweighting through the following methods: The number of decision trees is pruned to retain 50-80 core trees (60% of the original model trees); feature importance ranking is used to select the top 3 key features: electromagnetic wave frequency, temperature, and dielectric constant of the medium; a quantization weight matrix is ​​introduced to convert floating-point operations into 8-bit integer operations, and the model size is compressed to ≤500KB.

[0019] The electromagnetic wave sensor adopts a microstrip antenna structure with a size of ≤5cm×5cm and a transmission power of ≤10mW. The receiving end integrates an anti-interference filter circuit to eliminate electromagnetic scattering noise in complex media.

[0020] The embedded processing module uses a low-power FPGA chip, supports parallel hardware acceleration logic, and integrates a 32-bit soft-core processor.

[0021] The moisture content calculation process of the lightweight algorithm includes: Reflection coefficient collected by electromagnetic wave sensor Feature extraction is performed to obtain the resonant frequency offset. With amplitude attenuation rate ; Combination temperature ,humidity Parameters, constructing the input feature vector ; Input a lightweight random forest model, output water content The expression is: ; in, The number of decision trees after pruning (50≤ k ≤80), For the first The output of each decision tree.

[0022] A moisture content intelligent detection system based on any of the above-mentioned detection terminals, characterized in that it includes: At least three of the aforementioned detection terminals; Wireless data aggregation node: Receives detection data from various terminals via LoRa technology, performs spatiotemporal consistency calibration, and integrates data from multiple nodes; Mobile App: Supports visualization of detection data, query of historical records, and threshold alarms.

[0023] The multi-node data fusion employs a weighted average algorithm, with the weighting coefficients dynamically adjusted based on the detection accuracy of each terminal. The expression is as follows: ; in, W The moisture content after fusion. For the first The weight of each terminal, For the first The detection values ​​of each terminal, For the first Historical standard deviation of each terminal This represents the average standard deviation of all terminals.

[0024] The wireless data aggregation node supports up to 32 terminal accesses, with a communication distance of ≥500m. It adopts a dynamic time slot allocation mechanism to avoid data conflicts, and the data transmission delay is ≤100ms.

[0025] 1. Terminal hardware design: It employs an ultra-small microstrip antenna, covering electromagnetic wave frequencies from 1 to 10 GHz, and uses AD conversion to achieve reflection coefficient acquisition; it also includes a temperature sensor and a humidity auxiliary sensor; it uses a low-power FPGA chip, integrating customized parallel hardware logic for Fast Fourier Transform (FFT) acceleration, supporting parallel real-time extraction of electromagnetic wave features; and it is powered by a lithium battery, supporting a low-power battery life design.

[0026] 2. Lightweight Algorithm Implementation: Original random forest model optimization: Using 5000 soil samples, the three features with the greatest impact on water content were selected through the Gini coefficient. , (Temperature), the number of decision trees was pruned from 100 to 60; INT8 quantization was introduced to deeply compress and optimize model parameters, while improving inference speed and achieving efficient operation on the terminal side.

[0027] Combination temperature T ,humidity H Parameters, constructing the input feature vector ; Input a lightweight random forest model, output water content M The expression is: ; 3. System workflow: Three terminals were evenly deployed in the farmland monitoring area, spaced 5-10 meters apart, and initial parameter configuration was completed via Bluetooth. Each terminal collected data every 30 seconds, outputting and storing the moisture content locally. An anti-interference filter circuit was integrated into the receiver using a microstrip antenna structure to eliminate electromagnetic wave scattering noise in complex media. The single-point moisture content was calculated in real time locally using a lightweight improved random forest algorithm embedded on the terminal side. Subsequently, each terminal uploads data to the aggregation node via LoRa module according to the TDMA mechanism. The terminals upload data to the aggregation node via LoRa every hour on the hour. The aggregation node performs spatiotemporal calibration and then calculates the regional fused water content using a dynamic weighted fusion algorithm. W The system generates a continuous regional moisture content heat map by combining inverse distance weighted interpolation, and finally uploads it to the cloud via GPRS. The mobile APP displays the moisture content distribution heat map in real time.

[0028] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, component splitting or combination, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An intelligent electromagnetic detection terminal for water content in complex media, characterized in that, include: Sensor module: integrates 1-10GHz electromagnetic wave sensor, temperature sensor and humidity auxiliary sensor, used to collect electromagnetic wave reflection coefficient, ambient temperature and humidity data in complex media; Embedded processing module: It adopts a low-power FPGA chip, integrates embedded customized hardware acceleration logic, and carries a hardware implementation unit for lightweight algorithms on the terminal side to perform parallel local processing on the data collected by the sensor module. Storage module: Used to store raw sensor data, algorithm model parameters, and detection results; Output module: Real-time output of moisture content detection values ​​and equipment operating status; Power module: Uses a rechargeable lithium battery, supporting continuous operation for ≥8 hours.

2. The intelligent electromagnetic detection terminal for water content in complex media according to claim 1, characterized in that: The lightweight algorithm is an improved random forest algorithm, which achieves lightweighting through the following methods: Prune the number of decision trees, retaining 50-80 core trees; By ranking features by importance, the top three key features—electromagnetic wave frequency, temperature, and dielectric constant of the medium—were selected. By introducing a quantized weight matrix, floating-point operations are converted into 8-bit integer operations, and the model size is compressed to ≤500KB.

3. The intelligent electromagnetic detection terminal for water content in complex media according to claim 1, characterized in that: The electromagnetic wave sensor adopts a microstrip antenna structure with a size of ≤5cm×5cm and a transmission power of ≤10mW. The receiving end integrates an anti-interference filter circuit to eliminate electromagnetic scattering noise in complex media.

4. The intelligent electromagnetic detection terminal for water content in complex media according to claim 1, characterized in that, The embedded processing module uses a low-power FPGA chip, supports parallel hardware acceleration logic, and integrates a 32-bit soft-core processor.

5. The intelligent electromagnetic detection terminal for water content in complex media according to claim 1, characterized in that, The moisture content calculation process of the lightweight algorithm includes: Reflection coefficient collected by electromagnetic wave sensor Feature extraction is performed to obtain the resonant frequency offset. With amplitude attenuation rate ; Combination temperature ,humidity Parameters, constructing the input feature vector ; Input a lightweight random forest model, output water content The expression is: ； in, The number of decision trees after pruning. For the first The output of each decision tree.

6. A moisture content intelligent detection system based on any one of the detection terminals described in claims 1-5, characterized in that, include: At least three detection terminals as described in any one of claims 1-5; Wireless data aggregation node: Receives detection data from various terminals via LoRa technology, performs spatiotemporal consistency calibration, and integrates data from multiple nodes; Mobile App: Supports visualization of detection data, query of historical records, and threshold alarms.

7. The intelligent moisture content detection system according to claim 6, characterized in that: The multi-node data fusion adopts a weighted average algorithm, and the weight coefficients are dynamically adjusted according to the detection accuracy of each terminal. The expression is: ； in, W The moisture content after fusion. For the first The weight of each terminal, For the first The detection values ​​of each terminal, For the first Historical standard deviation of each terminal This represents the average standard deviation of all terminals.

8. The intelligent moisture content detection system according to claim 6, characterized in that, The wireless data aggregation node supports up to 32 terminal accesses, with a communication distance of ≥500m. It adopts a dynamic time slot allocation mechanism to avoid data conflicts, and the data transmission delay is ≤100ms.

Images