Portable water quality detection system with multi-sensor fusion, self-calibration and image recognition functions

Abstract

The invention provides a portable water quality detection system based on multi-sensor fusion, self-calibration and image recognition, and relates to the technical field of water quality detection.Water quality parameters are collected through multi-sensor fusion, data are processed in combination with the deep learning self-calibration and image recognition technology, and a comprehensive water quality analysis result is generated; the system integrated communication module uploads data to a cloud end in real time, low-power-consumption power management is achieved, portable deployment of field and emergency scenes is supported, the detection accuracy, intelligence and real-time monitoring capacity are remarkably improved, and efficient water quality safety supervision is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of water quality detection, and particularly to a portable water quality detection system integrating multi-sensor fusion, self-calibration, and image recognition. Background Art

[0002] Water quality monitoring is a key link in ensuring drinking water safety and environmental health. Current water quality detection mainly relies on laboratory analysis and traditional sensor instruments, such as electrochemical sensors and spectrometers. Although these methods have high measurement accuracy, they have obvious deficiencies: expensive equipment, complex operation, and long time consumption, which are not suitable for on-site real-time detection. Many institutions still use the method of manual sampling and sending for inspection, and calculate the water quality index (WQI) through laboratory analysis, which is time-consuming and costly. Traditional monitoring technologies also have limitations in obtaining real-time and comprehensive data, and need to improve remote collaboration and real-time data acquisition capabilities.

[0003] The above-mentioned prior art has the following defects and deficiencies:

[0004] (1) High system complexity: A typical water quality monitoring system consists of multiple different instruments, which are complex to deploy and maintain and require professional operation. Different sensors work independently, lacking a unified platform integration, which increases the system complexity.

[0005] (2) Insufficient intelligence: Most existing devices lack intelligent analysis and automatic calibration functions and cannot adaptively adjust according to environmental changes. Sensors drift over time and require frequent manual calibration, and traditional methods cannot automatically compensate for sensor errors. Without AI assistance, it is difficult to achieve timely early warning and autonomous judgment of abnormal conditions.

[0006] (3) Slow response speed: There is often a large delay from water sampling to obtaining results in water quality detection. Laboratory tests take several hours to several days and it is difficult to reflect water quality changes in a timely manner. Even for on-line monitoring instruments, limited by the response characteristics of sensors and data transmission frequency, it is difficult to achieve true real-time response. When water quality suddenly changes, traditional systems are difficult to capture and respond in a timely manner.

[0007] (4) Poor data fusion: Current multi-parameter water quality monitoring often presents each parameter separately, lacking an effective data fusion algorithm for comprehensive analysis. Multi-sensor data often exists in an isolated form, unable to corroborate or cooperate with each other for judgment, and it is difficult to obtain a comprehensive assessment of water quality and make full use of the complementary information of multi-source data.

[0008] These deficiencies lead to room for improvement in the real-time performance, intelligence, and accuracy of existing water quality monitoring systems. Therefore, there is an urgent need for a new type of intelligent water quality detection system integrating multi-sensor fusion, having self-calibration ability, and combining image recognition to improve monitoring efficiency and result reliability. Summary of the Invention

[0009] To overcome the deficiencies of the prior art, the object of the present invention is to provide a portable water quality detection system with multi-sensor fusion, self-calibration and image recognition. Through the synergistic effect of multi-sensor data fusion, deep learning self-calibration and image recognition technologies, the accuracy and intelligence of water quality detection are significantly improved; its portable integrated design and low-power long-endurance characteristics can be flexibly deployed in various scenarios such as the wild and emergencies to achieve real-time monitoring and remote warning, greatly improving the efficiency of water quality safety supervision.

[0010] To achieve the above object, the present invention provides the following solutions:

[0011] A portable water quality detection system with multi-sensor fusion, self-calibration and image recognition, comprising:

[0012] A sensor module, used to synchronously collect the physical and chemical parameters of the water body to obtain sensor data;

[0013] An image acquisition module, used to acquire the image of the water sample or reagent reaction to obtain the water sample image;

[0014] A data processing module, respectively connected to the sensor module and the image acquisition module, used to self-calibrate the sensor data based on a deep learning algorithm, compensate for sensor drift and environmental interference to obtain preprocessed data, and use a convolutional neural network to perform feature extraction and pattern recognition on the water sample image acquired by the image acquisition module to obtain an image recognition result, and comprehensively analyze the preprocessed data and the image recognition result through a multi-modal fusion algorithm to generate a water quality analysis result;

[0015] A communication module, connected to the data processing module, used to upload the water quality analysis result, the preprocessed data and the image recognition result to an external cloud server and receive remote control instructions;

[0016] A display and interaction module, connected to the data processing module, used to locally display the water quality analysis result, the preprocessed data and the image recognition result and provide a user operation interface;

[0017] A power module, connected to the data processing module, integrating a rechargeable battery and a power management circuit, used to extend the endurance time of the data processing module through a low-power strategy;

[0018] A cloud platform interface module, respectively connected to the communication module and the external cloud server, used to realize data interaction with the cloud server and support remote monitoring and big data analysis.

[0019] Preferably, the sensor module includes a pH sensor, a conductivity sensor, a dissolved oxygen sensor, a temperature sensor, and a turbidity sensor, which are used to monitor the acidity, salinity, oxygen content, temperature, and turbidity of water in real time.

[0020] Preferably, the deep learning algorithm of the data processing module is trained by comparing the historical sensor data with the standard reference values to dynamically adjust the calibration parameters when the sensor performance drifts, and the Kalman filter algorithm is used to fuse multi-sensor data.

[0021] Preferably, the convolutional neural network is integrated with an attention mechanism, and the convolutional neural network is used to focus on the key areas related to water quality in the water sample image; the image recognition results include at least one of reagent colorimetric analysis, abnormal water body color detection, or suspended matter distribution evaluation.

[0022] Preferably, the communication module supports multiple wireless transmission protocols; the wireless transmission protocols include Wi-Fi, cellular mobile network, and low-power wide area network; the wireless transmission protocols use encryption protocols to ensure the security of data transmission.

[0023] Preferably, the power module reduces energy consumption through dynamic frequency modulation, sensor power supply on demand, and intermittent sleep strategies to support the continuous operation of the data processing module for more than 12 hours in the preset working mode.

[0024] Preferably, the display and interaction module includes a touch screen and a buzzer; the display and interaction module supports local parameter setting, historical data query, and abnormal status alarm, and is connected to a mobile terminal through Bluetooth to achieve near-field interaction.

[0025] Preferably, the cloud platform interface module provides a water quality data visualization interface, supports threshold alarm, remote firmware upgrade, and multi-device collaborative management, and generates a long-term trend analysis report.

[0026] Preferably, both the sensor module and the image acquisition module are detachable structures.

[0027] Preferably, the multi-modal fusion algorithm includes at least one of the following methods:

[0028] Dynamic weighted fusion of sensor data and image recognition results based on Kalman filter;

[0029] Use a stacked autoencoder to extract the temporal features of sensor data, and jointly input the features of the image recognition results into a deep neural network for non-linear fusion.

[0030] According to the specific embodiments provided by the present invention, the following technical effects are disclosed by the present invention:

[0031] The present invention provides a portable water quality detection system for multi-sensor fusion, self-calibration, and image recognition, including: a sensor module for synchronously collecting physical and chemical parameters of water bodies to obtain sensor data; an image acquisition module for acquiring images of water samples or reagent reactions to obtain water sample images; a data processing module connected to the sensor module and the image acquisition module respectively, for self-calibrating the sensor data based on deep learning algorithms to compensate for sensor drift and environmental interference, obtaining preprocessed data, and using a convolutional neural network to extract features and perform pattern recognition on the water sample images acquired by the image acquisition module to obtain image recognition results, and comprehensively analyzing the preprocessed data and the image recognition results through a multi-modal fusion algorithm to generate water quality analysis results; a communication module connected to the data processing module for uploading the water quality analysis results, the preprocessed data, and the image recognition results to an external cloud server and receiving remote control instructions; a display and interaction module connected to the data processing module for locally displaying the water quality analysis results, the preprocessed data, and the image recognition results and providing a user operation interface; a power module connected to the data processing module, integrating a rechargeable battery and a power management circuit for extending the battery life of the data processing module through a low-power strategy; a cloud platform interface module connected to the external cloud server for realizing data interaction with the cloud server and supporting remote monitoring and big data analysis. Through the synergistic effect of multi-sensor data fusion, deep learning self-calibration, and image recognition technologies, the present invention significantly improves the accuracy and intelligence of water quality detection; its portable integrated design and low-power long battery life characteristics can be flexibly deployed in various scenarios such as the wild and emergencies to achieve real-time monitoring and remote warning, greatly improving the efficiency of water quality safety supervision. Description of the Drawings

[0032] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the following described drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0033] Figure 1 It is a schematic diagram of the system structure provided by the embodiment of the present invention;

[0034] Figure 2 It is a schematic diagram of the technical route provided by the embodiment of the present invention;

[0035] Figure 3 It is a schematic diagram of the module structure provided by the embodiment of the present invention;

[0036] Figure 4Flow chart of water quality detection and analysis provided by the embodiments of the present invention;

[0037] Figure 5 Schematic diagram of the system advantages provided by the embodiments of the present invention. Detailed implementation manners

[0038] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0039] The purpose of the present invention is to provide a portable water quality detection system with multi-sensor fusion, self-calibration and image recognition. Through the synergistic effect of multi-sensor data fusion, deep learning self-calibration and image recognition technologies, the accuracy and intelligence of water quality detection are significantly improved; its portable integrated design and low-power long-endurance characteristics can be flexibly deployed in various scenarios such as the wild and emergencies to achieve real-time monitoring and remote warning, greatly improving the efficiency of water quality safety supervision.

[0040] To make the above objects, features and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific implementation manners.

[0041] Figure 1 Schematic diagram of the system structure provided by the embodiments of the present invention, as Figure 1 shown, the present invention provides a portable water quality detection system with multi-sensor fusion, self-calibration and image recognition, including:

[0042] A sensor module for synchronously collecting physical and chemical parameters of water bodies to obtain sensor data;

[0043] An image acquisition module for obtaining images of water samples or reagent reactions to obtain water sample images;

[0044] A data processing module, respectively connected to the sensor module and the image acquisition module, for self-calibrating the sensor data based on a deep learning algorithm to compensate for sensor drift and environmental interference to obtain preprocessed data, and using a convolutional neural network to perform feature extraction and pattern recognition on the water sample images obtained by the image acquisition module to obtain image recognition results, and comprehensively analyzing the preprocessed data and the image recognition results through a multi-modal fusion algorithm to generate water quality analysis results;

[0045] A communication module, connected to the data processing module, is used to upload the water quality analysis result, the preprocessing data, and the image recognition result to an external cloud server and receive remote control instructions;

[0046] A display and interaction module, connected to the data processing module, is used to locally display the water quality analysis result, the preprocessing data, and the image recognition result and provide a user operation interface;

[0047] A power module, connected to the data processing module, integrates a rechargeable battery and a power management circuit and is used to extend the battery life of the data processing module through a low-power strategy;

[0048] A cloud platform interface module, respectively connected to the communication module and the external cloud server, is used to implement data interaction with the cloud server and support remote monitoring and big data analysis.

[0049] Specifically, the system of the present invention adopts a modular design and is integrated by multiple functional modules, including but not limited to: a sensor module, an image acquisition module, a data processing module, a communication module, a display and interaction module, a cloud platform interface module, and a power and energy consumption management module, etc. As Figure 2 and Figure 3 shown, the main functions and roles of each module are as follows:

[0050] (1) Sensor module: It contains a variety of water quality sensors and is used to collect the physical and chemical parameters of water bodies. Typical sensors include pH sensors, dissolved oxygen (DO) sensors, electrical conductivity (EC) sensors, temperature sensors, turbidity sensors, etc., covering key indicators such as the acidity and alkalinity of water quality, dissolved oxygen content, salinity / total dissolved solids, temperature, and turbidity. Through the collaborative work of multiple sensors, various parameters of water quality can be monitored in real time to achieve a comprehensive perception of the water quality status. The sensor module is usually designed in the form of a pluggable probe for easy on-site rapid deployment and replacement.

[0051] (2) Image acquisition module: It consists of a micro camera and a supplementary light source and is used to obtain images of water samples or reagent reaction results. The camera captures image information such as changes in the color and transparency of water bodies or the color reaction of test paper reagents, providing additional evidence that cannot be quantified by the naked eye. The visual information obtained through the image sensor can assist in detecting, such as abnormal water body colors, the distribution of suspended matter particles, or reagent colorimetric results. This module is usually located near the sensor probe to ensure clear close-up images of water samples are obtained.

[0052] (3) Data Processing Module: The core control unit of the system, including an embedded processor or a microcontroller unit (MCU) and a memory. This module processes and analyzes the data obtained by the sensor module and the image acquisition module in real time. Its sub-functions include: data preprocessing (filtering, denoising, and normalizing the signals of each sensor), intelligent calibration (running a deep learning-based self-calibration algorithm to compensate for sensor errors), image recognition (running a convolutional neural network CNN model to extract and analyze the features of the collected water sample images), and multi-sensor data fusion (combining the values of each sensor and the results of image recognition to determine the water quality status). Through the above intelligent algorithms, the data processing module realizes high-level analysis of the original data, greatly improving the measurement accuracy and intelligence level.

[0053] (4) Communication Module: Responsible for the wired / wireless communication interface between the system and the outside world, realizing data transmission and remote control. Wireless communication can adopt methods such as Wi-Fi, cellular mobile networks (such as NB-IoT / 4G), or low-power wide-area networks (such as LoRa) to upload the detection data to the cloud platform or user terminal in real time. For example, in a campus or urban environment, data can be transmitted to the cloud through Wi-Fi / 4G, and in remote areas, long-distance low-power communication such as LoRa can be used to send data to a centralized gateway. The communication module also supports short-range communication such as Bluetooth, so that on-site personnel can use devices such as mobile phones to connect to the system to read data or issue commands.

[0054] (5) Display and Interaction Module: Used for local display of detection results and providing a user interaction interface. It can include a low-power LCD / OLED screen or LED indicator lights to display key water quality parameter values, device status, and alarm information, etc. Users can interact with the system through buttons, touchscreens, or mobile phone apps to achieve function configuration (such as setting the measurement interval, starting calibration), data query, and result storage, etc. This module improves the usability of the system, enabling non-professional users to easily view and understand water quality information.

[0055] (6) Cloud Platform Interface Module: Connects to the cloud server through the communication module to realize remote storage and management of data. The cloud platform is used to receive real-time water quality data uploaded from the device, perform big data storage and in-depth analysis, and provide a visual interface for remote users to monitor. Managers can view the historical trends and statistical reports of multiple devices in the cloud and can set thresholds to achieve automatic early warning for exceeding the standard. The cloud platform can also send configuration updates or algorithm upgrade packages to realize remote maintenance and function expansion of the front-end device. Through cloud integration, collaborative management of a large-scale water quality monitoring network can be realized, improving the sharing and utilization value of data.

[0056] (7) Power and Energy Consumption Management Module: It includes a lithium battery power supply unit, a power management circuit, and an energy consumption optimization strategy to ensure the long-term stable operation of the system. The power module provides the required voltage for each functional module and has functions such as overcharge / overdischarge protection and voltage monitoring. The energy consumption management strategy includes measures such as power supply on demand for sensors, dynamic frequency modulation of the processor, and intermittent operation of the communication module: when the system is idle or the monitored environment is stable, the sensors and the processor are put into the sleep or low-power mode; they are woken up at preset intervals to collect and send data, thus significantly extending the battery life. Through optimizing the hardware and software design, this system can work continuously for several hours or even longer under typical settings, meeting the requirements for battery life in field portable monitoring.

[0057] The above modules cooperate to form the intelligent water quality detection and analysis system of the present invention: the sensor and image modules acquire multi-source water quality information, the data processing module conducts intelligent analysis and fusion, the communication module interacts with the cloud platform, and finally the results are presented through the display module and the power module provides lasting power support.

[0058] As Figure 4 shown, the water quality detection and analysis process of the present invention includes the following stages: data acquisition → data preprocessing → image processing → multimodal fusion → cloud transmission → display and interaction → energy consumption management. The specific steps and main algorithms adopted in each stage are as follows:

[0059] (1) Data Acquisition: The sensor module and the image module acquire water quality data according to the set period or trigger condition. Each sensor (pH, conductivity, temperature, etc.) in the sensor module obtains real-time values through ADC analog-to-digital conversion; the camera takes images of the water sample when needed. The system can set the sampling frequency according to requirements. For example, all sensor readings are collected once per minute, and images are acquired when necessary. To ensure data reliability, metadata such as timestamps and sensor status are recorded during the acquisition stage. If reagents need to be added to the water sample for colorimetric reaction, the control unit coordinates the acquisition module to obtain images at appropriate reaction time points. The raw data obtained in the data acquisition stage will enter the subsequent processing process.

[0060] (2) Data Preprocessing: Clean and filter the multi-source raw data to improve clarity and signal-to-noise ratio. For sensor values, digital filtering algorithms are used to smooth random noise and transient interference. For example, the moving average filtering method is applied to calculate the average value of the readings within the sliding window, and the formula is as follows:

[0061]

[0062] Where \(x[n]\) is the current reading of the sensor, \(y[n]\) is the filtered value, and \(N\) is the window length. By moving average, the influence of high-frequency noise on measurement can be effectively reduced. For outliers affected by sudden interference, median filtering can be combined to detect and remove the outlier points. In addition, the preprocessing stage also includes the calibration conversion of sensor data: converting the original voltage or frequency value into the corresponding physical quantity unit and correcting it with the calibration coefficient obtained from the most recent calibration. For image data, preprocessing includes color correction, white balance adjustment, and ROI (Region of Interest) extraction, etc. For example, intercepting the image part of the reagent reaction area and adjusting the brightness and contrast to reduce the influence of ambient light changes. After preprocessing, a clean and reliable sensor reading sequence and normalized images are obtained, preparing for subsequent analysis.

[0063] (3) Image processing (CNN structure summary): The preprocessed image is analyzed by a deep learning model to extract water quality-related feature information. The system uses a Convolutional Neural Network (CNN) to automatically identify the water sample image. Its structure includes multiple convolutional layers, pooling layers, and fully connected layers, which are used to extract image features step by step and make classification / regression judgments. To improve the model's attention ability, an attention mechanism (Attention Module) is introduced: different weights are assigned to the extracted feature maps to emphasize the attention to key image regions. For example, by calculating the importance score \(e\) i and normalizing it via the Softmax function, the weight coefficient of the \(i\)-th feature can be obtained:

[0064]

[0065] where \(\alpha\) i represents the attention weight of feature \(i\).

[0066] This attention mechanism enables the CNN to focus on areas with the most significant reagent color reaction or obvious suspension particle distribution, thus improving the recognition accuracy. The output of the CNN model varies according to application requirements: for quantitative analysis, the model can regress the corresponding water quality parameter values (e.g., predicting the residual chlorine concentration based on the reagent color); for qualitative analysis, the model can give the discrimination result of the water quality status (such as "normal / exceeding the standard"). The CNN model in this system has been trained with a large number of samples and can adapt to various shooting light conditions and water sample changes. By extracting high-level semantic features through deep convolution and strengthening key information with attention, the image processing module can convert complex visual information into quantifiable water quality indicators or auxiliary judgment results.

[0067] (4) Multimodal Fusion: The numerical data from sensors is fused with the results from image processing to comprehensively evaluate the water quality status. The fusion algorithm comprehensively utilizes the complementary information of different modal data to improve the reliability and accuracy of the determination. One implementation method is to adopt sensor fusion algorithms such as Kalman filtering: regarding the measured values of each sensor and the results inferred from the image as the observation inputs of the true water quality state, and using the recursive estimation of Kalman filtering to obtain the optimal estimation of the water quality parameters. Kalman filtering dynamically allocates weights according to the uncertainties of each sensor through a prediction-correction cycle to update the optimal estimation value. For example, for a certain water quality parameter, Kalman filtering will calculate the "Kalman gain" to balance the error between the sensor measurement z k and the prior estimation, so as to obtain the updated estimation after fusion Through this recursive calculation, it can accurately approximate the true value even in the presence of noise in multi-source data. Another fusion method is to construct a multimodal deep learning network: inputting the sensor numerical values and image features into a fusion model (such as a densely connected neural network), and the network automatically learns the associations between different modalities to achieve non-linear information fusion. For example, for the prediction of nutrient salt concentration, a two-level stacked autoencoder can be used to extract the features of the sensor sequence and fuse the image features extracted by the CNN to jointly predict the total nitrogen and total phosphorus contents. Regardless of the method adopted, the fusion result can provide a more comprehensive water quality assessment, such as giving a comprehensive water quality index or a conclusion on whether there are abnormal situations. Multimodal fusion makes full use of the information advantages of both sensors and images, making the judgment of the system more robust and reliable.

[0068] (5) Cloud Transmission: The water quality data and results obtained through local processing and fusion are sent to the cloud platform and / or user terminal through the communication module. The communication module selects an appropriate transmission method according to the network environment: when there is Wi-Fi or wired network, the data is uploaded through TCP / IP; in the outdoor environment, 4G / 5G cellular networks or Internet of Things protocols (MQTT, CoAP) can be used to send the data; in remote areas, low-power wide-area networks such as LoRa / NB-IoT are used to report regularly. The transmitted data includes the current readings of each sensor, the water quality indicators obtained through fusion calculation, the on-site device ID, the timestamp, etc., and an encryption protocol is used to ensure the transmission security. After receiving the data, the cloud stores it and can further trigger the analysis and visualization processes in the cloud (see Step 6 for details). In addition, the cloud can also send control instructions to the front-end devices, such as adjusting the sampling frequency, upgrading the algorithm model, etc. For scenarios without public network connection, the system also supports sending the data to local mobile terminals (such as sending the results to a mobile phone App through Bluetooth). Through a stable and reliable communication link, it ensures that the monitoring data is delivered to the backend in a timely manner, realizing remote monitoring and management.

[0069] (6) Display and interaction: The system provides data display and interaction functions both locally and remotely. Locally, the processor sends key results to the display module for real-time presentation. For example, the LCD screen scrolls to display the current water quality parameter values, and a buzzer or LED flashes to warn in case of abnormalities. Users can browse historical records or trigger manual measurements by pressing buttons. Remotely, the cloud platform aggregates data from the field and generates charts and reports for users to view. For example, through the Web interface or mobile phone app, users can view the pH and turbidity curves of a device in the last 24 hours, or water quality maps of multiple monitoring points. Once a parameter exceeds the preset threshold, the cloud platform will highlight it and notify relevant personnel via SMS / push. Users can also operate the device on the remote interface, such as starting a calibration, taking photos of the site, and other control commands. These commands are sent from the cloud and received and executed by the on-site communication module. By combining local real-time display with remote visualization, a friendly human-computer interaction experience is formed to ensure that data can be "seen and used well", making it convenient for users to grasp the water quality situation in a timely manner and operate the device.

[0070] (7) Energy management: The entire process is implemented with an intelligent energy management strategy to balance monitoring timeliness and power consumption. The system schedules the working status of each functional module according to the preset working mode and the current battery power: during the data collection interval, the sensor and camera are powered off and put into sleep mode, the processor enters low-power standby mode, and only the real-time clock is kept running for wake-up; after completing a data processing and transmission, if the next measurement cycle has not yet arrived, the processor will go into sleep mode again. The communication module immediately turns off the RF circuit after sending data in batches to reduce the power consumption of continuous communication. When the monitoring environment is stable for a long time, the system can automatically extend the sampling interval and enter the "power saving mode"; once a key parameter fluctuates violently or a remote command is received, it can switch to the "intensive monitoring mode" to increase the sampling frequency. In terms of hardware, the power management circuit continuously monitors the battery voltage. When the power is lower than the threshold, it notifies the processor to send a low-power warning to the cloud to remind maintenance personnel to replace the battery or charge it. Through the above-mentioned energy optimization of hardware and software, the system can minimize power consumption while ensuring the necessary monitoring frequency, and achieve long-term stable operation under battery power conditions.

[0071] In summary, after a series of steps including data collection, preprocessing, image recognition, and multi-source fusion, the system can finally deliver reliable water quality information to users and display it through a good interactive interface. In addition, the self-calibration and energy management mechanism throughout the whole process ensures the accuracy and efficiency of the system in long-term operation.

[0072] The intelligent portable water quality detection system of the present invention can be widely used in various scenarios, and is particularly suitable for the following environments and meets their special needs:

[0073] (1) Campus water quality monitoring: It can be used to monitor the water quality of drinking water and domestic water on campus, such as in universities, middle schools, and primary schools. For example, it can be installed at the outlet of the water purification equipment in the school cafeteria to continuously monitor the pH, residual chlorine, and turbidity of the water, ensuring the drinking water safety of teachers and students. Once the indicators are abnormal, the logistics department can be immediately notified to take measures. The portability of this system also facilitates regular inspections of drinking water points on campus, replacing manual sampling and sending for inspection, and improving efficiency. Due to the good campus network conditions, the device can be connected to the campus Wi-Fi to upload data to the campus environmental protection monitoring platform for centralized management. At the same time, the low-maintenance feature of the system reduces the burden on campus operation and maintenance personnel - the self-calibration function ensures that the device remains accurate and reliable after long-term operation without the need to frequently calibrate the sensor. For the relatively stable water quality in the campus environment, this system can set a longer sampling interval to enter the power-saving mode, extending the working time of a single deployment. In short, in the campus scenario where high reliability and ease of use are required and the maintenance manpower is limited, the system of the present invention can provide all-weather intelligent water quality protection at a relatively low cost.

[0074] (2) Water body monitoring in rural areas: For well water, mountain springs, or small-scale centralized water supplies in remote rural areas, this system can give full play to its advantages of low maintenance and easy deployment. Traditional rural water quality monitoring is often difficult to implement due to the lack of electricity and network, while this system is built-in with a battery and long-distance wireless communication (such as NB-IoT / LoRa) and can operate independently without infrastructure. For example, the device can be placed or installed beside the village wells, regularly monitor the water quality, and send the data to the regional water management platform through the cellular Internet of Things. Managers can understand the water quality changes without having to visit the site. In case of heavy rain and floods that may contaminate the well water, the real-time monitoring and rapid alarm functions of the system can help issue a drinking water warning in a timely manner to ensure the safety of villagers. Due to the lack of professional technical personnel on-site in rural areas, the self-calibration ability of the device is particularly important - it automatically eliminates sensor drift during long-term operation, ensuring the credibility of the data. At the same time, the modular protection design of the device can adapt to harsh outdoor conditions (dustproof and splash-proof), reducing the maintenance frequency. Even when maintenance is required, replacing the battery or the sensor probe is very simple and does not require complex tools. In the rural water source scenario where it is scattered, lacks electricity, and lacks network, this system can continuously provide water quality supervision with minimal manual intervention, greatly improving the level of rural drinking water safety monitoring.

[0075] (3) Emergency and disaster response: After sudden environmental pollution accidents or natural disasters, this system can be used as a rapid emergency water quality detection tool to assess the pollution situation of water sources on-site. For example, after earthquakes or floods, the water supply system may be damaged. This system can be carried to the scene by rescue personnel to quickly screen the water quality of temporary water sources and disaster relief water supply points to detect whether harmful pollutants exceed the standard. Since the equipment is ready to use and easy to operate, non-professionals can use it after a short training, saving time for large-scale disaster relief. When suspicious water samples are found, this system can immediately transmit data back to the command center through a 4G satellite portable station to assist in scientific decision-making. This system responds quickly, taking only dozens of seconds from sampling to getting results, and can complete the initial screening of water quality at multiple points in a short time, saving a lot of time compared to traditional laboratory testing. In emergency scenarios, power supply and resupply are unstable. The long-life battery and low-power consumption characteristics of this system ensure that it can still work continuously for several hours without external power, and can be charged through vehicle power or temporary power when necessary for continued use. In addition, the system is designed to be rugged, with good shock resistance and waterproof performance, and can operate normally in the harsh environment of the disaster site. In summary, in emergency monitoring scenarios where rapid equipment deployment and timely and reliable results are required, this system provides an ideal solution.

[0076] Through the adaptability analysis for different scenarios, it can be seen that the water quality detection system of the present invention has scene versatility and flexible deployment capabilities. Whether in a standardized environment such as an urban campus or a wild environment full of uncertainties such as rural areas and disaster areas, this system can meet the monitoring needs with its intelligence, autonomy and portability, providing technical support for various water quality safety applications.

[0077] As Figure 5 shown, the beneficial effects of the present invention are as follows:

[0078] (1) Multi-sensor data fusion, comprehensive and accurate detection: Integrate multiple water quality sensors to simultaneously monitor the temperature, pH, turbidity, conductivity, dissolved oxygen, etc. of the water body. Through data fusion algorithms, comprehensive analysis of the readings of different sensors can be carried out to obtain a more comprehensive water quality evaluation, reducing the impact of single-sensor measurement deviation on the results. The cross-validation of each parameter improves the ability to detect outliers, significantly enhancing the measurement accuracy and reliability.

[0079] (2) Deep learning self-calibration, reducing maintenance requirements: The system is built with an intelligent calibration mechanism based on deep learning, which can adaptively compensate for sensor drift and environmental interference. During long-term use of sensors, zero drift or sensitivity changes may occur. The present invention realizes automatic calibration of sensors by regularly collecting calibration data and using a trained neural network model for comparison and correction. Accurate measurement can be maintained without frequent manual calibration, improving the long-term operation stability of the equipment and reducing the maintenance workload.

[0080] (3) Image recognition for auxiliary judgment to enrich the monitoring dimension: An image recognition module is introduced to intelligently analyze the appearance of water samples and reagent reactions. By using algorithms such as CNN to identify the color of the water body, transparency, or color changes of test strips, it can detect indicators that are difficult to quantify by the naked eye (such as algal growth, color reactions of pollutants, etc.), providing a new basis for water quality assessment. Combining the image recognition results with sensor data helps to detect abnormalities (such as sudden water color changes) and improve the discrimination accuracy, expanding the monitoring dimension of the system.

[0081] (4) Cloud platform integration for remote real-time monitoring: This system supports uploading data to the cloud platform to achieve remote monitoring and centralized data management. Managers can view the water quality data of multiple on-site devices in real time through a computer or mobile phone, realizing the linkage of "cloud + terminal". When the water quality parameters at a monitoring point exceed the standard abnormally, the cloud platform can immediately trigger an alarm to notify relevant personnel, with a faster response. At the same time, the big data stored in the cloud can be used for long-term trend analysis and model optimization to improve the overall monitoring level.

[0082] (5) Portable and integrated design for flexible deployment: The device is small in size and light in weight, integrated with battery power supply and wireless communication, without the need for external power supply and complex wiring. Field personnel can hold it or place it in the on-site water body to quickly carry out detections. The device can be used immediately, significantly shortening the setup time, and is suitable for mobile monitoring needs in different scenarios. The portable design enables it to play a role in scenarios where traditional devices are difficult to cover in a timely manner, such as remote areas and disaster sites.

[0083] (6) Low power consumption and long battery life to adapt to long-term operation in the wild: The system adopts low-power components and an intelligent sleep-wake-up mechanism to minimize energy consumption. In the typical working mode, the device can operate continuously for several hours or even more than a dozen hours, supporting all-weather on-site monitoring requirements. With solar power supply or backup batteries, it can achieve longer-term unattended operation. This high-energy efficiency design meets the requirements for device battery life in remote areas and long-term deployments.

[0084] (7) Real-time response and rapid early warning: With parallel data acquisition by multiple sensors and real-time local embedded AI processing, this system can complete the process from data acquisition to result output within seconds. When water quality parameters mutate, the system almost immediately issues a warning locally and reports it through the network, achieving a rapid response to water quality anomalies, significantly superior to traditional methods that rely on manual sample submission. This real-time nature secures valuable time for emergency decision-making and helps to take timely measures to prevent pollution spread.

[0085] In summary, by organically combining technologies such as multi-sensor fusion, deep learning calibration, and image recognition, the present invention significantly improves the intelligent level and practicality of the water quality detection system, having comprehensive advantages that cannot be compared with traditional technologies.

[0086] In the present specification, the various embodiments are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. For the same or similar parts among the various embodiments, reference can be made to each other.

[0087] In this article, specific examples are used to elaborate on the principles and implementation manners of the present invention. The descriptions of the above embodiments are only used to help understand the method of the present invention and its core idea. At the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. A portable water quality detection system for multi-sensor fusion, self-calibration and image recognition, characterized in that, Including: A sensor module for synchronously collecting physical and chemical parameters of water body to obtain sensor data; An image acquisition module for acquiring images of water samples or reagent reactions to obtain water sample images; A data processing module, connected to the sensor module and the image acquisition module respectively, for self-calibrating the sensor data based on a deep learning algorithm, compensating for sensor drift and environmental interference to obtain preprocessed data, and using a convolutional neural network to extract features and perform pattern recognition on the water sample images acquired by the image acquisition module to obtain image recognition results, and comprehensively analyzing the preprocessed data and the image recognition results through a multimodal fusion algorithm to generate water quality analysis results; A communication module, connected to the data processing module, for uploading the water quality analysis results, the preprocessed data and the image recognition results to an external cloud server and receiving remote control instructions; A display and interaction module, connected to the data processing module, for locally displaying the water quality analysis results, the preprocessed data and the image recognition results and providing a user operation interface; A power module, connected to the data processing module, integrating a rechargeable battery and a power management circuit, for extending the battery life of the data processing module through a low-power consumption strategy; A cloud platform interface module, connected to the communication module and the external cloud server respectively, for realizing data interaction with the cloud server and supporting remote monitoring and big data analysis.

2. The portable water quality detection system for multi-sensor fusion, self-calibration and image recognition according to claim 1, wherein The sensor module includes a pH sensor, a conductivity sensor, a dissolved oxygen sensor, a temperature sensor and a turbidity sensor for real-time monitoring of the acidity, salinity, oxygen content, temperature and turbidity of the water body.

3. The portable water quality detection system for multi-sensor fusion, self-calibration and image recognition according to claim 1, characterized in that, The deep learning algorithm of the data processing module is trained by comparing historical sensor data with standard reference values to dynamically adjust calibration parameters when sensor performance drifts, and a Kalman filtering algorithm is used to fuse multi-sensor data.

4. The portable water quality detection system for multi-sensor fusion, self-calibration and image recognition according to claim 1, characterized in that, The convolutional neural network integrates an attention mechanism, and the convolutional neural network is used to focus on key regions related to water quality in the water sample images; the image recognition results include at least one of reagent colorimetric analysis, abnormal water body color detection or suspended matter distribution evaluation.

5. The portable water quality detection system for multi-sensor fusion, self-calibration and image recognition according to claim 1, characterized in that, The communication module supports multiple wireless transmission protocols; the wireless transmission protocols include Wi-Fi, cellular mobile network, and low-power wide area network; the wireless transmission protocols use encryption protocols to ensure data transmission security.

6. The portable water quality detection system for multi-sensor fusion, self-calibration and image recognition according to claim 1, wherein, The power module reduces energy consumption through dynamic frequency modulation, sensor power supply on demand and intermittent sleep strategies to support the data processing module to continuously operate for more than 12 hours in a preset working mode.

7. The portable water quality detection system for multi-sensor fusion, self-calibration and image recognition according to claim 1, characterized in that, The display and interaction module includes a touch screen and a buzzer; the display and interaction module supports local parameter setting, historical data query and abnormal state alarm, and realizes near-field interaction by connecting with a mobile terminal through Bluetooth.

8. The portable water quality detection system for multi-sensor fusion, self-calibration and image recognition according to claim 1, characterized in that, The cloud platform interface module provides a water quality data visualization interface, supports threshold alarm, remote firmware upgrade and multi-device collaborative management, and generates a long-term trend analysis report.

9. The portable water quality detection system for multi-sensor fusion, self-calibration and image recognition according to claim 1, characterized in that, Both the sensor module and the image acquisition module are detachable structures.

10. The portable water quality detection system for multi-sensor fusion, self-calibration and image recognition according to claim 1, characterized in that, The multimodal fusion algorithm includes at least one of the following methods: Dynamic weighted fusion of sensor data and image recognition results based on Kalman filtering; Using a stacked autoencoder to extract the temporal features of sensor data, and jointly inputting the features of the image recognition results into a deep neural network for non-linear fusion.

Images