A water environment intelligent monitoring system

Abstract

The present invention discloses a water environment intelligent monitoring system, comprising: a camera and a water quality detection detector; a cloud-side intelligent recognition rule library storing intelligent recognition rules; an edge-side intelligent recognition rule library receiving and storing the intelligent recognition rules issued by the cloud-side intelligent recognition rule library; an edge analysis module extracting intelligent recognition rules from the edge-side intelligent recognition rule library, using the intelligent recognition rules to identify the water environment video data to obtain water body characteristics, outputting analysis results of the water body characteristics and / or various water quality indicators according to water environment evaluation rules, displaying and issuing alarms for abnormal results, and sending data that cannot be analyzed and determined to a cloud-side comprehensive analysis module; a cloud-side comprehensive analysis module receiving and analyzing the data sent by the edge analysis module, displaying and issuing alarms for abnormal results. The present invention constructs a video recognition model based on machine learning, combines it with the detector to form a comprehensive water environment monitoring system, and realizes intelligent recognition of the water environment.

Description

Technical Field

[0001] This invention involves the application of the Internet of Things (IoT) and video analysis technology in the field of intelligent water quality monitoring. More precisely, it involves a comprehensive management method for intelligent sewage operation monitoring based on the IoT and video recognition and intelligent analysis technology. This technology has led to the development of a more comprehensive and real-time intelligent water environment monitoring system. Background Art

[0002] Sewage discharged daily by urban residents and many industrial wastewaters contain significant amounts of organic pollutants, leading to severe organic pollution in most urban rivers and reservoirs. This has led to a decrease in urban water resources and increased treatment costs, posing a serious threat to the health of urban residents. The harm caused by water pollution is widespread and long-lasting, often taking considerable time to manifest. Furthermore, water pollution exacerbates water scarcity and deteriorates the ecological environment. Consequently, water environmental governance and the water industry have become increasingly important.

[0003] Routine water management inspections typically follow established inspection plans. Inspectors conduct a variety of inspections, including regular, daily, special, and seasonal, to identify water environment issues within their respective watersheds. Water quality monitoring typically involves periodic manual sampling and testing to obtain current water quality results. Whether performing routine inspections or submitting water quality inspections, these systems often lack real-time monitoring, preventing the rapid detection of water issues and 24 / 7 uninterrupted monitoring of water conditions. Summary of the Invention

[0004] To this end, the present invention provides a comprehensive analytical method for comprehensive water management supervision. Based primarily on the Internet of Things (IoT), machine learning, and video recognition analysis technologies, it acquires and standardizes data on on-site water quality, water levels, and issues such as floating debris in rivers, achieving integrated data collection, data transmission, and data processing, thereby enabling comprehensive monitoring and management of smart water services. The specific solutions employed by the present invention are as follows:

[0005] A water environment intelligent monitoring system, comprising:

[0006] Camera A, installed at each first detection point in the water area, is used to capture water environment video data;

[0007] Water quality detection detector B, installed at each second detection point in the water area, is used to detect various water quality indicators;

[0008] A cloud-based intelligent recognition rule library E2 stores intelligent recognition rules, including water environment evaluation rules and various video recognition models formed through machine learning;

[0009] The edge-side intelligent recognition rule base E1 receives and stores the intelligent recognition rules issued by the cloud-side intelligent recognition rule base E2;

[0010] The edge analysis module D extracts intelligent recognition rules from the edge-side intelligent recognition rule library E1, uses the video recognition model to identify the water environment video data to obtain water body characteristics, outputs analysis results of the water body characteristics and / or various water quality indicators according to the water environment evaluation rules, and transmits abnormal results to the edge-side alarm display module C1 for display and alarm, and sends data that cannot be analyzed and determined to the cloud comprehensive analysis module F;

[0011] The cloud comprehensive analysis module F receives and analyzes the data sent by the edge analysis module D, and transmits the abnormal results to the cloud alarm display module C2 for display and alarm.

[0012] Optionally, the first detection point and the second detection point are not completely the same.

[0013] Optionally, the edge alarm display module C1 and the cloud alarm display module C2 both use sound and / or light to provide alarm prompts.

[0014] Optionally, the cloud alarm display module C2 also receives and displays the alarm sent by the edge alarm display module C1.

[0015] Optionally, the water environment evaluation rules include a first water environment evaluation rule and a second water environment evaluation rule, the first water environment evaluation rule determines a first water environment evaluation level according to various water quality evaluation standards, and the second water environment evaluation rule determines a second water environment evaluation level by performing a graded evaluation on water body characteristics.

[0016] Optionally, a video recognition model identifies water quality by the color of the water surface in the region's rivers. The training process is to extract the water surface colors of the region at different time periods and different water quality levels for machine learning to derive a second water environment evaluation grade for the region's different time periods and different water quality levels.

[0017] Optionally, various water quality detection detectors B are distributed in different areas to detect various factors affecting water quality, including natural water source detectors, urban water supply system detectors, water inlet detectors, and pollution source detectors.

[0018] Optionally, the camera A supports infrared, panoramic 360-degree rotation, and synchronous audio, and the edge analysis module D is integrated into the camera device or arranged independently.

[0019] Optionally, for data that does not match the water environment evaluation rules, the edge analysis module sends an alarm signal to the edge alarm display module C1, and the cloud comprehensive analysis module F sends an alarm signal to the cloud alarm display module C2.

[0020] Optionally, it also includes an edge data storage module G1, which stores water environment video data, various water quality index data collected by the detector, and analysis result data of the edge analysis module D;

[0021] The cloud data storage module G2 is used to store the collected data, analysis results and alarm records transmitted to the cloud.

[0022] This invention combines a video recognition model constructed using machine learning with a detector to form a comprehensive water environment monitoring system. Furthermore, the terminal edge analysis module effectively handles some of the data analysis and storage functions. This approach truly realizes intelligent identification of the water environment, real-time monitoring, and a more accurate water environment monitoring model tailored to local conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 1 is a diagram showing the overall framework of the water environment intelligent monitoring system according to an embodiment of the present invention;

[0024] Figure 2 1 is a flowchart showing the working process of the water environment intelligent monitoring system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0025] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0026] like Figure 1 As shown, the intelligent water environment monitoring system of this embodiment includes a camera A, a water quality detection detector B, an edge analysis module D, an edge-side intelligent recognition rule base E1, a cloud-side comprehensive analysis module F, a cloud-side intelligent recognition rule base E2, an edge data storage module G1, a cloud-side data storage module G2, and a data transmission module H. It is used to monitor the water environment conditions of the water area, including water quality, water level, water surface conditions, etc.

[0027] Multiple cameras A can be installed at primary detection points in the water area, such as rivers, reservoirs, sewage outfalls, and water level markers. Camera A is used to capture real-time scenes, such as water color, floating objects, and water level data. Cameras include, but are not limited to, infrared, 360-degree panoramic, rotatable, and synchronized audio.

[0028] Among them, various water quality detection detectors B are distributed at the second detection point of the water area. The second detection point can be exactly the same as the first detection point, or completely different, or not exactly the same. The water quality detection detector B is used to detect various factors that affect water quality, including but not limited to natural water source detectors (set in the river), urban water supply system detectors (set in the urban water supply system), water outlet detectors (set at the water terminal), and pollution source detectors (set at possible pollution sources). The data collected in this part will be used for analysis by the terminal edge analysis and processing subsystem. The water quality detector can include, for example, a temperature sensor, a turbidity sensor, a conductivity sensor, a microbial sensor, a COD sensor, an NH3-N sensor, a dissolved oxygen sensor, etc., or it can be an integrated detector containing these sensors.

[0029] Among them, edge analysis module D extracts intelligent recognition rules from edge-side intelligent recognition rule base E1. These intelligent recognition rules include multiple video recognition models formed through machine learning, as well as water environment evaluation rules. For some data, after being analyzed by the edge analysis module, only abnormal analysis results and corresponding monitoring data need to be transmitted to the cloud platform, without transmitting qualified analysis results and monitoring data to the cloud comprehensive analysis module for further analysis. For example, if there is a video recognition model that determines the water quality category based on the color of the river surface in a certain area, the water surface color data collected by camera A in this area will be analyzed by edge analysis module D. Data that passes the analysis will not be transmitted to the cloud comprehensive analysis module, while data and results that fail the analysis will be sent to the cloud comprehensive analysis module.

[0030] The video recognition model and related evaluation rules in the edge-side intelligent recognition rule library E1 are all issued by the cloud-based comprehensive analysis module, so the edge-side intelligent recognition rule library is a subset of the cloud-based comprehensive intelligent recognition rule library. The water environment evaluation rules include a first water environment evaluation rule and a second water environment evaluation rule. The first water environment evaluation rule determines a first water environment evaluation level based on various water quality evaluation standards, and the second water environment evaluation rule performs a graded evaluation of water features extracted from the video to determine a second water environment evaluation level. The edge analysis module D can either use the water features extracted from the video alone in combination with the second water environment evaluation rule to evaluate and output the second water environment level, or use the detector detection data alone in combination with the first water environment evaluation rule to evaluate and output the first water environment level. It can also comprehensively consider the evaluation results of the first and second water environment evaluation rules and output a comprehensive water environment level, for example, a weighted summation output. Determining an abnormal result refers to comparing the water environment evaluation level with a set threshold level. If the threshold level is exceeded, it is considered an abnormal result.

[0031] The video recognition model can be trained by machine learning, and the video recognition model can be trained by various data such as training data related to various water colors and water environments, or training data related to the water environment such as the size, number, density of floating objects on the water body. The video recognition model can be trained by extracting images from the video in the form of picture frames for recognition, and determining the second water environment level from the second water environment evaluation rules based on the recognition results.

[0032] There can be multiple video recognition models, such as floating object recognition model, water quality recognition model (through water color analysis), etc. Different second water environment evaluation rules can be set for each video recognition model. For example, for the floating object recognition model, the second water environment evaluation rule can be set according to the size of the floating object. The floating object coverage area is less than 10cm 2 , the number is less than 3 corresponding to one water environment level, and the floating objects cover an area greater than 10cm 2 Less than 30cm 2 , the number is less than 3 corresponding to one water environment level, and the floating objects cover an area greater than 30cm 2 Less than 1m 2 , with more than three corresponding to one water environment level. As the area covered by floating objects increases and their number increases, the corresponding second water environment level decreases. By matching this second water environment evaluation rule with the video recognition model, the second water environment level can be output according to the rule after the video recognition model identifies the result.

[0033] For example, corresponding to the water quality recognition model, the second water environment evaluation rule can be set according to the different colors of the water quality, and the water quality can be identified by the color of the water surface in a certain area. Then the training of the water quality recognition model is to extract the water surface colors of different water quality levels in different time periods of the area and perform machine learning to derive the rules of different levels of water quality in different time periods of the area. The different time periods can be, for example, the four seasons of spring, summer, autumn and winter. The water quality recognition model compares the recognition results with the second water environment evaluation rule to determine the corresponding second water environment level (of course, judging the water quality by the color of the water body is not accurate enough here, and the comprehensive water quality level can also be determined by combining with the data of the detector). Each first water environment evaluation rule and the second water environment evaluation rule are identified and coded, and each specific rule corresponds to a code. And preferably, each rule is also identified as whether it belongs to edge-side intelligent recognition or cloud-side intelligent recognition.

[0034] Detectors at different locations may use the same rule to determine their water environment levels. For example, data from COD sensors at different locations all use the same first water environment evaluation rule to determine the first water environment level.

[0035] Real-time analysis of video images captured by camera A. Since video images are typically large, transmitting them to cloud-based comprehensive analysis module F consumes significant time and storage space. Therefore, edge analysis module D utilizes real-time video data for edge-side recognition whenever possible, rather than transmitting it to cloud-based comprehensive analysis module F for further analysis. Edge analysis module D can be integrated within the camera device or designed as a separate data analysis and acquisition box. The implementation method is not important in this invention, but rather the functionality of the module.

[0036] After receiving relevant data from cameras or detectors, the edge analysis module D first checks the edge-side intelligent recognition rule library E1 to see if local intelligent recognition rules can be applied. If so, it calls the video recognition model and water environment assessment rules for analysis. Normal data after analysis is saved in the data storage module according to the set validity period. If abnormal data is found, the edge alarm display module C1 is called and the relevant alarm content is simultaneously transmitted to the cloud alarm display module C2.

[0037] For example, if camera A captures floating objects on the water surface, the edge analysis module uses its video recognition model to analyze the video frame and determines whether it is abnormal data based on the water environment evaluation rules. For example, if the coverage area of ​​the floating objects on the water surface is greater than 1m 2 If the water level is abnormal, the edge alarm display module C1 will issue an alarm. The alarm result is also transmitted back to the cloud alarm display module C2. After receiving the alarm, the control room staff will check the video to confirm that there is indeed large garbage on the water surface and notify the patrol personnel to clean it up on the spot.

[0038] The cloud-based comprehensive analysis module F primarily analyzes data transmitted by camera A and various water quality sensors that have not yet been processed by the edge analysis module. It uses the cloud-based intelligent recognition rule module E2 as the recognition model and basis for identification. For example, the current recognition model, which uses machine learning, only achieves an accuracy rate of 80%. Relying solely on this model is not very accurate. If a water quality sensor is available at the measured point, the data can be combined with the sensor data for analysis. If the sensor data is unavailable, it is returned to the cloud-based comprehensive analysis module, where control room personnel manually analyze the video data. For another example, sensor data at a monitoring point is easily affected by seasonal and environmental factors, and there is no single, fixed monitoring standard. The cloud-based comprehensive analysis module can use real-time data from the local area, such as rainfall, temperature, and drainage volume, to conduct a comprehensive analysis and refine the rules. This is part of the water technology section and will not be explained in detail here. After processing by the comprehensive analysis module, the water environment level is determined.

[0039] After receiving the data to be processed, the cloud-based comprehensive analysis module F invokes the cloud-side intelligent identification rule E2. If no matching water environment assessment rules are found, an alarm is issued to notify relevant technical personnel for manual analysis. For example, if a detector returns data on the concentration of a chemical component at a sewage outlet, but no relevant rules for that detector are found in the water environment assessment rules, the cloud-based alarm display module C2 is invoked to alert control room personnel through an audio signal that unanalyzable data has been received and requires manual analysis.

[0040] If a matching water environment evaluation rule is found, and the water environment evaluation rule indicates that it needs to be combined with a certain detector data for comprehensive analysis, the data storage module is called to obtain the real-time data of the detector for comprehensive analysis. Of course, when configuring the water environment evaluation rule, you can configure which video recognition model needs to correspond to which second water environment evaluation rule, which detector needs to correspond to which first water environment evaluation rule, and which video recognition model cooperates with which detector to perform comprehensive evaluation. For example, the rule of associating water quality color with the pollutant composition in the water. The cloud-based comprehensive analysis module receives data on abnormal water quality color returned by a camera, for example, the returned color is black, and the video recognition model is called. It needs to be combined with the data of a certain detector for comprehensive analysis. It is found that the manganese content returned by the detector data is seriously exceeded. According to the water quality color and the degree of exceeding the manganese content, the final comprehensive water environment level is obtained and displayed in the form of an alarm.

[0041] Cloud-side intelligent recognition rule base E2: This contains the video recognition model and various water environment assessment rules. Rule base E2 can be customized based on needs. As machine learning covers more recognition scenarios and water quality standards change, it can be updated in parallel. Furthermore, updates to the cloud-side intelligent recognition rule base E2 are synchronized with the edge-side intelligent recognition rule base E1.

[0042] After receiving the data to be processed, the cloud-side comprehensive analysis module F calls the cloud-side intelligent identification rule library E2. If a matching water environment evaluation rule is found, the relevant rules and third-party interfaces are called to analyze the data. If it is abnormal data, the cloud-side alarm display module C2 needs to be called. For example, the cloud-side comprehensive analysis module F receives the water level data of a river, finds the water environment evaluation rules for water level identification in the rule library E2, and marks it as needing to call the three-party interface (current region, current weather, current rainfall, etc.) for comprehensive analysis. For normal data after analysis, it is saved in the data storage module according to the set effective storage time. If it is abnormal data, the alarm display module C2 needs to be called.

[0043] The data storage module G1 is on the cloud side and contains all collected data, analysis results, and alarm records transmitted to the cloud. The data storage module can use a centralized conventional database for storage or a decentralized encryption method based on blockchain.

[0044] The data storage module G2 belongs to the terminal side. It can store some video data within the effective time and the data collected by the detector for the cloud comprehensive analysis module to call.

[0045] The data transmission module H receives various data transmitted back from the edge analysis module, camera and related water quality detectors on the terminal side, and sends the intelligent recognition rules sent from the cloud to the edge analysis module.

[0046] The edge alarm display module C1 records and issues alerts for measured points whose water environment levels fail to meet standards after being detected by the video recognition model and detectors. C1 is deployed on the terminal side. Alarm alerts include, but are not limited to, sound and lighting levels. After processing by the edge analysis module, unqualified data, such as faces appearing on the river surface after video recognition, is returned to the alarm display module. The edge alarm display module C1 then issues an alert in the form of sound. At the same time, the alarms generated by the edge alarm display module C1 need to be synchronously returned to the cloud alarm display module.

[0047] The cloud alarm display module C2 is used to display unqualified data after analysis and processing by the cloud comprehensive analysis module. The cloud alarm display module C2 will provide a combined warning in the form of sound and light. It can also receive alarms from the C1 module.

[0048] The following describes the workflow of the water environment intelligent monitoring system. The cloud-side intelligent recognition rule base E2 synchronizes some rules to the edge-side intelligent recognition rule base E1. After receiving data from the camera and detector, the edge analysis module D goes to the edge-side intelligent recognition rule base E1 to search for water environment evaluation rules. For example, if it is video data, the second water environment evaluation rule is checked for evaluation. If it is detector data, the first water environment evaluation rule is checked for evaluation. Normal data is stored in the data storage module G2, and abnormal data is stored in the edge alarm display module C1.

[0049] If the edge-side intelligent recognition rule base E1 does not have corresponding water environment evaluation rules, the edge analysis module D sends the data to the cloud-side comprehensive analysis module F. The cloud-side comprehensive analysis module F checks the water environment evaluation rules in the cloud-side intelligent recognition rule base E2 and can perform evaluation in combination with the first evaluation rule and the second evaluation rule.

[0050] If the cloud-side intelligent recognition rule library E2 lacks a corresponding water environment assessment rule, an alert is sent to technical personnel for manual analysis. If only sensor data is available, the cloud-based comprehensive analysis module F can access relevant evaluation data through a third-party interface. For example, for water level data sent by a sensor, the cloud-based comprehensive analysis module F can access current weather and rainfall data for the area, then comprehensively analyze the water level and provide a water level assessment.

[0051] Of course, the present invention may have many other embodiments. Without departing from the spirit and essence of the present invention, those skilled in the art may make various corresponding changes and modifications based on the present invention, but these corresponding changes and modifications shall fall within the scope of protection of the claims of the present invention.

Claims

1. A water environment intelligent monitoring system, characterized in that: include: Camera A, installed at each first detection point in the water area, is used to capture water environment video data; Water quality detection detector B, installed at each second detection point in the water area, is used to detect various water quality indicators; A cloud-based intelligent recognition rule library E2 stores intelligent recognition rules, including water environment evaluation rules and various video recognition models formed through machine learning; The edge-side intelligent recognition rule base E1 receives and stores the intelligent recognition rules issued by the cloud-side intelligent recognition rule base E2; The edge analysis module D extracts intelligent recognition rules from the edge-side intelligent recognition rule library E1, uses the video recognition model to identify the water environment video data to obtain water body characteristics, outputs analysis results of the water body characteristics and / or various water quality indicators according to the water environment evaluation rules, and transmits abnormal results to the edge-side alarm display module C1 for display and alarm, and sends data that cannot be analyzed and determined to the cloud comprehensive analysis module F; The cloud comprehensive analysis module F receives and analyzes the data sent by the edge analysis module D, and transmits the abnormal results to the cloud alarm display module C2 for display and alarm. The water environment evaluation rules include a first water environment evaluation rule and a second water environment evaluation rule. The first water environment evaluation rule determines a first water environment evaluation grade according to various water quality evaluation standards, and the second water environment evaluation rule determines a second water environment evaluation grade by performing a graded evaluation on water body characteristics. Different second water environment evaluation rules are set corresponding to each video recognition model.

2. The water environment intelligent monitoring system according to claim 1, characterized in that: The first detection point and the second detection point are not completely the same.

3. The water environment intelligent monitoring system according to claim 1, characterized in that: The edge alarm display module C1 and the cloud alarm display module C2 both use sound and / or light to issue alarm prompts.

4. The water environment intelligent monitoring system according to claim 1, characterized in that: The cloud alarm display module C2 also receives and displays the alarms sent by the edge alarm display module C1.

5. The water environment intelligent monitoring system according to claim 1, characterized in that: A video recognition model identifies water quality by the color of the river surface in a region. Its training process is to extract the water surface colors of the region at different time periods and different water quality levels for machine learning to derive the second water environment evaluation level of the region at different time periods and different water quality levels.

6. The water environment intelligent monitoring system according to claim 1, characterized in that: Various water quality detection detectors B are distributed in different areas and are used to detect various factors that affect water quality, including natural water source detectors, urban water supply system detectors, water outlet detectors, and pollution source detectors.

7. The water environment intelligent monitoring system according to claim 1, characterized in that: The camera A supports infrared, panoramic 360-degree rotation, and synchronous audio, and the edge analysis module D is integrated into the camera device or arranged independently.

8. The water environment intelligent monitoring system according to claim 1, characterized in that: For data that does not match the water environment evaluation rules, the edge analysis module sends an alarm signal to the edge alarm display module C1, and the cloud comprehensive analysis module F sends an alarm signal to the cloud alarm display module C2.

9. The water environment intelligent monitoring system according to claim 1, characterized in that: It also includes an edge data storage module G1, which stores water environment video data, various water quality index data collected by the detector, and analysis result data of the edge analysis module D; The cloud data storage module G2 is used to store the collected data, analysis results and alarm records transmitted to the cloud.

Images