A video multi-parameter monitoring device for drainage pipe networks based on AI edge computing
By integrating water quality sensors and video acquisition modules into the drainage network, the AI edge computing device solves the problems of false alarms in data matching and unstable communication in the existing system, and realizes efficient pollution monitoring and source tracing capabilities.
Patent Information
- Application Number
- CN202611153764.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-31
- Publication Date
- 2026-11-10
AI Technical Summary
In existing drainage network monitoring systems, water quality monitoring terminals and video surveillance equipment are deployed independently, resulting in inaccurate spatiotemporal matching of data, frequent false alarms, loss of early warning capabilities when communication signals are unstable, and traditional systems relying on cloud analysis leading to excessive bandwidth and storage loads, making it impossible to identify low-concentration visible pollution.
An integrated device based on AI edge computing is adopted, which integrates water quality sensors, video acquisition modules and edge computing controllers to achieve local fusion analysis of multimodal data. The AI inference chip performs real-time data verification and anomaly detection, and only uploads key data to reduce communication and storage burden.
It enables accurate differentiation between contamination and false alarms, reduces communication and storage costs, adapts to complex downhole environments, forms a credible digital evidence chain, and supports rapid contamination source tracing and emergency response.
Smart Images

Figure CN123065518A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of drainage network monitoring technology, and more specifically to a drainage network video multi-parameter monitoring device based on AI edge computing. Background Technology
[0002] Drainage pipe networks are the core nodes for water environment risk prevention and control. Deploying online monitoring equipment at key points in the pipe network can provide real-time early warning of water quality anomalies and support pollution source tracing. This plays a crucial role in improving the efficiency of emergency response to sudden water pollution and safeguarding regional water environment safety.
[0003] However, most current mainstream monitoring solutions deploy water quality monitoring terminals and video surveillance equipment separately. The clocks and communication links of these two sets of equipment are independent, making it impossible to accurately match water quality data with on-site images in time and space. After an alarm is triggered, the actual scene cannot be verified simultaneously, and it is difficult to quickly distinguish between false alarms caused by sensor malfunctions or disturbances in the underground environment, significantly slowing down the progress of pollution source tracing and disposal. Furthermore, existing systems rely entirely on cloud-based intelligent data analysis; uploading the complete raw video stream results in huge bandwidth and cloud storage loads. Traditional water quality sensors can only detect compliant physicochemical indicators and cannot identify low-concentration visual pollution such as oil films and water discoloration. Simultaneously, the entire system is highly dependent on the network; fluctuations in underground pipeline communication signals directly cause a loss of early warning capabilities, resulting in poor adaptability to complex underground working conditions.
[0004] Therefore, how to provide a video multi-parameter monitoring device that integrates AI edge computing capabilities and enables local fusion analysis of multimodal data is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention
[0005] In view of this, the present invention provides a video multi-parameter monitoring device for drainage pipe networks based on AI edge computing, which solves the above problems.
[0006] To achieve the above objectives, the present invention adopts the following technical solution: This invention discloses a video multi-parameter monitoring device for drainage pipe networks based on AI edge computing, comprising: a protective shell, a mounting bracket, a multi-parameter water quality sensor group, a video acquisition module, a supplementary lighting device, an edge computing controller, a communication module, and a power supply module; The protective housing is a sealed box, which is internally divided into an independent equipment compartment and a sensor compartment. It is fixed to the inside of the drainage network inspection well by the mounting bracket. The multi-parameter water quality sensor group is installed in the sensor compartment at the bottom of the protective housing; the probe of each sensor extends into the water in the pipe, and the signal output terminal of each sensor is connected to the first data input terminal of the edge computing controller. The video acquisition module and the supplementary lighting device are fixedly installed in the viewing window at the front end of the protective housing. The video signal output terminal of the video acquisition module is connected to the second data input terminal of the edge computing controller, and the control signal input terminal of the supplementary lighting device is connected to the first control output terminal of the edge computing controller. The edge computing controller is located inside the equipment compartment and is used to locally fuse water quality data and video image data, and perform AI analysis. The communication module is electrically connected to the second control output terminal of the edge computing controller and is used to upload the data and analysis results of the edge computing controller. The power supply module is fixedly installed inside the equipment compartment and provides unified power to the multi-parameter water quality sensor group, the edge computing controller, the video acquisition module, the supplementary lighting device, and the communication module.
[0007] Furthermore, the mounting bracket is an adjustable angle universal bracket, with its first end fixedly connected to the back of the protective shell and its second end fixedly connected to the manhole wall or the inner wall of the pipe, supporting and fixing the entire device at a predetermined height so that the collection window at the front end of the device faces the direction of the incoming water from the pipe.
[0008] Furthermore, the multi-parameter water quality sensor group includes a pH sensor, a COD sensor, an ammonia nitrogen sensor, and a conductivity sensor. Each sensor probe penetrates the sensor chamber and is immersed in the water body to collect raw water quality data of the water body in the pipeline in real time and transmit it to the edge computing controller. The raw water quality data includes pH value, COD concentration, ammonia nitrogen concentration, and conductivity parameter.
[0009] Furthermore, the multi-parameter water quality sensor group also includes a radar level gauge, a pressure level gauge, and a Doppler flow meter; the antenna of the radar level gauge is installed at the bottom of the protective housing, facing the water surface, for non-contact measurement of the water level; the probe of the pressure level gauge extends into the water through the sensor chamber for water pressure monitoring; the probe of the Doppler flow meter is placed in the water flow channel of the sensor chamber for flow monitoring.
[0010] Furthermore, the video acquisition module is an industrial-grade high-definition network camera, and the video signal output terminal of the video acquisition module is electrically connected to the second data input terminal of the edge computing controller via a coaxial cable or network cable; The front end of the viewing window is equipped with a transparent protective lens; The supplementary lighting device is an LED array arranged around the lens. It automatically turns on according to the instructions of the edge computing controller when the light is insufficient, providing auxiliary lighting for video acquisition. The lamp body of the LED array is fixed inside the viewing window cabin.
[0011] Furthermore, the edge computing controller integrates a microprocessor, an AI inference chip, and a storage unit; The storage unit is used to receive and temporarily store the raw water quality data collected by the multi-parameter water quality sensor group; and to receive and temporarily store the video image data collected by the video acquisition module. The AI inference chip performs local real-time analysis of video image data through a lightweight visual semantic segmentation and target detection model to identify visual pollution features; The microprocessor performs local correlation analysis between water quality data and visual recognition results to determine whether a pollution event has occurred and the type of event; and sends corresponding control commands to the supplementary lighting device and communication module based on the determination results; the event types include: real pollution events, suspected sensor malfunctions or interference, and visual alarm events.
[0012] Furthermore, the edge computing controller uniformly drives all sensors and video acquisition modules, assigning consistent timestamps and spatial tags to all water quality data and video image data; and synchronously storing them to the local storage unit to form a spatiotemporally unified digital evidence chain.
[0013] Furthermore, the communication module is a 4G / 5G / NB-IoT wireless communication module, which is integrated on the edge computing controller circuit board or electrically connected to the edge computing controller circuit board via a board-to-board connector; the antenna of the communication module is led out of the protective housing via an RF cable. The communication module is used to upload the fusion judgment results, key frame images or short video clips generated by the edge computing controller to the remote monitoring platform.
[0014] Furthermore, the power supply module is a rechargeable lithium battery pack or a combination of a battery pack and a solar charging controller, which is fixedly installed inside the equipment compartment.
[0015] Furthermore, the equipment compartment, the sensor compartment, and the viewing window compartment are isolated from each other; all external wiring locations are equipped with waterproof aviation plugs and waterproof sealed interfaces.
[0016] As can be seen from the above technical solution, compared with the prior art, the present invention provides a multi-parameter video monitoring device for drainage pipe networks based on AI edge computing, which has the following beneficial effects: This invention employs an integrated sealed housing that combines water quality sensing, video acquisition, and edge computing units. Leveraging a built-in AI inference chip, it performs multimodal fusion analysis of water quality physicochemical data and video footage locally. Through dual verification using numerical indicators and visual features, it can accurately distinguish between real pollution, sensor disturbances, and purely visual anomalies, significantly reducing invalid alarms and minimizing the waste of emergency resources. AI visual semantic recognition can capture low-concentration visible pollution, such as oil films and colored wastewater, which traditional sensors cannot detect, filling blind spots in pipeline monitoring and identifying early signs of pollution. The device only uploads key frames and feature data of anomalies, eliminating the need to transmit the full original video stream, significantly reducing downhole communication bandwidth usage and cloud storage costs. Simultaneously, the edge controller, equipped with a local storage unit, can continuously monitor and cache abnormal evidence during network outages, automatically re-transmitting data after network recovery, thus reducing reliance on the cloud and stable networks. All sensing and video modules are driven by a single controller with unified timing, ensuring data and images are naturally synchronized, forming a complete and reliable digital evidence chain, facilitating pollution source tracing and regulatory enforcement accountability.
[0017] This invention relies on AI edge computing to achieve multimodal local intelligent analysis, taking into account monitoring accuracy, transmission cost and adaptability to complex downhole environments. Attached Figure Description
[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.
[0019] Figure 1 This is a schematic diagram of the overall structure of the present invention.
[0020] In the diagram: 1. Protective housing; 2. Mounting bracket; 3. Multi-parameter water quality sensor group; 4. Video acquisition module; 5. Lighting device; 6. Edge computing controller; 7. Communication module; 8. Power supply module. Detailed Implementation
[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0022] This invention discloses a video multi-parameter monitoring device for drainage pipe networks based on AI edge computing, such as... Figure 1As shown, it includes: a protective housing 1, a mounting bracket 2, a multi-parameter water quality sensor group 3, a video acquisition module 4, a supplementary lighting device 5, an edge computing controller 6, a communication module 7, and a power supply module 8; The protective housing 1 is a sealed box made of corrosion-resistant material, which is divided into an independent equipment compartment and a sensor compartment. It is fixed to the inside of the drainage network inspection well by the mounting bracket 2. The multi-parameter water quality sensor group 3 is installed in the sensor compartment at the bottom of the protective housing; the probes of each sensor extend into the water in the pipe, and the signal output terminals of each sensor are connected to the first data input terminal of the edge computing controller through a waterproof aviation plug; The video acquisition module 4 and the supplementary lighting device 5 are fixedly installed in the viewing window compartment at the front of the protective housing. The video signal output terminal of the video acquisition module 4 is connected to the second data input terminal of the edge computing controller, and the control signal input terminal of the supplementary lighting device 5 is connected to the first control output terminal of the edge computing controller through a waterproof sealing interface. The edge computing controller 6 is located inside the equipment compartment and is used to locally fuse water quality data and video image data, and perform AI analysis. The communication module 7 is electrically connected to the second control output terminal of the edge computing controller 6, and is used to upload data and analysis results from the edge computing controller. The power supply module 8 is fixedly installed in the equipment compartment and provides unified power supply for the multi-parameter water quality sensor group 3, edge computing controller 6, video acquisition module 4, supplementary lighting device 5 and communication module 7.
[0023] In one specific embodiment, the mounting bracket 2 is an adjustable angle universal bracket, with its first end fixedly connected to the back of the protective shell and its second end fixedly connected to the manhole wall or the inner wall of the pipe, supporting and fixing the entire device at a predetermined height so that the collection window at the front end of the device faces the direction of the incoming water from the pipe.
[0024] In one specific embodiment, the multi-parameter water quality sensor group 3 includes a pH sensor, a COD sensor, an ammonia nitrogen sensor, and a conductivity sensor. Each sensor probe penetrates the sensor chamber and is immersed in the water body to collect raw water quality data of the water body in the pipeline in real time and transmit it to the edge computing controller. The raw water quality data includes pH value, COD concentration, ammonia nitrogen concentration, and conductivity parameter.
[0025] In one specific embodiment, the multi-parameter water quality sensor group 3 also includes a radar water level gauge, a pressure water level gauge, and a Doppler flow meter; the antenna of the radar water level gauge is installed at the bottom of the protective housing, facing the water surface, for non-contact measurement of water level; the probe of the pressure water level gauge extends into the water through the sensor chamber for water pressure monitoring; the probe of the Doppler flow meter is placed in the water flow channel of the sensor chamber for flow monitoring.
[0026] In one specific embodiment, the video acquisition module 4 is an industrial-grade high-definition network camera, and the video signal output terminal of the video acquisition module 4 is electrically connected to the second data input terminal of the edge computing controller through a coaxial cable or a network cable. A transparent protective lens is installed at the front of the viewing window; The supplementary lighting device 5 is a high-brightness LED array arranged around the lens. When the light is insufficient, it will automatically turn on according to the instructions of the edge computing controller to provide auxiliary lighting for video acquisition. The lamp body of the LED array is fixed inside the viewing window cabin.
[0027] In one specific embodiment, the edge computing controller 6 integrates a microprocessor, an AI inference chip, and a storage unit; The storage unit is used to receive and temporarily store the raw water quality data collected by the multi-parameter water quality sensor group; and to receive and temporarily store the video image data collected by the video acquisition module. The AI inference chip uses a lightweight visual semantic segmentation and target detection model to perform local real-time analysis of video image data and identify visual pollution features such as oil film, color anomalies, and floating objects. The microprocessor performs local correlation analysis between water quality data and visual recognition results to determine whether a pollution event has occurred and the type of event; and sends corresponding control commands to the supplementary lighting device and communication module based on the judgment results; the event types include: real pollution events, suspected sensor malfunctions or interference, and visual alarm events.
[0028] Specifically, the edge computing controller integrates an AI inference chip, enabling local correlation analysis of water quality and visual data. When water quality data exceeds standards but visual data is normal, the device can determine it as sensor interference or malfunction, and either not report it or only record it at a low level. When visual data detects an anomaly but water quality data does not exceed standards, the device can proactively report a visual warning, filtering out invalid alarms. In traditional solutions, all data is uploaded to the cloud before judgment, and the cloud can only see a single-dimensional value, unable to distinguish between actual pollution and sensor interference. This invention integrates the AI inference chip locally, giving the device the ability to perform multimodal data fusion analysis. When data exceeds standards, the device can immediately retrieve visual data from the same moment for cross-validation—if the visual data is normal, it indicates that the sensor may be covered by silt or experiencing momentary interference; if the visual data is also abnormal, it confirms actual pollution. Through this "multimodal dual verification" mechanism, the problem of false alarms is fundamentally solved, ensuring that every alarm is worth responding to.
[0029] The AI inference chip incorporates a lightweight visual semantic segmentation model, enabling pixel-level understanding of real-time video streams. It proactively identifies pollution features that traditional sensors cannot detect, such as oil film coverage areas, water color gradients, and abnormal floating object outlines. Even if pollutant concentrations do not reach the detection limits of water quality sensors (e.g., colorless transparent oil films, low-concentration dyed wastewater), the AI visual model can still achieve accurate identification based on visual semantic features such as morphology, gloss, and motion trajectory, expanding the perception dimension from "physicochemical parameters" to "visual semantics." Traditional water quality monitoring technologies, limited by sensor principles, can only respond to specific physicochemical indicators (e.g., pH, COD, ammonia nitrogen). For pollutants such as oils, colored dyes, and suspended particles, a considerable concentration often needs to accumulate before being detected by sensors, by which time the pollution has often already spread. This invention introduces AI visual understanding capabilities into the monitoring front end, enabling the device to "understand" water anomalies. AI visual models do not rely on concentration thresholds. Instead, they learn the visual characteristics of a large number of pollution samples (spectral reflectance characteristics of oil films, color spatial distribution of colored wastewater, and edge morphology of floating objects) to achieve keen perception of early pollution events, fill the monitoring blind spots that traditional sensors cannot cover, and gain valuable lead time for emergency response.
[0030] After the edge computing controller completes the analysis and feature extraction of multimodal data locally, it only uploads the feature data and key frames (KB level) of abnormal events through the communication module, instead of the entire raw data stream (GB level). Compared to traditional solutions that require uploading all raw data to the cloud, this invention can significantly reduce the amount of data transmitted. Traditional monitoring solutions require uploading all raw data (including massive video streams) to the cloud, which puts a lot of pressure on network bandwidth and cloud storage, especially in environments with poor 4G signals underground, where transmitting large amounts of data is inefficient and uneconomical. Traditional monitoring solutions require uploading all raw data (including continuous video streams) to cloud servers for processing and storage, which puts enormous pressure on network bandwidth and cloud storage resources. Especially in underground pipeline environments, 4G / 5G communication signals are often unstable, and transmitting massive amounts of raw data presents both technical bottlenecks and high communication costs. This invention integrates the AI inference chip locally on the device, enabling the data analysis and feature extraction process to be completed at the data source, and only uploading the analysis results and key evidence to the cloud. By adopting an "edge-device collaboration" architecture, the cloud computing power requirements are pushed down to the terminal devices, fundamentally solving the bottleneck problem of large-scale raw data transmission and significantly reducing the dependence on network bandwidth and cloud storage costs.
[0031] Even when the communication network is interrupted, the edge computing controller can still operate normally, continuously performing local data acquisition, video analysis, and anomaly detection. Anomaly data and keyframe images are temporarily stored in the storage unit and automatically re-uploaded after the network is restored. The device has independent operating capabilities and is unaffected by network conditions. Traditional monitoring devices, however, are highly dependent on network connectivity—raw data must be uploaded to the cloud for analysis and processing, and video data must also be uploaded to the cloud for viewing and assessment. Once the network is interrupted, such devices lose their basic monitoring and early warning functions. This invention integrates intelligent analysis functions locally on the device; the network only handles the function of reporting results. Even in the event of a network interruption, the device can still normally complete the entire process of monitoring, identifying, and recording abnormal events. After the network is restored, the system automatically re-uploads the abnormal data stored during the network outage to the backend. This is suitable for environments with unstable communication signals, such as underground pipelines, fundamentally improving the system's reliability and environmental adaptability.
[0032] In one specific embodiment, the edge computing controller 6 uniformly drives all sensors and video acquisition modules, assigning consistent timestamps and spatial tags to all water quality data and video image data, so that the multimodal data generated by the AI inference chip has completely consistent timestamps and spatial tags; and synchronously stores it to the local storage unit to form a spatiotemporally unified digital evidence chain.
[0033] Specifically, because all sensing modules are driven by the same edge computing controller, the multimodal data generated by the AI inference chip has completely consistent timestamps and spatial tags. When post-event tracing or liability determination is required, investigators can directly retrieve the records stored in the device to obtain original multimodal data that is perfectly matched in time and space and labeled by AI, forming a complete and credible digital evidence chain. In contrast, in traditional monitoring solutions, data from different modalities come from different devices. Even if correlation is established at the software level, its integrity, originality, and credibility may still be questionable—data can be tampered with, videos can be edited, and timestamps from different devices may deviate. This invention ensures, from a hardware structure perspective, that multimodal data is of the same source, same entity, and synchronized from the initial collection. The AI inference chip completes identification and labeling locally at the same time the data is generated. All original data and AI identification results are stored in the same device, forming a physical binding relationship. This "source physical uniformity" based on AI edge computing simplifies the process of building the evidence chain, making the resulting conclusions more objective, coherent, and credible, effectively supporting the technical needs of regulatory enforcement and incident liability determination.
[0034] In one specific embodiment, the communication module 7 is a 4G / 5G / NB-IoT wireless communication module, which is integrated on the edge computing controller circuit board or electrically connected to the edge computing controller circuit board via a board-to-board connector; the antenna of the communication module is led out of the protective housing via an RF cable. The communication module 7 is used to upload the fusion judgment results, key frame images or short video clips generated after processing by the edge computing controller to the remote monitoring platform.
[0035] In one specific embodiment, the power supply module 8 is a rechargeable lithium battery pack or a combination of a battery pack and a solar charging controller, which is fixedly installed in the equipment compartment; the power output terminal of the power supply module 8 is electrically connected to the power input terminal of the edge computing controller, the power input terminal of the multi-parameter water quality sensor group, the power input terminal of the video acquisition module, the power input terminal of the supplementary lighting device, and the power input terminal of the communication module, respectively, to provide working power for each component.
[0036] In one specific embodiment, the equipment compartment, sensor compartment, and viewing window compartment are isolated from each other; all external wiring locations are equipped with waterproof aviation plugs and waterproof sealed interfaces to achieve waterproof protection downhole.
[0037] In one specific embodiment, when wastewater flows from upstream of the pipeline through the monitoring area of the device, the various modules of the device work together in the following manner: (1) Routine monitoring phase: A multi-parameter water quality sensor array collects water quality parameters in real time and sends the data to the edge computing controller. The video acquisition module collects video images in real time and sends the video stream to the edge computing controller. The AI inference chip built into the edge computing controller performs real-time analysis of the video stream and continuously identifies visual features such as oil film, color anomalies, and floating objects. The edge computing controller performs local comparison and analysis between the water quality data and the visual recognition results. If no anomalies are found, the edge computing controller controls the communication module to only upload basic hydrological data (heartbeat packets) periodically to maintain low power consumption.
[0038] (2) Abnormal Triggering Phase: When the edge computing controller detects any of the following situations through local analysis: water quality data exceeds a preset threshold and video identifies visual anomalies (oil film / color abnormality / floating matter), it is determined to be a "real pollution event"; water quality data exceeds a preset threshold, but the video image is normal, it is determined to be a "suspected sensor malfunction or interference", and the alarm level is reduced; water quality data is normal, but video identifies visual anomalies, it is determined to be a "visual alarm event", and the edge computing controller immediately performs the following linkage operations: sends an activation command to the supplementary lighting device to enhance on-site lighting; sends an upload command to the communication module to package and upload the judgment result, the data exceeding the standard, key frame images or short video clips to the remote monitoring platform.
[0039] (3) Emergency Response Phase: The remote monitoring platform receives and displays the fused data packets uploaded by the device, allowing command personnel to view on-site images and data with zero latency and make judgments and decisions. If the event continues, the device continues to monitor and upload updated data. When the water quality data returns to normal and the video anomaly disappears, the edge computing controller determines that the event has ended, stops alarms and uploads, and resumes normal monitoring mode.
[0040] (4) Operating mode under network outage conditions: When the communication network is interrupted, the edge computing controller continues to operate normally, continuously performing local data acquisition, video analysis, and anomaly detection. If a contamination event occurs, the edge computing controller will temporarily store the abnormal data and key frame images in the storage unit. When the network is restored, the edge computing controller will automatically retransmit the temporarily stored abnormal data to the remote monitoring platform.
[0041] The device of this invention works in collaboration with the backend system, and its workflow follows the steps of "routine monitoring, anomaly triggering, joint evidence collection, and decision support," providing closed-loop support for emergency command. The complete workflow is as follows: 1. System initialization: The integrated monitoring device deployed at key nodes of the drainage network is powered on and started. The built-in data acquisition controller, various sensors and video modules complete self-tests and enter the real-time monitoring standby state.
[0042] 2. Local Sensing: When polluted water flows through the device's monitoring points, the water quality and hydrological sensors are the first to detect changes in concentration and liquid level. The data acquisition controller within the device compares the collected values with preset alarm thresholds in real time. Once it is confirmed that the water quality parameters continuously exceed the standards or the liquid level is abnormal, an alarm is triggered.
[0043] 3. Linked Evidence Collection: Upon triggering the alarm, a local linkage command can be executed to send an activation signal to the supplementary lighting, enhancing on-site illumination, and to control the high-definition camera to switch to high-definition recording or high-frequency snapshot mode. The on-site video footage recorded by the camera and the water quality exceeding the standard data at the moment of triggering are linked and bound together by the same controller within the device to ensure strict time and space synchronization.
[0044] 4. Data Reporting: The data acquisition controller will prioritize uploading data such as trigger time, location information, exceedance data, and video clips to the "Emergency Map for Sudden Water Pollution Incidents" backend system via the communication module. After receiving the data, the backend system will accurately locate the alarm point on the electronic map and automatically merge the reported real-time data curves with the video footage synchronously transmitted from the device on the same interface to generate a preliminary pollution incident situation.
[0045] 5. Emergency Assessment and Decision Support: Command personnel conduct assessments based on the fusion of "data + video" information provided by the device. They can directly determine whether the alarm is due to equipment malfunction or a false alarm through video, and visually assess the form, color, and diffusion of pollutants. Based on the assessment conclusions, command personnel can utilize the backend system for dispatching (e.g., notifying upstream companies to limit emissions).
[0046] 6. Alarm cancellation and status restoration: When the device detects that the water quality data has continued to fall below the safety threshold, it can determine that the event has ended, support the device to stop the evidence collection mode, turn off the supplementary light, and resume normal inspection of the camera.
[0047] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.
[0048] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A multi-parameter video monitoring device for drainage pipe networks based on AI edge computing, characterized in that, include: The system includes a protective housing, mounting brackets, a multi-parameter water quality sensor array, a video acquisition module, a supplementary lighting device, an edge computing controller, a communication module, and a power supply module. The protective housing is a sealed box, which is internally divided into an independent equipment compartment and a sensor compartment. It is fixed to the inside of the drainage network inspection well by the mounting bracket. The multi-parameter water quality sensor group is installed in the sensor compartment at the bottom of the protective housing; the probe of each sensor extends into the water in the pipe, and the signal output terminal of each sensor is connected to the first data input terminal of the edge computing controller. The video acquisition module and the supplementary lighting device are fixedly installed in the viewing window at the front end of the protective housing. The video signal output terminal of the video acquisition module is connected to the second data input terminal of the edge computing controller, and the control signal input terminal of the supplementary lighting device is connected to the first control output terminal of the edge computing controller. The edge computing controller is located inside the equipment compartment and is used to locally fuse water quality data and video image data, and perform AI analysis. The communication module is electrically connected to the second control output terminal of the edge computing controller and is used to upload the data and analysis results of the edge computing controller. The power supply module is fixedly installed inside the equipment compartment and provides unified power to the multi-parameter water quality sensor group, the edge computing controller, the video acquisition module, the supplementary lighting device, and the communication module.
2. The drainage network video multi-parameter monitoring device based on AI edge computing according to claim 1, characterized in that, The mounting bracket is an adjustable angle universal bracket. Its first end is fixedly connected to the back of the protective shell, and its second end is fixedly connected to the manhole wall or the inner wall of the pipe, supporting and fixing the entire device at a predetermined height so that the collection window at the front end of the device faces the direction of the incoming water from the pipe.
3. The drainage network video multi-parameter monitoring device based on AI edge computing according to claim 1, characterized in that, The multi-parameter water quality sensor group includes a pH sensor, a COD sensor, an ammonia nitrogen sensor, and a conductivity sensor. Each sensor probe penetrates the sensor chamber and is immersed in the water body to collect raw water quality data of the water body in the pipeline in real time and transmit it to the edge computing controller. The raw water quality data includes pH value, COD concentration, ammonia nitrogen concentration, and conductivity parameter.
4. The drainage network video multi-parameter monitoring device based on AI edge computing according to claim 3, characterized in that, The multi-parameter water quality sensor group also includes a radar level gauge, a pressure level gauge, and a Doppler flow meter; the antenna of the radar level gauge is installed at the bottom of the protective housing, facing the water surface, for non-contact measurement of water level; the probe of the pressure level gauge extends into the water through the sensor chamber for water pressure monitoring; the probe of the Doppler flow meter is placed in the water flow channel of the sensor chamber for flow monitoring.
5. The drainage network video multi-parameter monitoring device based on AI edge computing according to claim 1, characterized in that, The video acquisition module is an industrial-grade high-definition network camera. The video signal output terminal of the video acquisition module is electrically connected to the second data input terminal of the edge computing controller via a coaxial cable or network cable. The front end of the viewing window is equipped with a transparent protective lens; The supplementary lighting device is an LED array arranged around the lens. It automatically turns on according to the instructions of the edge computing controller when the light is insufficient, providing auxiliary lighting for video acquisition. The lamp body of the LED array is fixed inside the viewing window cabin.
6. The drainage network video multi-parameter monitoring device based on AI edge computing according to claim 1, characterized in that, The edge computing controller integrates a microprocessor, AI inference chip, and storage unit; The storage unit is used to receive and temporarily store the raw water quality data collected by the multi-parameter water quality sensor group; and to receive and temporarily store the video image data collected by the video acquisition module. The AI inference chip performs local real-time analysis of video image data through a lightweight visual semantic segmentation and target detection model to identify visual pollution features; The microprocessor performs local correlation analysis between water quality data and visual recognition results to determine whether a pollution event has occurred and the type of event. Based on the judgment result, corresponding control commands are sent to the supplementary lighting device and communication module; the event types include: real pollution events, suspected sensor malfunctions or interference, and visual alarm events.
7. The drainage network video multi-parameter monitoring device based on AI edge computing according to claim 1, characterized in that, The edge computing controller uniformly drives all sensors and video acquisition modules, assigning consistent timestamps and spatial tags to all water quality data and video image data; and synchronously storing them to the local storage unit to form a spatiotemporally unified digital evidence chain.
8. The drainage network video multi-parameter monitoring device based on AI edge computing according to claim 1, characterized in that, The communication module is a 4G / 5G / NB-IoT wireless communication module, which is integrated on the edge computing controller circuit board or electrically connected to the edge computing controller circuit board via a board-to-board connector; the antenna of the communication module is led out of the protective housing via an RF cable. The communication module is used to upload the fusion judgment results, key frame images or short video clips generated by the edge computing controller to the remote monitoring platform.
9. The drainage network video multi-parameter monitoring device based on AI edge computing according to claim 1, characterized in that, The power supply module is a rechargeable lithium battery pack or a combination of a battery pack and a solar charging controller, and is fixedly installed inside the equipment compartment.
10. The drainage network video multi-parameter monitoring device based on AI edge computing according to claim 1, characterized in that, The equipment compartment, the sensor compartment, and the viewing window compartment are isolated from each other; all external wiring locations are equipped with waterproof aviation plugs and waterproof sealed interfaces.
