A method and system for monitoring progress of sewage treatment

Abstract

This invention discloses a method and system for monitoring wastewater treatment progress, relating to the field of wastewater treatment technology. It includes a panoramic image acquisition module for real-time acquisition of panoramic images of wastewater discharge ponds, providing basic data for feature region analysis; a feature region analysis module for marking high-pollution-risk areas based on historical sampling data and image processing technology; a dynamic sampling point optimization module for automatically adjusting sampling point positions and cycle counts based on the feature region distribution obtained from the feature region analysis module; a multi-source data fusion module for integrating sensor data, image data, and historical data to construct an LSTM neural network model; and a wastewater treatment progress assessment module for assessing treatment progress based on the output results of the LSTM neural network model. This invention, through the feature region analysis module, enables the dynamic sampling point optimization module to automatically adjust the sampling point positions and cycle counts based on the feature region analysis results, ensuring the representativeness of the monitoring data.

Description

Technical Field

[0001] This invention relates to the field of wastewater treatment technology, and in particular to a method and system for monitoring the progress of wastewater treatment. Background Technology

[0002] Wastewater treatment refers to the process of removing pollutants such as suspended solids, organic matter, heavy metals, and pathogens from wastewater through physical, chemical, and biological methods, so that it meets national discharge standards or reuse requirements. Water treatment progress monitoring is the process of tracking changes in key parameters in real time during the treatment process through technologies such as sensors, image recognition, and data analysis, assessing whether the treatment efficiency meets the standards, and predicting the remaining treatment time.

[0003] Conventional methods typically employ fixed-location sampling, which cannot adapt to the dynamic changes in the composition of wastewater in time and space. For example, there may be localized high-concentration pollution areas in wastewater discharge ponds, and fixed sampling points are prone to missing these critical areas. Conventional methods sample cyclically at fixed time intervals or along fixed paths, and cannot dynamically adjust the frequency according to the degree of pollution. For example, over-sampling in low-pollution areas wastes resources, while insufficient sampling in high-pollution areas leads to data distortion. Furthermore, conventional systems usually rely solely on online sensors to collect data, which are susceptible to the influence of sensor accuracy, calibration cycle, and environmental interference, easily resulting in measurement errors. Therefore, a method and system for monitoring the progress of wastewater treatment is proposed. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies, such as conventional methods that typically use fixed-location sampling, which cannot adapt to the spatiotemporal dynamic changes of wastewater composition; conventional methods that use fixed time intervals or fixed paths for cyclic sampling, which cannot dynamically adjust the frequency according to the degree of pollution; and conventional systems that typically rely solely on online sensors to collect data, which are easily affected by sensor accuracy, calibration cycles, and environmental interference. Therefore, this invention proposes a wastewater treatment progress monitoring method and system.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: A method for monitoring wastewater treatment progress, comprising: S1: The camera captures panoramic views, and the YOLO model identifies the contaminated area and marks its coordinates. S2, adjust the sampling point density according to the pollution weight, and use the A* algorithm to plan the shortest sampling path; S3, sensor-image-historical data fusion, LSTM model predicts processing progress; S4, the progress bar quantifies the remaining time, and a two-level threshold triggers a process adjustment warning.

[0006] A monitoring system used in a method for monitoring wastewater treatment progress includes: The panoramic image acquisition module is used to acquire panoramic images of the wastewater discharge pool in real time, providing basic data for feature region analysis; The feature region analysis module is used to mark high-risk pollution areas based on historical sampling data and image processing technology; The dynamic sampling point optimization module is used to automatically adjust the sampling point position and the number of iterations based on the feature region distribution obtained by the feature region analysis module. The multi-source data fusion module is used to integrate sensor data, image data, and historical data to build an LSTM neural network model; The wastewater treatment progress assessment module is used to assess the treatment progress based on the output results of the LSTM neural network model and generate a visual report. The system integration and control module coordinates operations to achieve full-process automation.

[0007] The above technical solution further includes: The implementation steps of the panoramic image acquisition module are as follows: Hardware deployment: An industrial-grade panoramic camera was installed above the discharge pool to cover the entire surface of the pool. Equipped with LED explosion-proof lights, ensuring image clarity in low-light environments; Image acquisition and transmission: The camera captures panoramic images at a preset frequency and transmits them via Ethernet and 5G. The image data is labeled with timestamps and camera IDs, making it traceable. Output: Original panoramic image file.

[0008] The implementation steps of the feature region analysis module are as follows: Historical data correlation: Historical sampling data is retrieved from the database, matched to the corresponding image coordinate system, and calibrated using a geographic information system (GIS) to generate a historical pollution distribution heat map, which serves as a benchmark for feature identification. Image feature extraction: Preprocessing: Noise reduction using Gaussian filtering and contrast enhancement using histogram equalization; Feature recognition: Color characteristics: By segmenting the color space using HSV, abnormal color areas are marked, such as oil stains which appear black; Texture features: LBP algorithm is used to detect areas of suspended matter accumulation; Deep learning model: Train the YOLO model to directly identify pollution types, such as foam and scum.

[0009] Feature region labeling: Overlay high COD areas, heavy metal suspension areas, and confidence scores onto the image. Pollution type label; Output the coordinate set of the feature region, such as .

[0010] The implementation steps of the dynamic sampling point optimization module are as follows: Sampling strategy generation: Weighting: The sampling weight for high-pollution areas with COD>500mg / L was set to 3, and the weight for low-pollution areas was set to 1. Weighting formula: ; Sampling point density calculation: High-weighted areas have 2 points per square meter, while low-weighted areas have 1 point per 5 square meters. Loop count optimization: Based on historical data volatility (e.g., COD standard deviation in a certain region > 15%), dynamically increase the number of cycles; Path planning: The A* algorithm is used to plan the movement path of the sampling device, minimize the total travel distance, avoid obstacles, and prioritize the coverage of high-weight areas. Output: Optimized set of sampling point coordinates, number of loops, and execution order.

[0011] The implementation steps of the multi-source data fusion module are as follows: Data alignment: The sampling point coordinates are matched with the real-time pH and dissolved oxygen data of the sensor by timestamp, and historical data is synchronized to the current time through time series interpolation (such as cubic spline interpolation); Feature engineering: Extract image features (such as the area of ​​the polluted area and the mean color) and sensor features (such as the concentration change rate). Constructing feature vectors: ; Model training: An LSTM neural network is used, with historical data as the training set, to predict the current wastewater treatment progress (such as the remaining treatment time).

[0012] Output: The fused cube and the parameters of the prediction model.

[0013] The implementation steps of the wastewater treatment progress assessment module are as follows: Progress calculation: The remaining processing time is output based on the LSTM neural network model (e.g., "Denitrification is expected to be completed in 2 hours"). The progress bar is dynamically updated based on the current processing efficiency (e.g., COD removal rate of 90%). Anomaly warning: When the predicted value exceeds the threshold (e.g., remaining time > 20% of the set value), an early warning signal is triggered (SMS / email notification). Visual presentation: The monitoring screen displays panoramic images, real-time sampling point locations, and processing progress curves, generating a PDF report containing data tables, trend charts, and recommended measures.

[0014] The implementation steps of the system integration and control module are as follows: Task scheduling: Work is triggered sequentially via Kafka message queues; Hardware control: Send instructions to the PLC to drive the sampling device to move along the optimized path; User interaction: A web interface is provided for operators to manually adjust parameters such as sampling frequency and warning threshold; Log management: Record operation logs and support fault traceability.

[0015] The present invention has the following beneficial effects: 1. In this invention, panoramic images are acquired through a panoramic image acquisition module. A camera captures panoramic images of the wastewater discharge pool. Combined with historical sampling data, feature regions are marked. Then, based on the feature region analysis results, the dynamic sampling point optimization module automatically adjusts the sampling point positions and the number of cycles to ensure that the monitoring data is representative.

[0016] 2. In this invention, multi-source data fusion is performed through a multi-source data fusion module, integrating sensor data, image data, and historical operational data, and then constructing a multidimensional dataset to improve monitoring accuracy. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the method steps of a wastewater treatment progress monitoring method and system proposed in this invention; Figure 2 This is a schematic diagram of the monitoring system in this invention.

[0018] The diagram shows: 1. Panoramic image acquisition module; 2. Feature region analysis module; 3. Dynamic sampling point optimization module; 4. Multi-source data fusion module; 5. Wastewater treatment progress assessment module; 6. System integration and control module. Detailed Implementation

[0019] 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.

[0020] like Figure 1 - Figure 2 As shown, the present invention proposes a method for monitoring the progress of wastewater treatment, comprising: S1: The camera captures panoramic views, and the YOLO model identifies the contaminated area and marks its coordinates. S2, adjust the sampling point density according to the pollution weight, and use the A* algorithm to plan the shortest sampling path; S3, sensor-image-historical data fusion, LSTM model predicts processing progress; S4, the progress bar quantifies the remaining time, and a two-level threshold triggers a process adjustment warning.

[0021] A monitoring system used in a method for monitoring wastewater treatment progress includes: Panoramic image acquisition module 1 is used to acquire panoramic images of the wastewater discharge pool in real time, providing basic data for feature region analysis. Feature region analysis module 2 is used to mark high-pollution-risk areas based on historical sampling data and image processing technology; The dynamic sampling point optimization module 3 is used to automatically adjust the sampling point position and the number of cycles based on the feature region distribution obtained by the feature region analysis module 2. Multi-source data fusion module 4 is used to integrate sensor data, image data and historical data to build an LSTM neural network model; Wastewater treatment progress assessment module 5 is used to assess the treatment progress based on the output results of the LSTM neural network model and generate a visual report. System integration and control module 6 coordinates operation to achieve full-process automation.

[0022] The implementation steps of the panoramic image acquisition module 1 are as follows: Hardware deployment: An industrial-grade panoramic camera was installed above the discharge pool to cover the entire surface of the pool. Equipped with LED explosion-proof lights, ensuring image clarity in low-light environments; Image acquisition and transmission: The camera captures panoramic images at a preset frequency and transmits them via Ethernet and 5G. The image data is labeled with timestamps and camera IDs, making it traceable. Output: Original panoramic image file.

[0023] The implementation steps of the feature region analysis module 2 are as follows: Historical data correlation: Historical sampling data (such as COD and ammonia nitrogen concentrations) are retrieved from the database, matched to the corresponding image coordinate system, and calibrated through a geographic information system (GIS) to generate a historical pollution distribution heat map, which serves as a benchmark for feature identification. Image feature extraction: Preprocessing: Noise reduction using Gaussian filtering and contrast enhancement using histogram equalization; Feature recognition: Color characteristics: By segmenting the color space using HSV, abnormal color areas are marked, such as oil stains which appear black; Texture features: LBP algorithm is used to detect areas of suspended matter accumulation; Deep learning model: Train the YOLO model to directly identify pollution types, such as foam and scum.

[0024] Feature region labeling: Overlay high COD areas, heavy metal suspension areas, and confidence scores onto the image. Pollution type label; Output the coordinate set of the feature region, such as .

[0025] The implementation steps of the dynamic sampling point optimization module 3 are as follows: Sampling strategy generation: Weighting: The sampling weight for high-pollution areas with COD>500mg / L was set to 3, and the weight for low-pollution areas was set to 1. Weighting formula: ; Sampling point density calculation: High-weighted areas have 2 points per square meter, while low-weighted areas have 1 point per 5 square meters. Loop count optimization: Based on historical data volatility (e.g., COD standard deviation in a certain region > 15%), dynamically increase the number of cycles; Path planning: The A* algorithm is used to plan the movement path of the sampling device, minimize the total travel distance, avoid obstacles, and prioritize the coverage of high-weight areas. Output: Optimized set of sampling point coordinates, number of loops, and execution order.

[0026] The implementation steps of the multi-source data fusion module 4 are as follows: Data alignment: The sampling point coordinates are matched with the real-time pH and dissolved oxygen data of the sensor by timestamp, and historical data is synchronized to the current time through time series interpolation (such as cubic spline interpolation); Feature engineering: Extract image features (such as the area of ​​the polluted area and the mean color) and sensor features (such as the concentration change rate). Constructing feature vectors: ; Model training: An LSTM neural network is used, with historical data as the training set, to predict the current wastewater treatment progress (such as the remaining treatment time).

[0027] Output: The fused cube and the parameters of the prediction model.

[0028] The implementation steps of the wastewater treatment progress assessment module 5 are as follows: Progress calculation: The remaining processing time is output based on the LSTM neural network model (e.g., "Denitrification is expected to be completed in 2 hours"). The progress bar is dynamically updated based on the current processing efficiency (e.g., COD removal rate of 90%). Anomaly warning: When the predicted value exceeds the threshold (e.g., remaining time > 20% of the set value), an early warning signal is triggered (SMS / email notification). Visual presentation: The monitoring screen displays panoramic images, real-time sampling point locations, and processing progress curves, generating a PDF report containing data tables, trend charts, and recommended measures.

[0029] The implementation steps of the system integration and control module 6 are as follows: Task scheduling: Work is triggered sequentially via Kafka message queues; Hardware control: Send instructions to the PLC to drive the sampling device to move along the optimized path; User interaction: A web interface is provided for operators to manually adjust parameters such as sampling frequency and warning threshold; Log management: Record operation logs and support fault traceability.

[0030] In this embodiment, during wastewater treatment progress monitoring, the panoramic image acquisition module 1 serves as the data entry point. A 4K camera captures a panoramic view of the pool surface every 5 minutes, with automatic LED explosion-proof lighting. Images are input with added timestamps and camera IDs and then stored in the NAS. The feature region analysis module 2 retrieves concurrent water quality data from the database, maps it to image pixels via GIS, and performs preprocessing including Gaussian filtering and histogram equalization. Then, it uses HSV color segmentation, LBP texture detection, and a YOLOv8 deep learning model to identify feature regions such as oil, foam, and scum, outputting a coordinate set with confidence. The dynamic sampling point optimization module 3 allocates sampling density based on COD concentration weights (high-pollution areas have a weight of 3 and a density of 2 points / m², while low-pollution areas have a weight of 1 and a density of 0.2 points / m²). It uses the A* algorithm to plan the optimal path to avoid obstacles. If the area's COD... If the standard deviation exceeds 15%, the number of cyclic samplings is increased; the multi-source data fusion module 4 aligns the sampling point coordinates with sensor data (pH, dissolved oxygen, etc.) according to timestamps, fills in missing values ​​with cubic spline interpolation, constructs a multi-dimensional vector containing water quality parameters and image features, and outputs a fused dataset after predicting the remaining treatment time using LSTM; the wastewater treatment progress assessment module 5 dynamically updates the progress bar according to the COD removal rate, and triggers an alarm if the predicted remaining time exceeds the set value by 20%, and synchronizes the panoramic image, sampling point location and progress curve through the monitoring screen to generate a PDF report containing trend charts and suggested measures; the system integration and control module 6 acts as the central hub, executes the sequence through the Kafka message queue scheduling module, sends robotic arm movement instructions to the PLC, provides a web interface for users to manually adjust parameters, and records a full-process log to support fault backtracking.

[0031] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method of monitoring progress of sewage treatment, characterized by, include: S1: The camera captures panoramic views, and the YOLO model identifies the contaminated area and marks its coordinates. S2, adjust the sampling point density according to the pollution weight, and use the A* algorithm to plan the shortest sampling path; S3, sensor-image-historical data fusion, LSTM model predicts processing progress; S4, the progress bar quantifies the remaining time, and a two-level threshold triggers a process adjustment warning.

2. A monitoring system for use in a method of monitoring progress of sewage treatment according to claim 1, characterised in that, include: The panoramic image acquisition module (1) is used to acquire panoramic images of the wastewater discharge pool in real time, providing basic data for feature region analysis. Feature region analysis module (2) is used to mark high-pollution-risk areas based on historical sampling data and image processing technology; The dynamic sampling point optimization module (3) is used to automatically adjust the sampling point position and the number of cycles based on the feature region distribution obtained by the feature region analysis module (2); The multi-source data fusion module (4) is used to integrate sensor data, image data and historical data to build an LSTM neural network model; The wastewater treatment progress assessment module (5) is used to assess the treatment progress based on the output results of the LSTM neural network model and generate a visual report. The system integration and control module (6) coordinates operation to achieve full-process automation.

3. A monitoring system for use in a method of monitoring the progress of sewage treatment according to claim 2, characterised in that, The implementation steps of the panoramic image acquisition module (1) are as follows: Hardware deployment: An industrial-grade panoramic camera is installed above the discharge pool to cover the entire pool surface. It is equipped with LED explosion-proof lights and has high image clarity in low-light environments. Image acquisition and transmission: The camera captures panoramic images at a preset frequency and transmits them via Ethernet and 5G. The image data is labeled with timestamps and camera IDs. Output: Original panoramic image file.

4. A monitoring system for use in a method of monitoring progress of sewage treatment according to claim 2, characterised in that, The implementation steps of the feature region analysis module (2) are as follows: Historical data correlation: Historical sampling data is retrieved from the database, matched to the corresponding image coordinate system, and calibrated using a geographic information system (GIS) to generate a historical pollution distribution heat map, which serves as a benchmark for feature identification. Image feature extraction: Preprocessing: Noise reduction using Gaussian filtering and contrast enhancement using histogram equalization; Feature recognition: Color characteristics: By segmenting the color space using HSV, abnormal color areas are marked, such as oil stains which appear black; Texture Features: The LBP algorithm is used to detect areas of suspended matter accumulation; Deep learning model: Train the YOLO model to directly identify pollution types, such as foam and scum; Feature region labeling: Overlaying on the image the high COD areas, heavy metal suspended areas and the confidence score of the pollution type label ; output feature region coordinate set, .

5. A monitoring system for use in a method of monitoring progress of sewage treatment according to claim 2, characterised in that, The implementation steps of the dynamic sampling point optimization module (3) are as follows: Sampling strategy generation: Weighting: The sampling weight for high-pollution areas with COD>500mg / L was set to 3, and the weight for low-pollution areas was set to 1. Weight formula: ; Sampling point density calculation: High-weighted areas have 2 points per square meter, while low-weighted areas have 1 point per 5 square meters. Loop count optimization: The number of cycles is dynamically increased based on the volatility of historical data. Path planning: The A* algorithm is used to plan the movement path of the sampling device, minimize the total travel distance, avoid obstacles, and prioritize the coverage of high-weight areas. Output: Optimized set of sampling point coordinates, number of loops, and execution order.

6. The monitoring system used in the wastewater treatment progress monitoring method according to claim 2, characterized in that, The implementation steps of the multi-source data fusion module (4) are as follows: Data alignment: The sampling point coordinates are matched with the real-time pH and dissolved oxygen data of the sensor by timestamp, and historical data is synchronized to the current time through time series interpolation; feature project: Extracting image features and sensor features; Constructing the feature vector: ; Model training: An LSTM neural network is used, with historical data as the training set, to predict the current sewage treatment progress. Output: The fused cube and the parameters of the prediction model.

7. A monitoring system for use in a method of monitoring progress of sewage treatment according to claim 2, characterised in that, The implementation steps of the wastewater treatment progress assessment module (5) are as follows: Progress calculation: Output the remaining processing time based on the LSTM neural network model; The progress bar is dynamically updated based on the current processing efficiency. Anomaly warning: The predicted value exceeds the threshold, triggering an early warning signal; Visual presentation: The monitoring screen displays panoramic images, real-time sampling point locations, and processing progress curves, generating a PDF report containing data tables, trend charts, and recommended measures.

8. A monitoring system for use in a method of monitoring progress of sewage treatment according to claim 2, characterised in that, The implementation steps of the system integration and control module (6) are as follows: Task scheduling: Work is triggered sequentially via Kafka message queues; Hardware control: Send instructions to the PLC to drive the sampling device to move along the optimized path; User interaction: A web interface is provided for operators to manually adjust parameters such as sampling frequency and warning threshold; Log management: Record operation logs and support fault traceability.

Images