Visual identification system for urban water pollution treatment

Abstract

The invention provides a visual identification system for urban water pollution treatment, aims to realize real-time monitoring and pollutant identification of urban water, and integrates an image acquisition module, a data processing module, an environment sensor module, a central processing system and a monitoring and alarm module. The image acquisition module acquires a water body image in real time through the unmanned aerial vehicle, the unmanned ship and a fixed monitoring point and identifies pollutants such as floating garbage, oil stain and algae in combination with a convolutional neural network model, the environment sensor acquires water quality parameters and generates a water quality state weight, and the central processing system predicts a pollution trend by using an autoregressive moving average model. And a water quality risk grade is generated by adopting a fuzzy logic algorithm, so that accurate early warning is realized. The visual identification system has the online learning capability, can adapt to environmental changes and continuously optimize the identification precision, has the advantages of wide coverage, high real-time performance and accurate identification compared with a traditional water body monitoring method, and provides an intelligent and automatic solution for urban water body pollution management.

Description

Technical Field

[0001] The present invention relates to the technical field of ecological restoration, and in particular to a visual recognition system for urban water pollution control. Background Art

[0002] With the acceleration of urbanization, urban ecological and environmental problems are becoming increasingly prominent, especially urban water pollution, which has become one of the key factors affecting the quality of life of urban residents and ecological balance. Urban water bodies (including rivers, lakes, reservoirs, etc.) are usually affected by multiple pollution sources such as industrial emissions, domestic sewage and agricultural runoff, leading to water quality deterioration, destruction of aquatic habitats, ecological imbalance and other problems. In order to effectively control these problems, it is urgent to establish an efficient and accurate water quality monitoring and management system to detect pollution in a timely manner, trace the source of pollution and take corresponding control measures.

[0003] Currently, urban water monitoring methods are mainly divided into two categories: traditional manual sampling and laboratory analysis, and automated sensor detection at fixed monitoring points.

[0004] For manual sampling and laboratory analysis, this method usually involves professionals collecting water samples on site and sending them to the laboratory for chemical, physical, and biological analysis. Although this method is relatively accurate, it has several obvious disadvantages: (1) Poor real-time performance. It takes a certain amount of time from water sample collection to analysis, making it difficult to achieve real-time monitoring. (2) High cost. The cost of manual sampling and laboratory analysis is high, making it difficult to cover the entire urban water area on a large scale. (3) Discontinuity. This method is difficult to conduct large-scale continuous monitoring and is prone to missing water quality emergencies.

[0005] For automated sensor detection at fixed monitoring points, this method usually installs sensors at fixed locations in rivers and lakes, and uses physical and chemical sensors to detect water quality parameters such as pH, dissolved oxygen, conductivity, temperature, etc. in real time. Although this method has certain advantages in terms of real-time and continuity, it also has obvious shortcomings: (1) Limited coverage. Fixed monitoring points can only detect water quality conditions at the installation location and it is difficult to cover the pollution distribution of the entire water body. (2) Problems with sensor accuracy and life. Sensors that have been used for a long time are easily affected by the water quality environment, and may malfunction or err, resulting in unstable data. (3) Lack of pollutant identification capabilities. Traditional sensors can only detect a single physical or chemical parameter and it is difficult to identify the specific types of pollutants and pollution sources in the water body.

[0006] In the existing water monitoring system, there are problems such as lack of intelligent and automated data analysis methods, lack of real-time and continuity, limited pollutant identification, insufficient data reliability and accuracy, etc. Therefore, we urgently need to design a visual recognition system for urban water pollution control to solve the above problems. Summary of the invention

[0007] The purpose of the present invention is to provide a visual recognition system for urban water pollution control in view of the shortcomings of the prior art, so as to solve the problems raised in the background technology.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] A visual recognition system for urban water pollution control, comprising an image acquisition module, a data processing module, an environmental sensor module, a central processing system, and a monitoring and alarm module, wherein the image acquisition module is used to acquire environmental image data and is applicable to, but not limited to, real-time monitoring of urban water bodies;

[0010] The data processing module analyzes the collected images and identifies pollutants and floating objects in the environment through deep learning algorithms;

[0011] The environmental sensor module is used to collect physical and chemical parameters, including but not limited to pH, dissolved oxygen, conductivity, and temperature;

[0012] The central processing system fuses the image data and sensor data and obtains the environmental status through multi-source data analysis;

[0013] The monitoring and alarm module automatically issues an alarm signal and generates corresponding treatment suggestions when pollution or abnormal conditions are detected based on the analysis results.

[0014] As a preferred technical solution of the present invention, the image acquisition module includes an unmanned aerial vehicle, an unmanned ship and a fixed monitoring point, and the collected image data is transmitted to the central processing system in real time via a wireless network. The unmanned aerial vehicle and the unmanned ship are used for environmental image acquisition in dynamic areas, and the fixed monitoring points are used for long-term image acquisition in fixed areas.

[0015] As a preferred technical solution of the present invention, the data processing module uses a convolutional neural network (CNN) for image recognition and constructs a pollutant detection model through a deep learning algorithm, and the model is expressed as follows:

[0016]

[0017] Among them, x is the input image feature vector, w i is the weight parameter, b is the bias term, and f(x) is the pollutant identification function. After being trained with a large number of environmental pollution image samples, the model can identify pollution features as a preliminary basis for judging the environmental status.

[0018] As a preferred technical solution of the present invention, the pollutant detection model is combined with an edge detection algorithm of the image to accurately extract the contour features of the pollutant. The edge detection function is:

[0019]

[0020] Among them, G(x,y) is the image edge gradient, and I is the pixel value of the environmental image. This edge detection algorithm is combined with a convolutional neural network to extract the precise contour of the pollutant and improve the accuracy of recognition.

[0021] As a preferred technical solution of the present invention, the data collected by the environmental sensor module is filtered by a Kalman filter algorithm to remove noise and improve data stability. The filtering algorithm is:

[0022]

[0023] in, is the state estimate after filtering, K k is the Kalman gain, z k is the measured value, and H is the observation matrix. The filtered sensor data is provided to the central processing system together with the visual data to obtain more stable environmental parameters.

[0024] As a preferred technical solution of the present invention, the multi-source data fusion module of the central processing system fuses the visual recognition data and the sensor data, and uses a weighted average method. The fusion weight is dynamically allocated according to the confidence of the data. The calculation formula is:

[0025] W fusion =αW vision +βW sensor

[0026] Among them, W fusion is the weight of the fused environment state, W vision is the visual data weight, W sensor is the sensor data weight, α and β are confidence weights, and α+β = 1. The fused data is used as the basis for further judgment of the environmental state.

[0027] As a preferred technical solution of the present invention, the monitoring and alarm module performs threshold judgment on the fused environmental state value. fusion When the preset threshold value T is exceeded, an alarm is triggered. The judgment conditions are:

[0028] W fusion >T

[0029] Wherein, T is the preset environmental safety threshold. When the threshold is exceeded, the system will issue an alarm signal and generate processing suggestions.

[0030] As a preferred technical solution of the present invention, the central processing system has a pollution trend prediction function, uses the fused environmental state data to perform time series analysis, and uses an autoregressive moving average model (ARMA) to predict the changing trend of the environmental state. The model is:

[0031]

[0032] Among them, W t represents the fusion value of the environment state at time t, φ i and θ j are the autoregressive and moving average coefficients, ∈ t is a white noise term. This prediction function provides pollution trend warning for environmental management.

[0033] As a preferred technical solution of the present invention, the central processing system performs a comprehensive evaluation on the environmental state value based on the fuzzy logic algorithm to calculate the environmental risk level. The fuzzy evaluation function is:

[0034] Risk Level = f(pH, DO, T)

[0035] Among them, pH represents acidity, DO represents dissolved oxygen concentration, T represents temperature, and RiskLevel represents the risk level of the environment. The system divides the environmental status into four levels: "excellent", "good", "slightly polluted", and "severely polluted" through fuzzy rules.

[0036] As a preferred technical solution of the present invention, the system performs data storage and model update through a cloud server, uploads all image data and sensor data to the cloud, and automatically updates the model through an online learning algorithm to improve pollutant identification and detection accuracy. The online learning algorithm is defined as follows:

[0037]

[0038] Among them, θ t is the model parameter at the tth moment, η is the learning rate, is the gradient of the loss function, which is based on the environment fusion value W fusion , used to dynamically adjust the model weights. The online learning algorithm ensures that the model continuously adapts to environmental changes and improves recognition accuracy.

[0039] Compared with the prior art, the visual recognition system for urban water pollution control of the present invention overcomes the problems of poor real-time performance, limited coverage, insufficient recognition accuracy, etc. of the traditional water monitoring system by integrating visual recognition technology, environmental sensors, data fusion and intelligent algorithms, and has the following beneficial effects:

[0040] 1. The visual recognition system for urban water pollution control of the present invention adopts an image acquisition method that combines drones, unmanned boats and fixed monitoring points, which can cover a large range of water bodies and realize all-weather, multi-angle pollution monitoring. Drones and unmanned boats can dynamically collect water body images over a large range, and fixed monitoring points provide long-term fixed-point monitoring, ensuring all-round coverage of water pollution. Compared with the traditional method that relies on manual sampling and fixed monitoring points, this system can realize real-time and continuous monitoring of water bodies, significantly improve the response speed of water quality monitoring, and help to timely discover and respond to pollution incidents.

[0041] 2. The visual recognition system for urban water pollution control of the present invention analyzes images through a convolutional neural network (CNN) model to achieve automatic recognition of pollutants such as floating garbage, oil pollution, and algae, and accurately identify the type, location, and density of pollutants. The system uses a multi-source data fusion algorithm to fuse sensor data (such as pH, dissolved oxygen, conductivity, etc.) with visual data, and uses Kalman filtering technology to remove noise from sensor data, greatly improving the accuracy and reliability of water quality monitoring data. The application of multi-source data fusion makes up for the limitations of a single data source, allowing the system to provide more comprehensive water quality analysis results.

[0042] 3. The visual recognition system for urban water pollution control of the present invention has the functions of pollution trend prediction and risk warning. It analyzes and predicts historical water quality data through the autoregressive moving average model (ARMA), and can identify potential pollution risks in advance. The system also uses fuzzy logic algorithms to comprehensively evaluate the water quality status and generate water quality risk levels (such as "excellent", "good", "mildly polluted", "severely polluted", etc.), helping managers to quickly understand the health status of water bodies and make decisions. The online learning function of the system can continuously optimize the recognition model, automatically adapt to environmental changes, maintain high recognition accuracy, and ensure the reliability and flexibility of long-term monitoring. This intelligent prediction and warning function makes this system have extremely high practical value in urban water management. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 This is a system block diagram of a visual recognition system for urban water pollution control proposed by the present invention. DETAILED DESCRIPTION

[0044] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0045] The following is combined with Figure 1 With reference to the accompanying drawings and a number of embodiments, specific embodiments of the present invention are described in detail.

[0046] Example 1: Water pollution monitoring based on drones

[0047] System structure and configuration: The system includes image acquisition module (UAV), data processing module (convolutional neural network recognition model), environmental sensor module, central processing system (data fusion module) and monitoring and alarm module.

[0048] The drone is equipped with a high-definition camera and a wireless communication module, which is used to regularly patrol urban rivers, lakes and other water bodies, collect water image data in real time, and upload the image data to the central processing system via a wireless network.

[0049] Data collection and sensor data processing: The drone takes images of the water surface while cruising, and uses the convolutional neural network (CNN) model to process the image data. The mathematical expression of image feature extraction and pollutant identification is:

[0050]

[0051] Among them, x is the input image feature vector, w i is the weight parameter obtained from model training, and b is the bias term. Through training with a large number of pollution image samples, the model can identify pollution features such as floating objects and oil pollution, providing a basis for the preliminary judgment of water quality status.

[0052] Fixed monitoring points are arranged along the river or lake along the flight path of the drone. Environmental sensors at the monitoring points collect water quality parameters such as pH, dissolved oxygen (DO), and conductivity. The data collected by the sensors are smoothed by Kalman filtering. The filtering algorithm is as follows:

[0053]

[0054] in, represents the state estimate after filtering, K k is the Kalman gain, z k is the sensor measurement value, and H is the observation matrix. This processing method removes sensor noise and improves the reliability of water quality parameters.

[0055] Multi-source data fusion and monitoring alarm: The central processing system fuses the image recognition data and the filtered sensor data, and calculates the weight of the fused environmental state through weighted average:

[0056] W fusion =αW vision +βW sensor

[0057] Among them, Wvision is the weight of visual recognition data, W sensor is the weight of sensor data, and weights α and β satisfy α + β = 1. The weights are adjusted dynamically to adapt to the reliability of visual and sensor data in different scenarios.

[0058] The system's fused water quality state value W fusion Perform threshold judgment. When W fusion When the preset threshold T is exceeded, an alarm is triggered, and the conditions are:

[0059] W fusion >T

[0060] After receiving the alarm information, the management personnel can take prompt response measures.

[0061] Pollution trend prediction and risk assessment: The central processing system uses the autoregressive moving average model (ARMA) to predict the trend of water pollution. The model is as follows:

[0062]

[0063] Among them, W t is the fusion value of water quality status at the current moment, φ i and θ j are the autoregressive and moving average coefficients, ∈ t is a white noise term. The prediction model provides an early warning of pollution risk.

[0064] Example 2: Detection of floating garbage based on unmanned boats

[0065] System structure and configuration: This embodiment uses an unmanned boat as an image acquisition module, equipped with a high-definition camera and a water quality sensor, to monitor floating garbage in lakes and rivers.

[0066] Image processing and pollutant identification: The unmanned ship collects water surface images while cruising and extracts the contour features of floating garbage through edge detection algorithms. The edge detection formula is as follows:

[0067]

[0068] Among them, G(x,y) is the image edge gradient, and I is the image pixel value. By combining edge detection and convolutional neural network, the system can identify floating garbage on the water surface.

[0069] Environmental sensor data collection and fusion: The dissolved oxygen and temperature sensors equipped on the unmanned boat collect water quality parameters in real time, remove data noise through the Kalman filter algorithm, and use it together with visual data to judge water quality.

[0070] Fuzzy logic risk assessment: The central processing system conducts a comprehensive assessment of environmental data based on fuzzy logic algorithms. The fuzzy assessment formula is:

[0071] Risk Level = f(pH, DO, T)

[0072] Among them, pH is acidity, DO is dissolved oxygen concentration, T is temperature, and RiskLevel represents the risk level. The algorithm divides the environmental status into "excellent", "good", "mildly polluted", "severely polluted" and other levels.

[0073] Example 3: Long-term monitoring of water quality at fixed monitoring points

[0074] System structure and configuration: Fixed monitoring points are set up in key areas of urban rivers. Each monitoring point is equipped with a high-definition camera and water quality sensor, and the data is transmitted to the central processing system through the 4G / 5G network.

[0075] Data collection and fusion processing: The camera regularly collects water body images and uses the CNN model to detect floating objects in the images. After the image recognition data is fused with the sensor data, the water quality state weight W is generated. fusion .

[0076] Online learning and adaptive updating: The system uses online learning algorithms to dynamically update the pollution identification model to improve identification accuracy. The online learning algorithm is as follows:

[0077]

[0078] Among them, θ t represents the model parameters, η is the learning rate, is the gradient of the loss function. The model is constantly adjusted to adapt to environmental changes in long-term monitoring.

[0079] Historical data analysis and early warning: The system generates water quality change reports based on historical data and predicts future water quality trends, providing managers with a basis for early deployment.

[0080] Example 4: Monitoring of sudden river pollution incidents

[0081] System structure and configuration: Mobile monitoring stations are deployed in high-risk areas of river pollution, equipped with high-definition cameras and sensors to monitor sudden pollution incidents in real time.

[0082] Pollution identification and data analysis: Image recognition algorithms are combined with sensor data to quickly identify pollutants in water bodies. When the sensor detects abnormal pH values ​​or heavy metal ions, the pollution level is determined through a multi-source data fusion algorithm.

[0083] Emergency Alarm and Response: The system triggers a high-level alarm based on the pollution level and notifies the emergency response team to handle the situation on site. This emergency response mode can quickly upload pollution data to the central processing system and generate a pollutant diffusion model to provide decision support for managers.

[0084] Dynamic learning and model updating: The system conducts online learning based on real-time data to improve emergency response capabilities and adapt to the handling requirements of different pollution incidents.

[0085] Example 5: Lake algae outbreak monitoring and early warning

[0086] System structure and configuration: UAVs and fixed monitoring points are deployed in the lake. The UAVs are equipped with high-definition cameras to regularly patrol the lake surface, and the fixed monitoring points are equipped with dissolved oxygen and temperature sensors.

[0087] Algae identification and density calculation: UAV image data is used to identify the distribution of algae on the lake surface through a deep learning model, using the formula Perform feature extraction and density prediction.

[0088] Trend prediction and risk warning: The system uses the ARMA model to predict the trend of algae density. The formula is as follows:

[0089]

[0090] Among them, W t is the fusion value of water quality status at the current moment, φ i and θ j are the autoregressive and moving average coefficients, ∈ t is a white noise term. The prediction model provides an early warning of pollution risk.

[0091] When the risk of algae outbreak is predicted, the system sends early warning information.

[0092] Dynamic adjustment and online learning: The system updates model parameters based on real-time data to adapt to changes in lake water quality and improve the accuracy of algae bloom warnings.

[0093] The contents not described in detail in this specification belong to the prior art known to professional and technical personnel in the field. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A visual recognition system for urban water pollution control, characterized by: Its components include an image acquisition module, a data processing module, an environmental sensor module, a central processing system and a monitoring and alarm module. The image acquisition module is used to collect environmental image data and conduct real-time monitoring of urban water bodies. The data processing module analyzes the collected images and identifies pollutants and floating objects in the environment through a deep learning algorithm; The environmental sensor module is used to collect physical and chemical parameters, including but not limited to pH value, dissolved oxygen, conductivity and temperature; The central processing system fuses the image data and the sensor data, and obtains the environmental status through multi-source data analysis. The environmental status is divided into four levels: "excellent", "good", "mildly polluted", and "severely polluted"; The monitoring and alarm module automatically sends out an alarm signal when pollution or abnormal conditions are detected based on the analysis results, and generates treatment suggestions corresponding to the environmental status.

2. A visual recognition system for urban water pollution control according to claim 1, characterized in that: The image acquisition module includes unmanned aerial vehicles, unmanned ships and fixed monitoring points, and transmits the collected image data to the central processing system in real time via a wireless network. The unmanned aerial vehicles and unmanned ships are used for environmental image acquisition in dynamic areas, and the fixed monitoring points are used for long-term image acquisition in fixed areas.

3. A visual recognition system for urban water pollution control according to claim 1, characterized in that: The data processing module uses a convolutional neural network for image recognition and builds a pollutant detection model through a deep learning algorithm. The model is expressed as follows: Among them, x is the input image feature vector, w i is the weight parameter, b is the bias term, and f(x) is the pollutant identification function.

4. A visual recognition system for urban water pollution control according to claim 3, characterized in that: The pollutant detection model combines the edge detection algorithm of the image to extract the contour features of the pollutant. The edge detection function is: Among them, G(x,y) is the image edge gradient, and I is the environment image pixel value.

5. The visual recognition system for urban water pollution control according to claim 1 is characterized in that: The data collected by the environmental sensor module is filtered by the Kalman filter algorithm to remove noise. The filtering algorithm is: in, is the state estimate after filtering, K k is the Kalman gain, z k is the measurement value, and H is the observation matrix.

6. The visual recognition system for urban water pollution control according to claim 1 is characterized in that: The multi-source data fusion module of the central processing system fuses the visual recognition data and the sensor data, and uses the weighted average method. The fusion weight is dynamically allocated according to the confidence of the data. The calculation formula is: W fusion =αW vision +βW sensor Among them, W fusion is the weight of the fused environment state, W vision is the visual data weight, W sensor is the sensor data weight, α and β are confidence weights, satisfying α+β=1.

7. The visual recognition system for urban water pollution control according to claim 1 is characterized in that: The monitoring and alarm module performs threshold judgment on the fused environmental state value. fusion When the preset threshold value T is exceeded, an alarm is triggered. The judgment conditions are: W fusion >T Wherein, T is the preset environmental safety threshold. When the threshold is exceeded, the system will issue an alarm signal and generate processing suggestions.

8. The visual recognition system for urban water pollution control according to claim 1 is characterized in that: The central processing system has the function of predicting pollution trends. It uses the fused environmental status data to perform time series analysis and adopts an autoregressive moving average model to predict the changing trend of environmental status. The model is: Among them, W t represents the fusion value of the environment state at time t, φ i and θ j are the autoregressive and moving average coefficients, ∈ t is the white noise term.

9. A visual recognition system for urban water pollution control according to claim 8, characterized in that: The central processing system performs a comprehensive evaluation of the environmental status value based on the fuzzy logic algorithm and calculates the environmental risk level. The fuzzy evaluation function is: Risk Level = f(pH, DO, T) Among them, pH represents acidity and alkalinity, DO represents dissolved oxygen concentration, T represents temperature, and RiskLevel represents the risk level of the environment.

10. The visual recognition system for urban water pollution control according to claim 1 is characterized in that: The system uses a cloud server to store data and update models. All image data and sensor data are uploaded to the cloud, and the model is automatically updated through an online learning algorithm to improve pollutant identification and detection accuracy. The online learning algorithm is defined as follows: Among them, θ t is the model parameter at the tth moment, η is the learning rate, is the gradient of the loss function, which is based on the environment fusion value W fusion , used to dynamically adjust model weights.

Images