A sewage discharge monitoring system for river outfalls based on profile analysis

Abstract

This invention discloses a river discharge outlet sewage monitoring system based on image analysis. The system includes a video monitoring module for collecting infrared and visible light video stream data of the sewage outlet detection area above and underwater. The collected real-time video stream data is transmitted to a pollution image analysis module and, under the control of a remote control module, to a pollution event storage module. A rainfall analysis module acquires real-time rainfall data and, in sunny weather, uses a flow velocity sensor in the sewage outlet detection area to determine if there is liquid flow. This system facilitates rapid determination of sewage discharge in sunny weather. In rainy weather, the video monitoring module is activated, and the pollution status is analyzed through the image model training module and the pollution image analysis module, which helps in sewage detection during rainy weather and facilitates better training of the Faster R-CNN model.

Description

Technical Field

[0001] This invention relates to the field of water quality monitoring technology, specifically to a sewage discharge monitoring system for river outfalls based on profiling analysis. Background Technology

[0002] Water quality monitoring is used to monitor and determine the status of pollutants in water bodies; the data obtained from monitoring should be scientifically calculated and processed, and alarms and records should be issued and recorded in a timely manner. All activities and measures must be in place to ensure the accuracy and reliability of water quality monitoring data.

[0003] Existing pollution video recognition systems for monitoring water quality at sewage outlets cannot analyze whether the water discharged into the river is rainwater or sewage on rainy days. Existing video sewage discharge recognition methods process individual frames of the image, which can only simply determine whether sewage is being discharged, but cannot record the duration of the discharge. It is also difficult to reduce the amount of computation during video analysis, and it is difficult to analyze the monitoring video of the sewage outlet detection area on rainy days. At the same time, it is not convenient to monitor the sewage outlet detection area at night or when there is insufficient light. Summary of the Invention

[0004] The purpose of this invention is to provide a sewage discharge monitoring system for river outfalls based on profile analysis, and to solve the following technical problems:

[0005] Existing pollution video recognition technologies for sewage outlet water quality monitoring are difficult to reduce the computational load during video analysis, and are also difficult to analyze monitoring videos of sewage outlet detection areas on rainy days. Furthermore, they are not convenient for monitoring sewage outlet detection areas at night or when there is insufficient light.

[0006] The objective of this invention can be achieved through the following technical solutions:

[0007] A river discharge monitoring system based on profile analysis includes:

[0008] The video monitoring module is used to collect video stream data of infrared and visible light monitoring of the sewage outlet detection area on the water surface, and to collect video stream data of infrared and visible light monitoring of the sewage outlet detection area underwater; the collected real-time video stream data is transmitted to the pollution image analysis module, and under the control of the remote control module, the real-time video stream data is transmitted to the pollution event storage module;

[0009] The rainfall analysis module is used to acquire real-time rainfall data. In sunny weather, it uses a flow velocity sensor in the sewage outlet detection area to determine whether there is liquid flow. If no liquid flow is detected, the video monitoring module is turned off. If liquid flow is detected, the video monitoring module is turned on, and it is determined that the sewage is being discharged on a non-rainy day. In rainy weather, the video monitoring module is turned on to monitor whether the sewage is in a polluted state.

[0010] The pollution event storage module is used to store all historical video stream data of the pollution period in the sewage outlet detection area, and periodically transmit all historical video stream data of the pollution period to the image model training module.

[0011] The image model training module acquires all historical video stream data during the pollution period, fuses the video stream data from infrared monitoring and visible light monitoring, and trains Faster R-CNN models for different ambient brightness levels using the fused data based on the ambient brightness when the video streams were captured.

[0012] The pollution image analysis module acquires real-time video stream data from the video surveillance module and selects a Faster R-CNN model tailored to the corresponding ambient brightness to analyze the real-time video stream data.

[0013] As a further aspect of the present invention, it also includes:

[0014] The remote control module, based on the analysis results of the rainfall analysis module and the pollution image analysis module, when there is sewage discharge on a non-rainy day and the real-time video stream data analysis indicates a pollution state, controls the sampling module to sample the water quality in the sewage outlet detection area and controls the alarm module to issue an abnormal sewage discharge alarm for the pollution event. The remote control module also controls the video monitoring module to transmit the real-time video stream data to the pollution event storage module for pollution event storage. When the real-time video stream data is not in a pollution state, the remote control module controls the video monitoring module to stop transmitting the real-time video stream data to the pollution event storage module.

[0015] The sampling module is used to sample the water quality in the sewage outlet detection area when the area is polluted.

[0016] The alarm module is used to issue abnormal discharge alarms in response to pollution events;

[0017] The abnormal discharge alarm includes sending abnormal discharge alarm information and satellite positioning information to a designated object via a wireless communication network.

[0018] As a further aspect of the present invention: the image model training module fuses video stream data from infrared monitoring and visible light monitoring, and trains Faster R-CNN models for different ambient brightness levels using the fused data, based on the ambient brightness during video stream capture, including the following steps:

[0019] S1: Obtain historical video stream data of the time period when pollution occurred in the sewage outlet detection area in the pollution event storage module, perform brightness detection on all visible light monitoring video frames in the historical video stream data, and obtain the average brightness of each visible light monitoring video frame.

[0020] S2: The obtained average brightness is used as the ambient brightness value for the video frame capture time of visible light monitoring;

[0021] S3: Divide the ambient brightness value into five ambient brightness levels, and assign all visible light monitoring video frames to the corresponding ambient brightness level groups based on the ambient brightness value;

[0022] S4: Register all infrared surveillance video frames in the historical video stream data according to the timestamp and all visible light surveillance video frames that are grouped together. Then, fuse the paired infrared surveillance video frames and visible light surveillance video frames using a multi-resolution fusion algorithm to generate a fused image.

[0023] S5: Train the Faster R-CNN model on the fused images of each group to obtain five Faster R-CNN models for different ambient brightness levels.

[0024] As a further aspect of the present invention: the brightness detection of all visible light monitoring video frames in the historical video stream data in step S1 includes the following steps:

[0025] Convert visible light surveillance video frames into grayscale images;

[0026] Sum the pixel values ​​of a grayscale image;

[0027] Calculate the total number of pixels in a grayscale image;

[0028] Dividing the total pixel value of a grayscale image by the total number of pixels yields the average pixel value of the grayscale image, which is the average brightness of the video frame monitored by visible light.

[0029] As a further aspect of the present invention: step S3 includes the following steps:

[0030] An ambient brightness level is defined as follows: 0-20 cd / m², "Night"; 21-50 cd / m², "Dark"; 51-120 cd / m², "Dark"; 121-200 cd / m², "Bright"; and greater than 200 cd / m², "Bright".

[0031] Based on the fact that the ambient brightness value of each visible light monitoring video frame falls within the range of the five ambient brightness levels, all visible light monitoring video frames are assigned to the corresponding ambient brightness level groups.

[0032] As a further aspect of the present invention: the registration of all infrared surveillance video frames in the historical video stream data according to all visible light surveillance video frames grouped with the same timestamp in step S4 includes the following steps:

[0033] Pair all infrared surveillance video frames in the historical video stream data with all visible light surveillance video frames that are grouped together according to the timestamp;

[0034] By comparing the features of the paired infrared surveillance video frames and the visible light surveillance video frames, and using edges and corners as registration features, the number of matching registration points is determined to be greater than 4.

[0035] The registration points of infrared monitoring video frames and visible light monitoring video frames are matched with each other. Through a phase correlation algorithm, the infrared monitoring video frames and visible light monitoring video frames are registered to have the same viewing angle and pixel size.

[0036] As a further aspect of the present invention: the pollution image analysis module selects a Faster R-CNN model tailored to the corresponding ambient brightness to analyze the real-time video stream data based on the real-time ambient brightness, including the following steps:

[0037] The system acquires real-time video stream data from the video surveillance module. Brightness is detected in the visible light monitoring video frames within the real-time video stream data to obtain the average brightness of the real-time visible light monitoring video frames, which is then used as the real-time ambient brightness value.

[0038] The infrared monitoring video frames and visible light monitoring video frames in the real-time video stream data are paired according to the timestamp, and the paired infrared monitoring video frames and visible light monitoring video frames are fused to generate a real-time fused image.

[0039] Based on the average brightness of the video frames monitored in real time, the real-time fused images are assigned to the Faster R-CNN model corresponding to the ambient brightness level for analysis. If at least one of the real-time fused images above water or below water is identified as polluted, the system is determined to be polluted. The analysis results of whether the real-time video stream data is polluted are obtained and transmitted to the remote control module.

[0040] As a further aspect of the present invention, the fusion of the infrared monitoring video frames and the visible light monitoring video frames includes the following steps:

[0041] The video frames from infrared monitoring and visible light monitoring are preprocessed separately. The preprocessing includes image denoising, enhancement, and filtering.

[0042] Color features, texture features, and reflectance features were extracted from video frames of infrared monitoring and visible light monitoring, respectively. Color features were extracted using a color histogram, texture features were extracted using a texture filter and a gray-level co-occurrence matrix, and reflectance features were calculated using a reflectance model.

[0043] The color features, texture features, and reflectance features of infrared surveillance video frames and visible light surveillance video frames are fused using a weighted average method to obtain a fused feature vector.

[0044] The fused feature vectors are transformed into the final fusion result using wavelet transform.

[0045] The fused image is then post-processed, including image enhancement, denoising, and sharpening.

[0046] As a further aspect of the present invention: the video surveillance module includes:

[0047] The waterborne video monitoring unit is used to collect infrared and visible light monitoring data on the water surface of the sewage outlet detection area at a fixed position above the sewage outlet detection area.

[0048] The underwater video monitoring unit, assigned to the corresponding ambient light level Faster R-, is used to collect infrared and visible light data from a fixed underwater location in the sewage outlet detection area.

[0049] As a further aspect of the present invention: the waterborne video monitoring unit includes: a waterborne monitoring camera and a waterborne infrared camera, both of which are aimed at the sewage outlet detection area, and the waterborne infrared camera is located below the waterborne monitoring camera;

[0050] The underwater video monitoring unit includes an underwater monitoring camera and an underwater infrared camera. Both the underwater monitoring camera and the underwater infrared camera are installed at a fixed position at the bottom of the sewage outlet detection area. The underwater monitoring camera and the underwater infrared camera face the same direction. The underwater infrared camera is located to one side of the underwater monitoring camera.

[0051] The beneficial effects of this invention are:

[0052] This invention utilizes a rainfall analysis module to determine the presence of liquid flow in the sewage outlet detection area under sunny weather conditions using a flow velocity sensor. If no liquid flow is detected, the video monitoring module is turned off; if liquid flow is detected, the video monitoring module is turned on, and the sewage discharge is determined to be on a non-rainy day. If the flow velocity sensor does not detect liquid flow, there is no sewage discharge, allowing for rapid determination of whether sewage discharge has occurred in the sewage outlet detection area. This helps reduce the computational load of the image model training module and the pollution image analysis module. The video monitoring module is only turned on under rainy weather conditions, and the pollution status is analyzed by the image model training module and the pollution image analysis module. This facilitates sewage discharge detection on rainy days and makes it easier to train the Faster R-CNN model.

[0053] This invention utilizes video stream data from infrared and visible light monitoring of the sewage outlet detection area on the water surface and underwater. This allows for the acquisition of visible light monitoring video stream data during rainy weather, facilitating thorough analysis of sewage discharge during rainy days. The addition of infrared monitoring also enables the acquisition of infrared video stream data at night and in low-light conditions where visible light monitoring features are not readily apparent, ensuring accurate video stream data analysis even at night. Furthermore, it facilitates the training of different models based on the ambient brightness conditions at the time of video capture, improving the effectiveness of model training. Attached Figure Description

[0054] The invention will now be further described with reference to the accompanying drawings.

[0055] Figure 1 This is a schematic diagram of the system of the present invention. Implementation

[0056] 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.

[0057] Please see Figure 1 As shown, this invention is a river discharge monitoring system based on profile analysis, comprising:

[0058] The video monitoring module is used to collect video stream data of infrared and visible light monitoring of the sewage outlet detection area on the water surface, and to collect video stream data of infrared and visible light monitoring of the sewage outlet detection area underwater; the collected real-time video stream data is transmitted to the pollution image analysis module, and under the control of the remote control module, the real-time video stream data is transmitted to the pollution event storage module;

[0059] The rainfall analysis module is used to acquire real-time rainfall data. In sunny weather, it uses a flow velocity sensor in the sewage outlet detection area to determine whether there is liquid flow. If no liquid flow is detected, the video monitoring module is turned off. If liquid flow is detected, the video monitoring module is turned on, and it is determined that the sewage is being discharged on a non-rainy day. In rainy weather, the video monitoring module is turned on to monitor whether the sewage is in a polluted state.

[0060] The pollution event storage module is used to store all historical video stream data of the pollution period in the sewage outlet detection area, and periodically transmit all historical video stream data of the pollution period to the image model training module.

[0061] The image model training module acquires all historical video stream data during the pollution period, fuses the video stream data from infrared monitoring and visible light monitoring, and trains Faster R-CNN models for different ambient brightness levels using the fused data based on the ambient brightness when the video streams were captured.

[0062] The pollution image analysis module acquires real-time video stream data from the video surveillance module and selects a Faster R-CNN model tailored to the corresponding ambient brightness to analyze the real-time video stream data.

[0063] Specifically, in sunny weather, the rainfall analysis module uses flow velocity sensors in the sewage outlet detection area to determine if there is liquid flow. If no liquid flow is detected, the video monitoring module is turned off; if liquid flow is detected, the video monitoring module is turned on, and the sewage discharge is determined to be non-rainy. If the flow velocity sensor does not detect liquid flow, there is no sewage discharge. This allows for quick determination of whether sewage discharge has occurred in the sewage outlet detection area, which helps reduce the computational load of the image model training module and the pollution image analysis module. The video monitoring module is only turned on in rainy weather. The pollution status is analyzed by the image model training module and the pollution image analysis module, which helps to detect sewage discharge in rainy weather and facilitates better training of the Faster R-CNN model.

[0064] Meanwhile, based on the video stream data from infrared and visible light monitoring of the sewage outlet detection area on the water, and the video stream data from infrared and visible light monitoring of the sewage outlet detection area underwater, video stream data from visible light monitoring on the water and underwater can be obtained during rainy weather. This facilitates a thorough analysis of sewage discharge during rainy days. Furthermore, the addition of infrared monitoring makes it easier to acquire infrared video stream data at night and in dimly lit environments where visible light monitoring features are not obvious, enabling accurate analysis of video stream data even at night. It also facilitates training different models based on the ambient brightness conditions at the time of different video captures, improving the effectiveness of model training.

[0065] In one embodiment of the present invention, it further includes:

[0066] The remote control module, based on the analysis results from the rainfall analysis module and the pollution image analysis module, determines whether the pollution occurs during non-rainy days when sewage discharge occurs and the real-time video stream data analysis indicates a pollution state. In these cases, the remote control module controls the sampling module to sample water quality in the sewage outlet detection area and controls the alarm module to issue an abnormal sewage discharge alarm. The remote control module also controls the video monitoring module to transmit the real-time video stream data to the pollution event storage module for storage. Conversely, if the real-time video stream data is not in a pollution state, the remote control module controls the video monitoring module to stop transmitting the real-time video stream data to the pollution event storage module.

[0067] The sampling module is used to sample the water quality in the sewage outlet detection area when the area is polluted.

[0068] The alarm module is used to issue abnormal discharge alarms in response to pollution events;

[0069] The abnormal discharge alarm includes sending abnormal discharge alarm information and satellite positioning information to a designated object via a wireless communication network.

[0070] Specifically, the remote control module facilitates the sampling module's ability to sample real-time pollution conditions, enabling timely and effective preservation of pollution evidence. Simultaneously, the video monitoring module transmits real-time video stream data to the pollution event storage module only when pollution occurs, aiding in better preservation of video stream data and facilitating the determination of discharge duration. Furthermore, the alarm module sends abnormal discharge alarm information and satellite positioning information to designated targets via a wireless communication network, enabling rapid location of the polluted discharge outlet detection area.

[0071] In one embodiment of the present invention, the image model training module fuses video stream data from infrared monitoring and visible light monitoring, and trains Faster R-CNN models for different ambient brightness levels using the fused data, based on the ambient brightness during video stream capture. This includes the following steps:

[0072] S1: Obtain historical video stream data of the time period when pollution occurred in the sewage outlet detection area in the pollution event storage module, perform brightness detection on all visible light monitoring video frames in the historical video stream data, and obtain the average brightness of each visible light monitoring video frame.

[0073] S2: The obtained average brightness is used as the ambient brightness value for the video frame capture time of visible light monitoring;

[0074] S3: Divide the ambient brightness value into five ambient brightness levels, and assign all visible light monitoring video frames to the corresponding ambient brightness level groups based on the ambient brightness value;

[0075] S4: Register all infrared surveillance video frames in the historical video stream data according to the timestamp and all visible light surveillance video frames that are grouped together. Then, fuse the paired infrared surveillance video frames and visible light surveillance video frames using a multi-resolution fusion algorithm to generate a fused image.

[0076] S5: Train the Faster R-CNN model on the fused images of each group to obtain five Faster R-CNN models for different ambient brightness levels.

[0077] Specifically, the brightness of video frames varies under different environments. By dividing ambient brightness values ​​into five levels and training the model separately for each level, the model's adaptability to different scenes can be enhanced, thereby improving its generalization ability. Training the model separately for different ambient brightness levels allows it to become more familiar with target behavior under various backgrounds, thus improving detection accuracy and precision. Simultaneously, it reduces false positives caused by changes in ambient brightness, improving the model's stability and robustness. This also leads to better results when used in a wider range of scenarios, providing monitoring equipment users with a more intelligent and efficient user experience.

[0078] Simultaneously, a multi-resolution fusion algorithm is used to register and fuse paired infrared and visible light surveillance video frames. Infrared and visible light surveillance complement each other, and their simultaneous use improves the accuracy and precision of target detection, thus enhancing the performance of the surveillance system. Furthermore, training the model using images from different modalities increases its adaptability to complex scenes and improves its robustness. By fusing images from different modalities, the trained model can be applied to more scenarios, improving its generalization ability.

[0079] In one embodiment of the present invention, step S1, which involves performing brightness detection on all visible light monitoring video frames in the historical video stream data, includes the following steps:

[0080] Convert visible light surveillance video frames into grayscale images;

[0081] Sum the pixel values ​​of a grayscale image;

[0082] Calculate the total number of pixels in a grayscale image;

[0083] Dividing the total pixel value of a grayscale image by the total number of pixels yields the average pixel value of the grayscale image, which is the average brightness of the video frame for visible light monitoring.

[0084] In one embodiment of the present invention, step S3 includes the following steps:

[0085] An ambient brightness level is defined as follows: 0-20 cd / m², "Night"; 21-50 cd / m², "Dark"; 51-120 cd / m², "Dark"; 121-200 cd / m², "Bright"; and greater than 200 cd / m², "Bright".

[0086] Based on the fact that the ambient brightness value of each visible light monitoring video frame falls within the range of the five ambient brightness levels, all visible light monitoring video frames are assigned to the corresponding ambient brightness level groups.

[0087] Specifically, the ambient brightness value is determined based on the average brightness of the video frames monitored by visible light. This facilitates the subsequent model training by framing the video frames into groups corresponding to the ambient brightness.

[0088] In one embodiment of the present invention, step S4, which involves registering all infrared surveillance video frames in the historical video stream data according to all visible light surveillance video frames grouped together with the same timestamp, includes the following steps:

[0089] Pair all infrared surveillance video frames in the historical video stream data with all visible light surveillance video frames that are grouped together according to the timestamp;

[0090] By comparing the features of the paired infrared surveillance video frames and the visible light surveillance video frames, and using edges and corners as registration features, the number of matching registration points is determined to be greater than 4.

[0091] The registration points of infrared monitoring video frames and visible light monitoring video frames are matched with each other. Through a phase correlation algorithm, the infrared monitoring video frames and visible light monitoring video frames are registered to have the same viewing angle and pixel size.

[0092] Specifically, by matching the registration points of infrared and visible light video frames and using a phase correlation algorithm, a high-precision multimodal image can be formed, thus increasing the accuracy of the subsequent fused image. This registration-fusing process eliminates differences between infrared and visible light images caused by viewing angles, ensuring that the infrared and visible light video frames have the same viewing angle and pixel size. Furthermore, image registration allows for correction and denoising, further improving the accuracy of subsequent analysis results.

[0093] In one embodiment of the present invention, the pollution image analysis module selects a Faster R-CNN model for the corresponding ambient brightness to analyze the real-time video stream data based on the real-time ambient brightness, including the following steps:

[0094] The system acquires real-time video stream data from the video surveillance module. Brightness is detected in the visible light monitoring video frames within the real-time video stream data to obtain the average brightness of the real-time visible light monitoring video frames, which is then used as the real-time ambient brightness value.

[0095] The infrared monitoring video frames and visible light monitoring video frames in the real-time video stream data are paired according to the timestamp, and the paired infrared monitoring video frames and visible light monitoring video frames are fused to generate a real-time fused image.

[0096] Based on the average brightness of the video frames monitored in real time, the real-time fused images are assigned to the Faster R-CNN model corresponding to the ambient brightness level for analysis. If at least one of the real-time fused images above water or below water is identified as polluted, the system is determined to be polluted. The analysis results of whether the real-time video stream data is polluted are obtained and transmitted to the remote control module.

[0097] In one embodiment of the present invention, fusing infrared surveillance video frames and visible light surveillance video frames includes the following steps:

[0098] The video frames from infrared monitoring and visible light monitoring are preprocessed separately. The preprocessing includes image denoising, enhancement, and filtering.

[0099] Color features, texture features, and reflectance features were extracted from video frames of infrared monitoring and visible light monitoring, respectively. Color features were extracted using a color histogram, texture features were extracted using a texture filter and a gray-level co-occurrence matrix, and reflectance features were calculated using a reflectance model.

[0100] The color features, texture features, and reflectance features of infrared surveillance video frames and visible light surveillance video frames are fused using a weighted average method to obtain a fused feature vector.

[0101] The fused feature vectors are transformed into the final fusion result using wavelet transform.

[0102] The fused image is then post-processed, including image enhancement, denoising, and sharpening.

[0103] Specifically, when color, texture, and reflectance features are fused into a single feature vector using a weighted average method, the combined ability of different features to describe a specific scene is more representative and robust than using a single feature. Furthermore, wavelet transform can adapt to feature variations at different scales and frequencies, while also possessing good image localization and compression properties. Therefore, this fusion approach can improve the precision and accuracy of the analysis.

[0104] In one embodiment of the present invention, the video surveillance module includes:

[0105] The waterborne video monitoring unit is used to collect infrared and visible light monitoring data on the water surface of the sewage outlet detection area at a fixed position above the sewage outlet detection area.

[0106] The underwater video monitoring unit, assigned to the corresponding ambient light level Faster R-, is used to collect infrared and visible light data from a fixed underwater location in the sewage outlet detection area.

[0107] In one embodiment of the present invention, the waterborne video monitoring unit includes: a waterborne monitoring camera and a waterborne infrared camera, both of which are aimed at the sewage outlet detection area, and the waterborne infrared camera is located below the waterborne monitoring camera.

[0108] The underwater video surveillance unit includes an underwater monitoring camera and an underwater infrared camera. Both are mounted in fixed positions at the bottom of the sewage outlet detection area. The underwater monitoring camera faces the same direction as the underwater infrared camera, which is located to one side of the underwater monitoring camera. This design ensures that the infrared and visible light monitoring angles are roughly the same, facilitating subsequent fusion processing of the video frames from the infrared and visible light monitoring.

[0109] In the description of this invention, it should be understood that the terms "upper," "lower," "left," and "right," etc., indicating orientation or positional relationships based on the orientation or positional relationships shown in the accompanying drawings, are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or a specific orientational structure and operation. Therefore, they should not be construed as limitations on the invention. Furthermore, "first" and "second" are only for descriptive purposes and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "multiple" means two or more.

[0110] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal communication between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0111] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A sewage discharge monitoring system for river outfalls based on profile analysis, characterized in that, include: The video monitoring module is used to collect video stream data of infrared and visible light monitoring of the sewage outlet detection area on the water surface, and to collect video stream data of infrared and visible light monitoring of the sewage outlet detection area underwater; the collected real-time video stream data is transmitted to the pollution image analysis module, and under the control of the remote control module, the real-time video stream data is transmitted to the pollution event storage module; The rainfall analysis module is used to acquire real-time rainfall data. In sunny weather, it uses a flow velocity sensor in the sewage outlet detection area to determine whether there is liquid flow. If no liquid flow is detected, the video monitoring module is turned off. If liquid flow is detected, the video monitoring module is turned on, and it is determined that the sewage is being discharged on a non-rainy day. In rainy weather, the video monitoring module is turned on to monitor whether the sewage is in a polluted state. The pollution event storage module is used to store all historical video stream data of the pollution period in the sewage outlet detection area, and periodically transmit all historical video stream data of the pollution period to the image model training module. The image model training module acquires all historical video stream data during the pollution period, fuses the video stream data from infrared monitoring and visible light monitoring, and trains Faster R-CNN models for different ambient brightness levels using the fused data based on the ambient brightness when the video streams were captured. The pollution image analysis module acquires real-time video stream data from the video surveillance module and selects a Faster R-CNN model tailored to the corresponding ambient brightness to analyze the real-time video stream data.

2. The river discharge monitoring system based on profile analysis according to claim 1, characterized in that, Also includes: The remote control module, based on the analysis results of the rainfall analysis module and the pollution image analysis module, when there is sewage discharge on a non-rainy day and the real-time video stream data analysis indicates a pollution state, controls the sampling module to sample the water quality of the sewage outlet detection area, and controls the alarm module to issue an abnormal sewage discharge alarm for the pollution event. The remote control module also controls the video monitoring module to transmit the real-time video stream data to the pollution event storage module for pollution event storage. When the real-time video stream data is not in a contaminated state, the remote control module controls the video monitoring module to stop transmitting the real-time video stream data to the contamination event storage module; The sampling module is used to sample the water quality in the sewage outlet detection area when the area is polluted. The alarm module is used to issue abnormal discharge alarms in response to pollution events; The abnormal discharge alarm includes sending abnormal discharge alarm information and satellite positioning information to a designated object via a wireless communication network.

3. The river discharge monitoring system based on profile analysis according to claim 1, characterized in that, The image model training module fuses video stream data from infrared and visible light surveillance. Based on the ambient brightness during video stream capture, it uses the fused data to train Faster R-CNN models for different ambient brightness levels, including the following steps: S1: Obtain historical video stream data of the time period when pollution occurred in the sewage outlet detection area in the pollution event storage module, perform brightness detection on all visible light monitoring video frames in the historical video stream data, and obtain the average brightness of each visible light monitoring video frame. S2: The obtained average brightness is used as the ambient brightness value for the video frame capture time of visible light monitoring; S3: Divide the ambient brightness value into five ambient brightness levels, and assign all visible light monitoring video frames to the corresponding ambient brightness level groups based on the ambient brightness value; S4: Register all infrared surveillance video frames in the historical video stream data according to the timestamp and all visible light surveillance video frames that are grouped together. Then, fuse the paired infrared surveillance video frames and visible light surveillance video frames using a multi-resolution fusion algorithm to generate a fused image. S5: Train the Faster R-CNN model on the fused images of each group to obtain five Faster R-CNN models for different ambient brightness levels.

4. The river discharge monitoring system based on profile analysis according to claim 3, characterized in that, Step S1, which involves brightness detection of all visible light monitoring video frames in the historical video stream data, includes the following steps: Convert visible light surveillance video frames into grayscale images; Sum the pixel values ​​of a grayscale image; Calculate the total number of pixels in a grayscale image; Dividing the total pixel value of a grayscale image by the total number of pixels yields the average pixel value of the grayscale image, which is the average brightness of the video frame monitored by visible light.

5. The river discharge monitoring system based on image analysis according to claim 3, characterized in that, Step S3 includes the following steps: The ambient light value is set between 0 and 20 cd / m². 2 Within this range, the ambient light level is "Nighttime," with an ambient light value ranging from 21 to 50 cd / m². 2 Within this range, the ambient brightness level is "dark," with ambient brightness values ​​ranging from 51 to 120 cd / m². 2 Within this range, the ambient brightness level is "relatively dim," with an ambient brightness value ranging from 121 to 200 cd / m². 2 Within the specified range, the ambient brightness level is "relatively bright," with an ambient brightness value greater than 200 cd / m². 2 Within the specified range, the ambient brightness level is "bright". Based on the fact that the ambient brightness value of each visible light monitoring video frame falls within the range of the five ambient brightness levels, all visible light monitoring video frames are assigned to the corresponding ambient brightness level groups.

6. The river discharge monitoring system based on profile analysis according to claim 3, characterized in that, Step S4, which involves registering all infrared surveillance video frames in the historical video stream data according to all visible light surveillance video frames grouped together with the same timestamp, includes the following steps: Pair all infrared surveillance video frames in the historical video stream data with all visible light surveillance video frames that are grouped together according to the timestamp; By comparing the features of the paired infrared surveillance video frames and the visible light surveillance video frames, and using edges and corners as registration features, the number of matching registration points is determined to be greater than 4. The registration points of infrared monitoring video frames and visible light monitoring video frames are matched with each other. Through a phase correlation algorithm, the infrared monitoring video frames and visible light monitoring video frames are registered to have the same viewing angle and pixel size.

7. The river discharge monitoring system based on profile analysis according to claim 1, characterized in that, The pollution image analysis module, based on the real-time ambient brightness, selects a Faster R-CNN model tailored to the corresponding ambient brightness to analyze the real-time video stream data, including the following steps: The system acquires real-time video stream data from the video surveillance module. Brightness is detected in the visible light monitoring video frames within the real-time video stream data to obtain the average brightness of the real-time visible light monitoring video frames, which is then used as the real-time ambient brightness value. The infrared monitoring video frames and visible light monitoring video frames in the real-time video stream data are paired according to the timestamp, and the paired infrared monitoring video frames and visible light monitoring video frames are fused to generate a real-time fused image. Based on the average brightness of the video frames monitored in real time, the real-time fused images are assigned to the Faster R-CNN model corresponding to the ambient brightness level for analysis. If at least one of the real-time fused images above water or below water is identified as polluted, the system is determined to be polluted. The analysis results of whether the real-time video stream data is polluted are obtained and transmitted to the remote control module.

8. The river discharge monitoring system based on image analysis according to any one of claims 3-7, characterized in that, The fusion of infrared surveillance video frames and visible light surveillance video frames includes the following steps: The video frames from infrared monitoring and visible light monitoring are preprocessed separately. The preprocessing includes image denoising, enhancement, and filtering. Color features, texture features, and reflectance features were extracted from video frames of infrared monitoring and visible light monitoring, respectively. Color features were extracted using a color histogram, texture features were extracted using a texture filter and a gray-level co-occurrence matrix, and reflectance features were calculated using a reflectance model. The color features, texture features, and reflectance features of infrared surveillance video frames and visible light surveillance video frames are fused using a weighted average method to obtain a fused feature vector. The fused feature vectors are transformed into the final fusion result using wavelet transform. The fused image is then post-processed, including image enhancement, denoising, and sharpening.

9. The river discharge monitoring system based on profile analysis according to claim 1, characterized in that, The video surveillance module includes: The waterborne video monitoring unit is used to collect infrared and visible light monitoring data on the water surface of the sewage outlet detection area at a fixed position above the sewage outlet detection area. The underwater video monitoring unit, assigned to the corresponding ambient light level Faster R-, is used to collect infrared and visible light data from a fixed underwater location in the sewage outlet detection area.

10. The river discharge monitoring system based on profile analysis according to claim 9, characterized in that, The waterborne video monitoring unit includes a waterborne monitoring camera and a waterborne infrared camera. Both the waterborne monitoring camera and the waterborne infrared camera are aimed at the sewage outlet detection area, and the waterborne infrared camera is located below the waterborne monitoring camera. The underwater video monitoring unit includes an underwater monitoring camera and an underwater infrared camera. Both the underwater monitoring camera and the underwater infrared camera are installed at a fixed position at the bottom of the sewage outlet detection area. The underwater monitoring camera and the underwater infrared camera face the same direction. The underwater infrared camera is located to one side of the underwater monitoring camera.

Images