Intelligent AI detection system and method for river channel occupation detection

Abstract

The invention relates to the field of river channel occupation detection, and discloses an intelligent AI detection system and method for river channel occupation detection, and the system comprises an image collection module, a controller, and a cloud platform. The image acquisition module is connected with the controller through a wired network line; the controller is in signal communication with the cloud platform through a wireless network; the image acquisition module is used for acquiring river channel images; the controller detects the collected riverway image through a carried riverway occupation detection model, and transmits a detection result to the cloud platform; according to the invention, the problems of low efficiency, many false alarms and insufficient coverage of a traditional monitoring means can be solved, the intelligent degree of river channel monitoring is effectively improved, all-weather automatic monitoring and localized data processing are realized, and the real-time performance, reliability and stability of the system are significantly enhanced.

Description

Technical Field

[0001] The present invention relates to the field of river channel encroachment detection, and in particular to an intelligent AI detection system and method for river channel encroachment detection. Background Art

[0002] With the acceleration of urbanization, the problem of illegal occupation of rivers has become increasingly prominent. Illegal buildings, accumulations, etc. occupying rivers not only affect the ecological environment of rivers, but also weaken the flood discharge capacity of rivers and increase the risk of natural disasters such as floods. At the same time, river bank erosion and retreat are also key issues affecting river management. Gradual erosion of river banks may lead to serious ecological damage and human property losses.

[0003] Traditional river monitoring methods mostly rely on manual inspections or fixed equipment monitoring, which is not only time-consuming and labor-intensive, but also difficult to achieve real-time monitoring of large areas. At the same time, traditional water level meters, laser rangefinders and other equipment can only provide single-dimensional information and cannot effectively identify the overall picture of river occupation and riverbank changes.

[0004] At the same time, traditional CRNN may not be able to effectively extract features when dealing with complex scenes (such as various natural lighting and weather changes during river monitoring), resulting in low recognition accuracy. In addition, traditional CRNN may face high computing resource requirements and large delays when processing real-time video, and may not be able to fully capture the key temporal features in the image when processing time series data, and its generalization ability to different data sets may also be insufficient. Summary of the invention

[0005] The purpose of the present invention is to provide an intelligent AI detection system and method for river occupancy detection, so as to solve the problems of low efficiency, many false alarms and insufficient coverage of traditional river occupancy monitoring means.

[0006] Specifically, the present invention provides an intelligent AI detection system for river occupancy detection, comprising: an image acquisition module, a controller, and a cloud platform;

[0007] The image acquisition module is connected to the controller via a wired network line;

[0008] The controller communicates with the cloud platform via wireless network signals;

[0009] The image acquisition module is used to acquire river channel images;

[0010] The controller detects the collected river channel images through the carried river channel occupancy detection model, and transmits the detection results to the cloud platform.

[0011] The beneficial effects provided by the present invention are: the system uses visual technology to monitor river occupation, river bank retreat, erosion status and water level changes in real time. The CRNN of this system adds an attention mechanism to enhance the accuracy of feature extraction, so that the model can more effectively identify important information related to river occupation and water level changes. In addition, the system adopts an adaptive learning rate adjustment strategy (such as AdamW optimizer) and specific data enhancement methods (including random cropping, rotation and color transformation) to accelerate the convergence speed of the model and improve robustness, ensuring that the model can perform well in complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 It is a schematic diagram of the system structure of the present invention;

[0013] Figure 2 1 is a schematic diagram of the structure of a convolutional recurrent neural network CRNN improved in an embodiment of the present invention;

[0014] Figure 3 A schematic diagram of the structure of a controller in an embodiment of the present invention;

[0015] Figure 4 It is a schematic diagram of the overall working process of the system of the present invention.

[0016] Above Figure 1 , 3 middle:

[0017] 1-camera; 2-improved convolutional recurrent neural network CRNN algorithm; 3-controller; 31-control module; 32-judgment module; 33-storage module; 34-receiving module; 35-sending module; 4-cloud platform. DETAILED DESCRIPTION

[0018] To make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0019] Before formally describing the present invention, the scheme of the present invention is first generally described for easy understanding.

[0020] Please refer to Figure 1 , Figure 1 It is a schematic diagram of the system structure of the present invention.

[0021] The present invention provides an intelligent AI detection system for river occupancy detection, comprising: an image acquisition module, a controller and a cloud platform;

[0022] The image acquisition module is connected to the controller via a wired network line;

[0023] The controller communicates with the cloud platform via wireless network signals;

[0024] The image acquisition module is used to acquire river channel images;

[0025] The controller detects the collected river channel images through the carried river channel occupancy detection model, and transmits the detection results to the cloud platform.

[0026] It should be noted that the image acquisition module is a camera.

[0027] It should be noted that the controller includes: a control module, a judgment module, a storage module, a sending module and a receiving module.

[0028] It should be noted that the control module is used to control the data acquisition frequency of the image acquisition module.

[0029] It should be noted that the judgment module is used to judge the working state of the image acquisition module.

[0030] It should be noted that the sending module is used to send the collected images and image detection results to the cloud platform; the receiving module is used to receive feedback instructions from the cloud platform.

[0031] It should be noted that the river occupancy detection model adopts an improved convolutional recurrent neural network CRNN, introduces an attention mechanism into the convolutional recurrent neural network CRNN, and adopts an adaptive learning adjustment strategy during model training.

[0032] An intelligent AI detection method for river occupancy detection is applied to the detection system, and the method comprises the following steps:

[0033] S1, controlling the acquisition frequency of the image acquisition module through the control module, and acquiring the acquired image;

[0034] S2, the collected images are detected by the river occupancy detection model to obtain the river occupancy type;

[0035] S3, storing the collected images and the corresponding river occupancy types locally through a storage module;

[0036] S4, sending the collected images and the corresponding river occupancy types to the cloud platform through the sending module;

[0037] S5. Receive feedback instructions from the cloud platform through the receiving module;

[0038] It should be noted that, during the detection process, the judgment module is also used to judge the abnormal situation of the image acquisition module in real time.

[0039] In the present invention, as an embodiment, the camera is Sony SNC-VM772R.

[0040] The core controller is RK3588, which can efficiently process image data from the camera and perform real-time analysis of river occupancy, erosion and water level changes.

[0041] The cloud platform can be any one of the cloud service platforms such as Baidu, Alibaba Cloud, Tencent Cloud, etc.

[0042] As a specific embodiment, the present invention deploys multiple cameras 1 in key areas along the river. These cameras have high-definition, wide-angle and night vision functions, and can collect real-time video data of the river and riverbank around the clock. Whether it is day or night, the camera can ensure clear image quality. The camera transmits the collected data to the controller 3 of the system through the network port connection, ensuring the efficiency and real-time performance of data transmission.

[0043] The camera 1 sends the collected video data of the river channel and river bank to the controller 3, and the controller 3 uploads the processed information to the cloud platform 4 through the network port communication, and then sends it to the expert system for secondary judgment. The controller 3 provides power for the camera 1.

[0044] Please refer to Figure 2 , Figure 2 1 is a schematic diagram of the structure of the improved convolutional recurrent neural network CRNN according to an embodiment of the present invention. In which, the attention mechanism module is added between the pooling layer and the time series analysis module, and the network is better able to extract features by adding the attention mechanism module.

[0045] The controller 3 has efficient image processing and analysis capabilities. The controller 3 processes the data collected by the camera 1 in real time, and runs the improved convolutional recurrent neural network CRNN algorithm 2 to perform spatial and temporal series analysis on the image data.

[0046] In addition, the system also uses an adaptive learning rate adjustment strategy, such as the AdamW optimizer, to accelerate the convergence of the model. This optimization method dynamically adjusts the learning rate so that the model can adapt to data characteristics more quickly during training, significantly shortening training time and improving efficiency. The introduction of adaptive learning rate not only accelerates the training process of the model, but also improves the performance of the final model.

[0047] In order to enhance the robustness of the model, this system has designed a variety of specific data enhancement methods, including random cropping, rotation, and color transformation. These methods not only increase the diversity of training data, but also improve the model's adaptability to different input conditions. For example, random cropping can simulate different perspectives and object scales, ensuring that the model can process images from different camera angles; rotation can effectively cope with changes in camera installation angles and enhance the adaptability of the model; color transformation helps the model to effectively identify under various lighting conditions. This diverse data enhancement strategy enables the model to better handle complex situations in real application environments and improve the accuracy and reliability of detection.

[0048] To verify the effectiveness of this system, several experiments were conducted. Table 1 below shows the detection accuracy of the model under different environmental conditions.

[0049] Table 1 Model detection accuracy under different environmental conditions

[0050] Environmental conditions Detection accuracy Detection time (ms) False Positive Rate Sunny day 95％ 50 2％ Cloudy during the day 93％ 55 3％ at night 90％ 60 5％ rain 85％ 65 7％

[0051] As can be seen from the table, the system can maintain a high detection accuracy under various environmental conditions, especially under sunny conditions during the day, with an accuracy of 95%. Although the accuracy drops slightly at night and on rainy days, it still remains above 90%, showing the good robustness of the system.

[0052] In the experiment, the average time for the system to process images was also kept within an acceptable range, ensuring the need for real-time monitoring. The false alarm rate was relatively low, indicating that the system has a high degree of accuracy in identifying river occupancy and water level changes.

[0053] Specifically, for the detection of river occupation, the system first extracts spatial features in the image through CNN to identify whether the river is occupied for a long time by illegal deposits, buildings, etc. Then, through RNN time series analysis, the system can determine whether these occupied objects are short-term floating objects or have a tendency to occupy the river for a long time, thereby achieving accurate identification and dynamic tracking of river occupation.

[0054] In terms of riverbank retreat detection, the system builds a baseline model of riverbank edge contours based on CRNN. ​​By comparing images at different time points, CRNN analyzes changes in riverbank contours to determine whether there are signs of riverbank retreat. Its CNN part extracts the edge and morphological features of the riverbank from the image, while the RNN part tracks the changes of these features over time, thereby capturing subtle trends in riverbank retreat.

[0055] For monitoring water level changes, the CRNN model uses visual technology to identify the boundaries of the water body, that is, the specific location of the water level line. Through the river channel images collected by the camera in real time, the system can accurately identify the water level line, and combine historical data to analyze the trend of water level changes and predict the risk of floods or droughts in advance.

[0056] In riverbank erosion detection, the CNN part of the CRNN model identifies the loss of surface soil or vegetation on the riverbank through changes in soil color, surface structure, and texture in the image. As erosion progresses, the RNN can analyze the rate, range, and depth of erosion through image changes over multiple time steps.

[0057] In order to ensure the accuracy of the detection results, this system has designed a two-level discrimination mechanism.

[0058] When the system detects an abnormal situation (such as river occupation, river bank retreat, erosion or water level changes), it will automatically generate a report including detailed information such as image data, timestamp, and detection results, and upload it to the cloud platform4 in real time for management personnel to view and analyze. Based on the first-level judgment, the detection data will also be transmitted to the second-level expert system for more in-depth analysis. The expert system combines historical data, geographic information, and environmental factors (such as weather forecasts, rainfall, etc.) to make a secondary judgment on the abnormal phenomenon to reduce the possibility of false alarms or missed alarms. The system can not only automatically monitor, but also issue intelligent warnings in the event of major abnormalities, reminding management departments to take emergency measures.

[0059] Specifically, see Figure 3 As shown, in this embodiment of the present invention, the controller 3 includes:

[0060] The control module 31 is used to control the frequency of data collection to meet different river monitoring requirements.

[0061] The judgment module 32 is used to judge the working state of the camera 1 and handle abnormal situations, such as equipment failure, signal loss or data abnormality.

[0062] The storage module 33 is used to locally store the collected river occupancy, river bank retreat, water level changes, and erosion data, and supports local storage and query of historical data.

[0063] The sending module 35 is used to send the collected data and analysis results to the cloud platform 4.

[0064] The receiving module 34 is used to receive the feedback instruction sent by the cloud platform 4.

[0065] Finally, please refer to Figure 4 , Figure 4 It is a schematic diagram of the overall working process of the system of the present invention.

[0066] Specifically, in the embodiment of the present invention, the camera 1 is a Sony SNC-VM772R, which supports wide dynamic range (WDR) technology, adapts to complex lighting conditions, has a built-in infrared function, supports nighttime monitoring, and has a sturdy housing that is waterproof and dustproof (IP66 level).

[0067] Specifically, in the embodiment of the present invention, the controller 3 uses RK3588. RK3588 has powerful processing capabilities, supports complex AI algorithms such as CNN and RNN, can efficiently process image data from the camera 1, and analyze river occupation, erosion and water level changes in real time.

[0068] Specifically, in the embodiment of the present invention, the cloud platform 4 can be any one of the cloud service platforms such as Baidu, Alibaba Cloud, Tencent Cloud, etc.

[0069] As a result, the cloud platform 4 can timely view and analyze the data uploaded by the system, comprehensively grasp the situations such as river occupation, river bank erosion, water level changes, etc., and provide data support and intelligent analysis for relevant decisions.

[0070] The beneficial effects of the present invention are:

[0071] 1. Through the combination of cameras and edge computing boards, the system can analyze in real time whether the river is occupied, eroded or the water level changes, and realize all-weather automatic monitoring.

[0072] 2. Using the spatial and temporal series analysis of the improved CRNN model, the system can accurately identify the occupancy and erosion of the river channel and avoid false positives and false negatives.

[0073] 3. The core controller improves the real-time and efficiency of data processing, can quickly perform data analysis on site, and reduce data transmission delays, thereby improving the system's response speed to emergencies.

[0074] 4. Through the combination of the automatic reporting system and the secondary judgment of the expert system, the system can quickly report abnormal situations while ensuring the accuracy and reliability of detection.

[0075] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. An intelligent AI detection system for river occupancy detection, characterized in that: include: Image acquisition module, controller and cloud platform; The image acquisition module is connected to the controller via a wired network line; The controller communicates with the cloud platform via wireless network signals; The image acquisition module is used to acquire river channel images; The controller detects the collected river channel images through the carried river channel occupancy detection model, and transmits the detection results to the cloud platform.

2. An intelligent AI detection system for river occupancy detection as claimed in claim 1, characterized in that: The image acquisition module is a camera.

3. The intelligent AI detection system for river occupancy detection according to claim 1, characterized in that: The controller comprises: a control module, a judgment module, a storage module, a sending module and a receiving module.

4. An intelligent AI detection system for river occupancy detection as claimed in claim 3, characterized in that: The control module is used to control the data acquisition frequency of the image acquisition module.

5. The intelligent AI detection system for river occupancy detection according to claim 3, characterized in that: The judging module is used to judge the working state of the image acquisition module.

6. The intelligent AI detection system for river occupancy detection according to claim 3, characterized in that: The storage module is used to store the images collected by the image collection module and the image detection results.

7. The intelligent AI detection system for river occupancy detection according to claim 3, characterized in that: The sending module is used to send the collected images and image detection results to the cloud platform; the receiving module is used to receive feedback instructions from the cloud platform.

8. The intelligent AI detection system for river occupancy detection according to claim 1, characterized in that: The river occupancy detection model adopts an improved convolutional recurrent neural network CRNN, introduces an attention mechanism into the convolutional recurrent neural network CRNN, and adopts an adaptive learning adjustment strategy during model training.

9. An intelligent AI detection method for river occupancy detection, applied to the detection system according to any one of claims 1 to 8, characterized in that: The method comprises the following steps: S1, controlling the acquisition frequency of the image acquisition module through the control module, and acquiring the acquired image; S2, the collected images are detected by the river occupancy detection model to obtain the river occupancy type; S3, storing the collected images and the corresponding river occupancy types locally through a storage module; S4, sending the collected images and the corresponding river occupancy types to the cloud platform through the sending module; S5. Receive feedback instructions from the cloud platform through the receiving module.

10. An intelligent AI detection method for river occupancy detection as claimed in claim 9, characterized in that: During the detection process, the judgment module is also used to judge the abnormal situation of the image acquisition module in real time.

Images