| सारांश: | Water leakage detection in household pipelines is crucial for maintaining the efficiency of water supply systems. However, residents often become aware of leakage issues only after receiving unexpectedly high water bills. Recent research has explored IoT-based monitoring systems and the application of Machine Learning (ML), particularly Deep Learning (DL), to enhance leak detection accuracy. A common drawback of many existing solutions is their intrusive nature, which often requires modifications to the current piping infrastructure. To address this limitation, this project proposes a non-intrusive smart water monitoring system that integrates computer vision and deep learning techniques for real-time leak detection. The system employs a YOLOv8 model to accurately detect and localize digital sensor displays, followed by an Optical Character Recognition (OCR) component to extract flow rate and pressure values. Monitoring is conducted in real time using a webcam connected to a Raspberry Pi. When the detected values fall below a predefined threshold, a Human Machine Interface (HMI) developed in Flask sends immediate alerts to users. The model is trained using the Google Colaboratory platform and deployed through Python in Visual Studio Code, with a Flask-based dashboard used to present the results. When combining the techniques YOLOv8 and PaddleOCR, the system achieves a detection accuracy of 81.82% with a processing time of 7.96 seconds per image, demonstrating its effectiveness for time-sensitive and practical leak detection applications.
|
|---|
