Research on River Floating Objects Detection Method Based on Deep Learning

The environmental issues have become increasingly prominent, and water environments have remained a major societal concern. Surface pollutants in rivers are one of the key factors damaging water environments. Given the large number and wide distribution of rivers, manual river inspections and the collection of floating objects are time-consuming and labor-intensive. In this study, a river surface floating objects detection method based on deep learning is proposed. To address the noise issues caused by interferences in natural scenes, an image preprocessing method based on median filtering and Laplace operator is studied. In order to select the most suitable deep learning algorithm, this work compares the YOLOv5, Faster-RCNN, and SSD algorithms. After comparative analysis, an optimized YOLOv5 model is proposed for detecting floating objects on river surfaces. Finally, a visualization platform is developed based on PyQt5, enabling real-time monitoring and display of floating objects in river surveillance videos. Experimental results demonstrate that this method can effectively detect floating objects on river surfaces, with recall rate of 0.741 for artificial trash, [email protected] of 0.798, and [email protected]:0.95 of 0.704. This system achieves real-time surface floating objects monitoring in rivers and open channels, providing crucial information for the protection and management of water environments.