We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
Research on Underwater Debris Detection Technology Based on YOLO Algorithm
Summary
Researchers built an AI-powered tool that uses smart cameras (the YOLO algorithm) to automatically spot and sort ocean trash into categories like plastic waste versus marine life, then display the results on an easy-to-use screen. This kind of technology matters because ocean plastic doesn't just harm sea creatures—it breaks down into microplastics that can end up in seafood and drinking water, so faster, more accurate trash detection could help cleanup efforts target pollution before it reaches our plates.
Marine debris not only severely pollutes the health and habitat of marine life but also indirectly threatens human life. Deep learning-based target detection algorithms have achieved significant results in underwater debris detection, with the YOLO algorithm being widely used by researchers due to its fast detection speed and high accuracy. However, the underwater environment is complex and variable, and underwater debris detection faces challenges such as low detection accuracy, false negatives and missed detections of small targets, and a large number of parameters. This paper mainly focuses on three aspects: experimental environment configuration, dataset construction, and platform design. The experimental environment setup primarily included Anaconda environment debugging and Python programming language installation. Dataset construction involved initial collection of underwater debris images, followed by image labeling and classification. The dataset was divided into two main categories: biological (bio) and plastic (plastic) debris. Finally, the platform was designed using PyQt5. The platform interface is divided into three areas: settings, detection, and data display. The settings area allows users to select images and videos for detection. The detection area displays the detected underwater debris in the currently detected images and videos. The data display shows the time taken to detect the images, the number of detected targets, their type, and the confidence level.