Automated River Cleaning Robot for Plastic Waste Segragation Using AI

Abstract

River plastic pollution threatens aquatic habitats, animals, and human health.  This paper describes an automated floating robot concept that detects and collects plastic debris from water surfaces using computer vision and machine learning.  Unlike standard cleaning systems that gather all forms of debris, the proposed robot use a TensorFlow Lite object recognition model running on a Raspberry Pi to target just plastic materials.   A Pi Camera records footage in real time onboard for identification.  When detection is more than 80% certain, a servo-powered loader moves the object to an interior storage area.  For verification, it takes a snapshot of the recognized item and logs time and GPS data using the NEO-6M module.  Robots are designed to be energy efficient and may be powered by solar energy.   Robots conduct exact recognition under a variety of situations.