Detecting Litter in Street Sweepers Using Deep Learning

Abstract

Street sweeping vehicles are essential equipment in our daily lives designed to clean streets and roads. With numerous mechanical components, they play a significant role in collecting all types of waste and contributing to environmental cleanliness. These vehicles typically consist of rotating brushes, collecting belts, and components involving water or air currents. Among these parts, brushes and vacuums are the most energy-consuming elements in street sweepers. Moreover, they are often operated in full power mode due to semi-automatic control systems, leaving the remaining control to the driver. However, this practice results in energy wastage and noise pollution. The aim of this study is to adjust vacuum suction in street sweepers according to the size of waste using image processing and deep learning techniques, thus achieving energy conservation. In this research, the YOLOv7 model and OpenCV are employed to train artificial intelligence for waste detection in street sweepers and accordingly regulate vacuum suction.