Transparent Plastic Bottle Detection and Depth Decision Method using YOLOv8 in Recyclable Waste Segregation Systems

Plastic poses a significant challenge as it takes a long time to decompose and can pollute the natural environment if incinerated or disposed of in landfills. Recent efforts have focused on leveraging robots and artificial intelligence at recycling sorting sites to tackle this issue. This paper introduces a recyclable waste segregation system and proposes a method for detecting and determining the depth of transparent plastic bottles using YOLOv8 within such systems. The system initially obtains depth information through RGB images captured by stereo cameras. Subsequently, the system employs YOLOv8 to detect plastic bottles and extract their instance segments from the captured images along with depth data. The system then calculates the center position along the z-axis for picking up the detected bottles. Furthermore, this paper addresses depth error correction, a crucial aspect considering the difficulty in accurately measuring the depth of transparent objects.