Depth Extraction and Object Positioning System for Recyclable Waste Classification using Delta Robots

In daily life, as the use of recyclable products such as PET, PE, PE, PP, glass, etc. increases, efforts to reuse them are also increasing. It is important to accurately classify and select the recyclable waste by material type to increase the reuse rate. Computer vision and delta robot technology are being used in parallel to automatically classify recyclable waste, but the problem is that the delta robot cannot pick the waste accurately because it does not recognize the location of transparent plastic and glass objects. In this paper, we propose a depth extraction and object positioning system to increase the picking efficiency of delta robots by accurately identifying the depth and location of recyclable waste recognized through computer vision. Stereovision technology is used to extract the depth of transparent waste in the image, and to recognize the waste and use the centroid of the waste as the picking point. The depth and coordinate value of this centroid are measured and transmitted to the delta robot. The performance evaluation results showed an error within 4 centimeters even though it was a transparent waste.