An Object Recognition and Volume Calculation Method Based on Yolov3 and Depth Vision

Abstract

In recent years, the logistics industry is in its full swing. The demand for logistics is increasing throughout the country, and this rising trend will continue during the following decades. With the substantial increase of workloads, traditional manual ways are increasingly unable to meet the new needs of the logistics industry, therefore new production and transportation modes are becoming imperative. Considering this, a new method was proposed to help detect, recognize and calculate the volume of rectangular objects automatically. In this method, a YOLO-based model was firstly applied to detect the object. Secondly, an RGB-D camera was used to obtain the depth information. Finally, after the length, width and height values were estimated from a single surface, the volume of the object can be calculated. Experimental results showed that the method is simpler and more accurate as compared with traditional three-dimensional reconstruction-based methods.