The new SSD image detection for quadcopter platform control simulation

Abstract

Image processing and machine learning has fast progress in this era. Machine learning also has been applicate in our daily life. One of machine learning application is an object detection. This paper shows a state-of-the-art of object detection is applicated on control system. This paper also gives a deliberation by comparing another trained state-of-the- A rt (YOLOv3-tiny-416 and SSD with MobileNet V2) to detect single object i.e. person. SSD is one state-of-the-art of object detection which concept is to detect object in a single forward (one-shot) to feature extraction and classifier. Our work develops the feature extractor layer of SSD to fit with our purpose and show our modified SSD successfully control quadcopter movements feedback and has an average precision (AP) 97% to detect person. Quadcopter controls is completed by a simulation in gazebo simulator.