A Technical Review on UAV for Object Detection

Abstract

According to a Forbes report published in 2018, the US drone industry has witnessed a dramatic commercial growth from merely $40 million in 2012 to a billion in 2017, as documented in the study conducted by McKinsey and Company in December 2017. McKinsey estimates that by 2026 the commercial drones will annually impact $31 billion to $46 billion on the country’s gross domestic product. Drones have been widely adopted for data capturing that can be used in defense, agriculture, emergency response, disaster management, conservation of endangered species, healthcare, etc. For most of the applications of UAVs, Real-time object detection is exceptionally crucial. In the last few years, considering the growth of the drone industry and the interest of around 300 companies making substantial investments of time and resources in drones, many technologies have emerged for advancements in the field of UAV, focusing on object detection and recognition for UAV. This paper summarizes a number of object detection techniques proposed to date by researchers. It reports the characteristics and requirements of UAVs from an object detection viewpoint. The objective of our research is to understand different architectures that are capable of detecting objects from aerial images. The main goal of this survey is to create an insight into an architecture that is accurate, fast, robust, and utilizes low computation power.KeywordsConvolutional neural network (CNN)Unmanned aerial vehicle (UAV)Object detection and recognition