Abstract
This paper presents the findings of a comprehensive literature and technology review which is aimed at identifying the current state of the art and future needs for object detection in automated systems. The work spans several application areas but the common denominator in each case is the technology related to the use of live images to improve the situational awareness of the automated system. This in turn should make the system capable of making decisions relative to its functional/mission requirements without the need for human intervention. Although it is acknowledged that advanced civil and military/space applications already encapsulate the technology or high-cost hardware and functionality required to deliver state of the art image recognition capability, this review will focus on low weight/size/cost solutions. The aim is to create a body of knowledge capable of informing a programme of work related to day-today consumer or industrially based applications. The overall programme of work aims to combine the size and cost advantages of single board computers with advanced methods of machine learning. The key technical challenge lies in how we can address the existing problem of image resolution impacting the accuracy in image recognition and its broader application for the control of automated systems. This review will focus on the current state of the art in object detection based on 2D image collection. The review will cover previous and ongoing work in the area covering academic research and industrial applications.