A Hybrid Color Plane Approach Towards Color Based Object Detection and Modeling of a Real-Time Gesture Based Intelligent Virtual Aid Using Artificial Neural Network

Augmented and Virtual Reality has become a highly researched subfield in the Computer Vision domain. Among various researched topics, object detection and tracking is a major field of interest, with various algorithms being developed for it. Various algorithms exist that employs different approaches to solving the problem of object detection. The approach that we take is a mixture of investigating and evaluating the previous approaches and coming up with a novel algorithm that is able to detect and track objects in real-time that is not tightly bound to specific colours. In order to implement a dynamic algorithm, we look into various colour planes and then take in multiple dimensions from a given colour plane and use a combination of the results obtained from the individual dimensions to get a result that is capable of detecting various range of colours. After the objects are satisfactorily detected, we proceed to track them and implement gestures that allow the users to elect a single colour to be tracked without any direct interactions, using a novel algorithm that we developed over the course of the research. After the election, the algorithm is expanded further to be able to integrate the ability to detect various gestures and patterns. The system is also capable of basic intelligence and is able to modify its phases as necessary based on the user's interaction with the system. As such most of the system is user action dependent and free of any physical interaction. All the while, the algorithm is designed to be able to do all the processing in real-time.