Stereo-Vision Based System For Object Detection And Recognition

Abstract

This paper proposes a method for detecting and recognizing the object using Stereo Vision, Scale-Invariant Feature Transform (SIFT) and Fast library for approximate Nearest Neighbors (FLANN) concept with its implementation on an embedded system. Using stereo vision on the microprocessor Raspberry Pi, the implemented system takes the two images produced as input, calculates the disparity map which provides the relative depth information. Using this map and the Scale-Invariant Feature Transform (SIFT), features are obtained and matched with a database having large collection of images. This implementation uses Fast Library for Approximate Nearest Neighbors (FLANN), which unlike the Brute-Force matching algorithm can support large databases. This system gives a voice output when the object is recognized by text to speech conversion.