Vision based obstacle avoidance and recognition system

Automation techniques have been applied in almost every field in past few years. Automated Guided Vehicle (AGV) are most often used in industries and inventories for object management. Obstacle avoidance being a necessary requirement for navigation in any vehicle, still faces many challenges in the field of automation due to uncertain nature of the surrounding environment. This paper presents the implementation of an obstacle avoidance method on an AGV using stereoscopic vision by creating a disparity map and measuring the relative distances of the objects in the scene. Not only it avoids collision but it also classifies the object into one of the specified categories using supervised learning algorithm. The integration of image classification with stereovision can allow the AGV to understand which objects to pick up and which ones to leave behind.