MOBILE ROBOT NAVIGATION IN OUTDOOR ENVIRONMENT USING MONOCULAR VISION

Abstract

Increasing number of road car accidents and high causality rate in these accidents demand an efficient method to manage road traffic and restrict people speed limits. Numerous traffic police officers are stationed throughout the country on every major avenue but many times they seem to be unable to control traffic flow and implement speed limits. Not only does this affect human lives but also have adverse effects on economy. A system needs to be developed that follows all the traffic laws, is reliable and also easy to use so that not only human lives can be saved but the huge economic losses associated with traffic accidents could also be reduced. Autonomous driving systems is the perfect solution for this problem as this system would follow traffic laws, greatly reduce number of accidents, and improve fuel efficiency by eliminating rash and distracted driving. This system would also be equally beneficial for children, elderly and disabled as they would not require help of other people to drop them at their destinations. Different International agencies are working on developing Autonomous driving systems like DARPA, Google, General Motors, BMW and many other universities from around the world. Unfortunately, in Pakistan, no government agency or university is working on autonomous driving cars. The main reason for this is the fact that equipment used for autonomous driving (sensors and processors) are very expensive, there is very little locally owned mapping data (such as Google maps and Google Street View etc.) and it will take years of research to make a car that is able to drive through public roads without any problems. Autonomous driving cars will not only save our assets, our people, but will also help in development of our country. This MS Thesis is concerned with the problem of Mobile Robot Navigation in outdoor structured environment using Monocular Vision. The developed algorithm will be generic enough to implement on any intelligent driving system. Navigation of mobile robot in its own environment is a challenging task. All programming and image processing was done in MATLAB using Image Processing Toolbox. This report presents a novel approach for navigation of mobile robot on the road using monocular vision only. The developed algorithm can be implemented on any vehicle or robot moving on road. A method for road segmentation and obstacle detection is presented. A state of the art method for distance estimation is also described. After that a method for obstacle avoidance is presented using previously gained information from monocular image.