PATH TRACKING WITH COMPUTER VISION USING NEURAL NETWORK

Abstract

Accurate localization and mapping is a problem of great interest and significance in mobile robotics. In most of the cases the environment is unstructured and is unknown beforehand. Building an autonomous robot in the real time environment is quite challenging and requires modeling and investigation. This work is done keeping in mind to track the path using the computer vision approach Attempt is made to make the robot completely autonomous and to control it from remote locations. A robotic car has been constructed and image processing operations like wiener filtering, thresholding and Hough transforms have been applied. Wireless communication is implemented to accomplish the above desired task in real time. A novel approach to steer the robot using stepper motor is designed and implemented by transmitting the control commands wirelessly. This thesis proposes a robust solution to path tracking by a robot in real time. The approach used is based on the computer vision in which the images are used as the source of the input data provided to the algorithm for the tracking. Edge detection techniques are implemented and other feature extraction routines. Image is feature extracted for the information present. The approach of neural networks is also taken into consideration where the input information is trained for the output. The further use of wireless communication accompanied by RS. 232 serial communication makes it suitable for remote applications. There has been use of PIC microcontroller 16F877A for the wireless reception at the robot where the commands received are output for the tracking. Here vision feedback approach is used which has emerged as the demanded and useful approach in modern scenario. The robot designed is _ an inexpensive solution to the remote controlled, autonomous and real time robotic applications. The work finds its applications in various areas like exploration of outer space, mines and oceans, nuclear research etc. where human presence is impractical, unsafe or simply to. avoid manual labor. Operations in such impenetrable regions can be performed by installing such type of robotic device there, while the communication can be established using a satellite channel The wireless technology discussed here will definitely give an edge to any work-done in the field of robotics by giving it the finesse of movement free of wired constraints.