Embedded Computer Vision System Applied to a Four-Legged Line Follower Robot

TL;DR

This paper demonstrates an embedded computer vision system that enables a four-legged robot to autonomously follow a line using open source tools, integrating perception and action in a simple, cost-effective manner.

Contribution

It introduces a practical implementation of line-following robotics using open source hardware and software, emphasizing simplicity, replicability, and educational value.

Findings

Robot successfully follows a line using color recognition techniques.

Multiple image processing approaches were tested for accuracy.

Open source tools proved effective for embedded robotics applications.

Abstract

Robotics can be defined as the connection of perception to action. Taking this further, this project aims to drive a robot using an automated computer vision embedded system, connecting the robot's vision to its behavior. In order to implement a color recognition system on the robot, open source tools are chosen, such as Processing language, Android system, Arduino platform and Pixy camera. The constraints are clear: simplicity, replicability and financial viability. In order to integrate Robotics, Computer Vision and Image Processing, the robot is applied on a typical mobile robot's issue: line following. The problem of distinguishing the path from the background is analyzed through different approaches: the popular Otsu's Method, thresholding based on color combinations through experimentation and color tracking via hue and saturation. Decision making of where to move next is based on the line center of the path and is fully automated. Using a four-legged robot as platform and a camera as its only sensor, the robot is capable of successfully follow a line. From capturing the image to moving the robot, it's evident how integrative Robotics can be. The issue of this paper alone involves knowledge of Mechanical Engineering, Electronics, Control Systems and Programming. Everything related to this work was documented and made available on an open source online page, so it can be useful in learning and experimenting with robotics.