Efficient counting and localizing objects of interest for agricultural automation

Abstract

This project has introduced an agricultural automation framework for counting and localizing the objects of interest by using a computer, a mobile phone and NXT Lego Mindstorm Robot. Its main purpose is to speed up the yield counting task in agricultural field especially when the yield is huge. Mobile phone is used for object recognition while computer is used to process, send and receive data. In this project, eggs are the objects of interest. The robot used is capable to navigate autonomously in the known environment using A- algorithm and knowing its location using Monte Carlo Localization. When robot's location and sensor range reading are retrieved, object's location can be calculated by simply implementing Trigonometry Identity concept. To count the number of eggs, other object such as stone will be excluded even if it exists in the system. They can be efficiently differentiated through object's contour size analysis and colour detection.