Aerial Monitoring On Mango Using Machine Vision Techniques

Oh, Kok Ken (2016) Aerial Monitoring On Mango Using Machine Vision Techniques. Project Report. UTeM, Melaka, Malaysia. (Submitted)

Text (24 Pages) Aerial Monitoring On Mango Using Machine Vision Techniques.pdf - Submitted Version Download (1MB)

Abstract

The entire project deals with development of shape identification algorithm and image training classification with Histogram of Oriented Gradient. The focus would be the process of detection and summing up the total number of mango on its tree with a drone and display it in MATLAB. The hypothesis made is the system could at least achieve the detection rate of 60%. The conventional method in harvesting mango has its limitation which leads to the degradation of mango's quality. Besides, the rate of production and the structure of the tree will be affected too. Previous researches have proven the significant role of machine vision to be embedded in agricultural technology. Nonetheless, the usage of a drone with an algorithm of image processing could be employed for a better and more precise mango's farming. The obvious benefit of utilizing drone is its capability to hover around the upper part of tree which is normally not reachable by human. The device could hover around mango trees and count the detected mango in a short operating time. It differentiates the mango from its background based on the images captured. The experiments were done in indoor as well as outdoor. Indoor experiment is less prone to noise meanwhile; outdoor experiment is sensible to noise. Eccentricity and form factor is the determining criteria for shape and size judgment. Machine vision's classification using training classifier would be another method under research. Analysis of obtainable result would be tabulated of based on confusion matrix. The detection rate in all experiments should exceed a detection rate of approximately eighty percent. In short, it provides a quick review for the mango grower, agricultural developer and investor.

Actions (login required)

View Item