Please use this identifier to cite or link to this item: https://dl.ucsc.cmb.ac.lk/jspui/handle/123456789/3926
Title: Identification of Lantana Camara Distribution Using Convolutional Neural Networks
Authors: Samarajeewa, T. M.
Issue Date: 2017
Abstract: Abstract Lantana camara is an exotic invasive plant that has been a major threat to the biodiversity of a number of countries around the world. This plant has introduced a number of hazards such as outpacing native trees and plants, changes in soil, devaluation of habitat, alteration of fire and more. This dissertation presents a novel method using Convolutional Neural Networks (CNNs) and image processing techniques to identify the distribution of Lantana camara plants with red, orange, or yellow colour flowers, in aerial images. The proposed method includes three stages for the detection of Lantana camara invaded places. The first step is the detection of possible flower patches in the aerial images. Second step is the recognition of Lantana camara flowers from the flower patches through classification of a convolutional Neural Network (CNN). The last step is marking Lantana camara flower presence in the original image. The resulting image is marked with Lantana camara flowers, which indicates the presence of Lantana camara plants in that image. Thresholding in L*a*b* colour space has been employed for the first step to segment possible flower patches from aerial image. For the second step, the BVLC (Berkeley Vision and Learning Center) distributed AlexNet has been used as the CNN architecture. The CNN has been trained to classify 967 flower species at an accuracy of 55.2%. The accuracy of recognizing Lantana camara flowers by the CNN is 94.6%. The accuracy of identifying all Lantana camara flowers in the original image is 40.71%. The proposed model was able to identify the presence of Lantana camara in aerial images successfully.
URI: http://hdl.handle.net/123456789/3926
Appears in Collections:SCS Individual/Group Project - Final Thesis (2017)

Files in This Item:
File Description SizeFormat 
THESIS.pdf4.43 MBAdobe PDFView/Open


Items in UCSC Digital Library are protected by copyright, with all rights reserved, unless otherwise indicated.