1. UKM Learning and Research Repository
2. Master Thesis / Tesis Master
3. Faculty of Engineering and Built Environment / Fakulti Kejuruteraan dan Alam Bina

Title:	Solid waste bin detection and classification using dynamic time warpingand gabor filter techniques
Authors:	Md. Shafiqul Islam (P59415)
Supervisor:	M. A. Hannan, Professor Dr.
Keywords:	Waste bin
Issue Date:	17-Apr-2014
Description:	The solid waste generation is increasing rapidly due to high rates of urbanization and industrialization and it has become a big challenge for municipal authorities. Numerous methods and systems have already been explored to overcome this challenge such as sensor embedded intelligent bin, RFID and GIS technologies, software-based routing and image processing techniques. The existing image processing systems for solid waste level classification cannot classify the solid waste bin level correctly without the bin position at the centre of the image. The systems consider the whole image area for feature extraction and consider the bin is at the centre position or near to the centre. However, capturing the bin image is a big challenge to keep the bin position at the centre of the image. As yet, there is no ideal system which can detect and classify the solid waste bin correctly. The main objective of this study is to develop an efficient image processing technique to solve the bin detection and classification problems. The system has four prime working stages: image pre-processing, bin detection, feature extraction and classification. The image preprocessing is based on removing the noise caused by uncontrolled external illuminations and weather conditions. Dynamic time warping (DTW) technique is used for detecting the bin and cropping the bin area and Gabor wavelet filter is developed for feature extraction of the solid waste bin. The waste bins have similar shapes, although the shapes cannot arrange up in X-axis because of different time sequences. The DTW was applied to find similarities within time domain series, the time axis of series have warped and averaging them to achieve an enhanced alignment. The constraints of the warping path, boundary conditions, monotonicity and step sizes, were influenced to crop the bin area. The waste images were enhanced by Gabor wavelet filter, which was a convolution process of filters coefficient matrix and the image. The Gabor filters parameters were specified considering the frequency levels and mask size for required features extraction. A Multi-Layer perceptron (MLP) classifier was used to classify the waste bin level and estimate the amount of waste inside the bin. The extracted image features were used to train the classifier for four different levels (low, medium, full and flow) classification. The area under the receiver operating characteristic (ROC) curves was used to statistically evaluate classification performance. The results of this developed system were compared to previous image processing based system with several similarity measures. The system demonstration using DTW with Gabor wavelet filter for feature extraction and an MLP classifier led to promising results with respect to the accuracy of waste level estimation (98.50%). This research has achieved its stated goal, which can be used to optimize the routing of solid waste collection system.,Master/Sarjana
Pages:	110
Publisher:	UKM, Bangi
Appears in Collections:	Faculty of Engineering and Built Environment / Fakulti Kejuruteraan dan Alam Bina

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.