Use this URL to cite or link to this record in EThOS: https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.784572
Title: On-line measurement of broken rice percentage from image analysis of length and shape
Author: Mallah, Charles D.
Awarding Body: Kingston University
Current Institution: Kingston University
Date of Award: 2013
Availability of Full Text:
Access from EThOS:
Access from Institution:
Abstract:
This thesis investigates the design of an on-line signal processing system to measure the percentage of a batch of rice grains that are broken. The objectives of the investigation were: to provide information contributing to the hardware design; to devise one or more approaches for each processing stage; to evaluate and validate these approaches, where necessary recommending the most suitable; and to verify the overall accuracy and robustness of the proposed system. The fundamental processing stages were proposed and, for each stage, the possible approaches were investigated. A configurable experimental apparatus was assembled to enable an iterative development of algorithm and hardware design: an important aspect of the methodology was to locate an acceptable operating point in the space of possible configurations. The processing stages comprised geometric camera calibration; segmentation of the grains from the background; detection of single grains; measurement of grain length and characterisation of shape; classification; and finally corrections for bias and conversion to the required unit of measurement, percentage by mass. For each stage, an analysis of error and uncertainty was undertaken. This provides an indication of the expected accuracy and bias of the proposed system. There were several elements of novel work completed in the course of this investigation. Some novel components were introduced into the calibration procedure; a tiled approach was used in the segmentation stage, to accommodate diverse illumination across the field of view. For the shape analysis, an innovative method is employed to estimate the posterior class density, and the analysis of expected bias includes a treatment of finite aperture but also how the probability of observing a single grain is conditional upon its length. The results obtained from the experimental apparatus and prototype device indicate that it is feasible to obtain on-line measurements that are well within the required tolerance. This would enable such a device to be deployed in a number of important industrial applications.
Supervisor: Orwell, James ; Hamid, Gabriel Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.784572  DOI: Not available
Keywords: Computer science and informatics
Share: