Characterisation of rice grains using digital imaging techniques

Visual characteristics of rice grains can be quantified through image acquisition and processing. In this research image capture hardware and dedicated image processing algorithms are developed as part of a system for rice grain characterisation. A series of experimental work was undertaken to test the effectiveness of the image processing algorithms. The imaging technique is non-intrusive and non-destructive of sample grains, and can be implemented at a low cost compared to other more established techniques.

In this thesis a review of techniques for the characterisation of rice grains is presented, with the main focus upon digital imaging approaches to understand the role it can play in isolation and in conjunction with other techniques. Following the review a physical rice grain image capture setup is designed, implemented and tested extensively with a range of cameras. A number of established image processing techniques are used with the custom built image capture setup to complete a novel system for the characterisation of rice grains. Extraction of features from the processed images is undertaken in order to test the features as being suitable to return descriptive characteristics of individual rice grains. In the course of these activity novel features, methodologies and feature analysis are created and implemented with wider potential applications beyond rice imaging. Results of these tests are given within this thesis. These tests take several forms in multiple sets of rice grain images, which were created, processed, and subject to different kinds of analysis to find different features.

Findings show that digital imaging can return a range of valid rice characteristics which are connectable to fraud and grading issues, identification of rice types, and assisting the physical measurement of rice grain properties. The performance of the system is discussed and suggestions given for the future development of the methodology.