Mahmood Golzarian

Phone: +61 8 8302 3614
Fax: +61 8 8302 3380
Email: mahmood.golzarian@postgrads.unisa.edu.au
Location: J1-12, Mawson Lakes Campus
Status: Full Time
Supervisors: Assoc. Prof. John Fielke, Dr. Jack Desbiolles
Expected Completion Date: January 2009

Project title:

Development of a computer vision based system to monitor crop establishment

Brief description of project:

Monitoring crop establishment over a period of days and weeks, which embraces detecting crop and weeds and counting seedling and weed numbers, provides a means to determine how successful the crop has established. The conventional method, using a quadrat, is a physically demanding and time-consuming task. This research suggests an efficient and innovative crop establishment monitoring system using machine vision. A novel new algorithm based on neural networks is to be developed to enhance the accuracy of the system. Ultimately, the developed system will be mounted on a moving vehicle to verify the result of the research in real time in situ.

Project details

Aim

to develop a novel machine vision system to monitor crop establishments

Objectives

to analyses the colour indices used for segmenting plants from the background and the best colour index that leaves the least amount of noise
to analytically compare the colour indices in order to highlight superior colour indices based on image noise levels and to maximize the segmentation quality
to optimise the segmentation and remove all noise regions affecting feature extraction, and to produce a clean (no noise) image with the least affected plant integrity under crop residue conditions
to investigate a quality way to connect the plant segments resulted after the segmentation process
to locate individual plants in a field in real time

Achievements

A comparison study was conducted to investigate the optimum colour index (segmentation process)
A high quality segmented images were produced using the optimum colour index, following by a novel noise removal complementary process
A novel method for optimizing the segmentation process for images with crop residue background was proposed and tested on 500 images
The isolated plant segments were connected to form an integrated plant shape in images

Areas of interest

Image Processing and neural networks applications in machine vision systems
Robotics and machine vision systems in Agriculture
Precision Agriculture

Publications

Golzarian, M., Fielke, J, Desbiolles, J and Lee, M.K. (2007) A novel approach to improve the plant-background segmentation in computer vision systems, The Society for Engineering in Agriculture Conference, September 2007, Adelaide, Australia.
Golzarian, M. (2007) Evaluation and review of several colour indices used for segmenting plant from non-plant regions in colour images, proceeding of the 12th International CSI Computer Conference, Tehran, Iran.
Golzarian, M., Desbiolles J. and Lee M. K. (2007) A novel colour-index evaluation method for plant segmentation from a soil background, the 6th European Conference on Precision Agriculture, 3-6 June, Skiathos, Greece.
Golzarian, M.R., Lee, S.H. and Desbiolles J. (2006) Colour segmentation techniques for plant identification, E-Proceeding of the 2006 International Manufacturing Leaders Forum, Taiwan..

News Story: Separating weeds from the wheat

top^