Research aimed at food safety, market access, market development and trade of plant industries

The Spectral Imaging (formerly known as SISM) project advances Australian capabilities in key emerging sensing technologies beyond the visible spectrum with a strong focus in national prosperity and well being. The project is one of the first of its kind worldwide combining a broad range of technologies to “see” beyond the visible spectrum and addressing the spectral image understanding problem making use of structural and statistical pattern recognition and computer vision techniques.

The project has two main areas

• Hyperspectral sensing

Hyperspectral imaging is an information-rich representation of the object under study in which each pixel or sample is comprised by a number of wavelength-indexed measurements. As a result, hyperspectral sensing is particularly well suited for non-intrusive material identification and recognition tasks with application in the areas of security, environment and biology.

• Thermal and near-IR

Thermal and near-infrared imaging is a key technology in night-vision and surveillance applications. Thermography can also be used to monitor blood flow, body temperature and perform early, non-intrusive diagnosis of some kinds of cancer.

This technology can, potentially, save farmers and quarantine officers hundreds of hours in laborious and time-consuming pest inspections. As a result, the project produces research relevant to Australia’s DPI future in a multidisciplinary environment with collaborators from a number of state-government departments and research institutions nationwide. We are focusing on Early detection of plant pathogens Smart trapping technologies for plant pests  These efforts have yielded promising preliminary results on the identification of a number of pests, such as some varieties of fruit fly and moth. Other promising results are those pertaining plant pathogens such as apple scab and wheat rust, amongst others.

What has the project achieved?

• The project produces internationally competitive research on methods aimed at
  

• Hyperspectral image understanding and spectroscopy techniques for the biosecurity sector
  
• Image database search and retrieval methods for spectral imaging systems
  
• Reliable material recognition and identification algorithms with a large margin or accuracy and very low false negative rates.

• It has produced more than 30 peer-reviewed publications in first-tier academic venues. It has also demonstrated the technologies developed at the project in a number of commercial and academic venues.
  
• The project has the capability to integrate hyperspectral imaging and sensing systems as turnkey, general purpose research vehicles.
• It provides education and training on the use of close range, ground-based spectroscopy systems.
  

Research team

• Jun Zhou
• Antonio Robles-Kelly
• Tim Raupach
• Ran Wei
• Cong Phuoc Huynh
• Pattaraporn Khuwuthyakorn
• Quang Anh Nguyen

Participants

• Australian National University
• Collaborative Research Centre for National Plant Biosecurity, Australia
  
• University of York, United Kingdom
• Zhejiang University, P.R. China.

Awards

The project was awarded the Best Paper Prize at the 6th IEEE Workshop on Object Tracking and Classification Beyond and in the Visible Spectrum (OTCBVS'09) for its work on hyperspectral image descriptors reported in the paper

• P. Khuwuthyakorn, A. Robles-Kelly and Jun Zhou, "An affine Invariant Hyperspectral Texture Descriptor Based Upon Heavy-tailed Distributions and Fourier Analysis", OTCBVS, 2009

Best Paper

• Z. Fu and A. Robles-Kelly and T. Caelli and R. Tan, “On Automatic Absorption Detection for Imaging Spectroscopy: A Comparative Study”, In IEEE Transactions on Geoscience and Remote Sensing, Vol. 45(11-2), 3827-3844, 2007.

Other Selected Publications (For a complete list, click here)

Journals

• §S. Prakash and A. Robles-Kelly, “A Semi-supervised Approach to Space Carving”, In Pattern Recognition, Article in press,2009.
• J. Zhou, L. Cheng and W. Bischof, "Spatial-Temporal Modeling of Interactive Image Interpretation", In Spatial Vision,  Article in press, 2009. 
• A. Robles-Kelly and E. R. Hancock, “Shape-from-shading using the Heat Equation”. In IEEE Transactions on Image Processing. Vol. 16(1), 7-21, 2007.
• §A Torsello, A. Robles-Kelly and E. R. Hancock, “Discovering Shape Classes using Tree Edit-distance and Pairwise Clustering”. In International Journal of Computer Vision. Vol. 72 (3), 259-285, 2007.  

Conferences

• §Z. Fu and A. Robles-Kelly, “An Instance Selection Approach to Multiple Instance Learning”, In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2009.
• J. Zhou and A. Robles-Kelly, "A Quasi-random Sampling Approach to Image Retrieval", In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2008. §
• J. Zhou, L. Cheng and W. Bischof, "Prediction and Change Detection In Sequential Data for Interactive Applications", In 23rd AAAI Conference on Artificial Intelligence, 2008. 
• C. P. Huynh and A. Robles-Kelly, "A NURBS-Based Spectral Reflectance Descriptor with Applications in Computer Vision and Pattern Recognition", In Proceedings of the IEEEConference on Computer Vision and Pattern Recognition, 2008.  

Patents

• C. P. Huynh and A. Robles-Kelly,"Compact Representation of a Reflectance Spectrum", Provisional Specification, 2009.

§