|
|
PhD Opportunities in Computer Vision or Pattern Recognition
| Posted on 2007-06-19
|
NICTA and the CRC on National Plant Biosecurity (CRCNPB) are combining their efforts to offer PhD opportunities to first class applicants which are motivated to pursue research in the areas of computer vision or pattern recognition at the Spectral Imaging and Source Mapping (SISM) project.
At present, we are focusing on the following research topics:
Hyperspectral vision methods for automatic recognition and classification
Recent developments in computer vision and graphics have opened-up the possibility of recovering spectral traces of plants and insects for purposes of classification, identification and recognition. This project hinges in the idea that hyperspectal imaging technologies based upon hyperspectral and UV can be used to recover discriminant information and design robust classifiers. The output of these classifiers can then be used as input to statistical pattern recognition and machine learning algorithms.
Following this rationale, the work will focus on the development of algorithms and methods which can be used to perform identification and classification of plant pests in laboratory and field settings. From the scholarly point of view, the idea is to explore the use of kernel methods and optimisation techniques that can be utilised for the purposes of discriminant learning and classification. These are novel technologies that can be used to enhance early warning trapping networks. These methods are quite general in nature and can be easily applied to other pattern recognition settings relevant to the current joint effort between NICTA and the CRCNPB regarding the potential of spectral imaging and digital shape/pattern recognition technologies for automatic detection of plant pests.
3D data acquisition via confocal vision
Confocal microscopy allows for the collection of optical sections of the specimens under study. These optical sections, i.e. images with a very small focal depth, can be used to build a 3D image of the sample so as to recover the 3D structure of the specimen. In this project, we aim at developing robust techniques for the processing of 3D images acquired making use of confocal microscopy. These methods can be used to recover CAD models from archived insects and will permit the integration of 3D models into widely accessible databases for purposes of reference, training and collection management. It will also enhance Australian capabilities in the use, research and development of confocal microscopy.
The work will focus on the development of algorithms and methods which can be used to recover CAD models making use of confocal microscopy. Confocal imagery processing usually involves a denoising step, a segmentation operation and a volumetric shape analysis stage. This is aimed at producing a 3D representation of the specimen or an image sequence which can emulate rotations and other spatial transformations. The idea is to develop techniques that can be utilised to design and deploy tools for the semi-automatic, efficient processing of confocal imagery. These are novel methods that are quite general in nature and can be used to enhance databases, better document collections and other tasks.
The SISM project pursues research in key emerging imaging and sensing technologies (hyperspectral cameras, infrared cameras, etc.) that can see beyond what is visible to the human eye. The project has a strong focus in national prosperity and well being and employs technologies which are non-intrusive and non-destructive. These sensing techniques have a wide variety of applications in biosecurity, ecology, human performance and health, food quality assurance and surveillance.
For further information on the projects above, please contact Dr. Antonio Robles-Kelly.
|
|
