Novel algorithms for image classification and visualisation of multispectral datasets

用于图像分类和多光谱数据集可视化的新算法

• 批准号: 516103-2017
• 负责人: Braidy, Nadi
• 金额: $ 1.82万
• 依托单位: Université de Sherbrooke
• 依托单位国家: 加拿大
• 项目类别: Engage Grants Program
• 财政年份: 2017
• 资助国家: 加拿大
• 起止时间: 2017-01-01 至 2018-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=639630
• 关键词: Novel; algorithms; image; classification; visualisation

The recent technological improvements in microscopy now allows the collection of extensive information ofsamples at the nanoscale. However, in many cases, the amount of data collected surpasses the currentvisualization capacity making their interpretation and analysis impractical. While two and three physicaldimensions visualization is quite intuitive, the interpretation of datasets with a higher number of dimensionssuch as with chemical tomography, is less intuitive. Nevertheless, the information provided by extradimensions is essential when attempting to uncover similarities or correlations essential for materialsdevelopment (e.g. distinguish phases in a material, identify and map impurities).The research group of Pr. Nadi Braidy have been successful in developing a classification algorithm to simplifythe interpretation of multi-spectral datasets. However, the efficiency of the algorithm and the code is notadapted to large datasets (i.e. several giga/terabytes). The goal of this project is to develop and implementscientific and software improvements to the existing algorithm to improve its computational efficiency. Therevamped algorithm will be implemented into Object Research System (ORS) Inc. visualization software,Dragonfly.In the framework of this Engage, the collaboration between Pr. Braidy and ORS will provide ORS with a toolto handle multi-modal datasets in its software while providing support in improving its efficiency.

最近的显微技术的进步，现在允许收集样品在纳米尺度上的广泛信息。然而，在许多情况下，收集的数据量超过了目前的可视化能力，使其解释和分析不切实际。虽然两个和三个物理维度的可视化是非常直观的，但具有更高维度的数据集的解释，如化学断层扫描，则不那么直观。然而，当试图揭示材料开发所必需的相似性或相关性时，额外维度提供的信息是必不可少的（例如区分材料中的相，识别和映射杂质）。Nadi Braidy教授的研究小组成功地开发了一种分类算法，以简化多光谱数据集的解释。然而，算法和代码的效率并不适用于大型数据集（即几千兆/太字节）。该项目的目标是开发和实施对现有算法的科学和软件改进，以提高其计算效率。改进后的算法将在对象研究系统（ORS）公司中实现。可视化软件Dragonfly。在这个Engage的框架内，Braidy教授和ORS之间的合作将为ORS提供一个在其软件中处理多模态数据集的工具，同时为提高其效率提供支持。