SGER: New Approaches for Design of Nonlinear Discriminants

SGER：非线性判别式设计的新方法

• 批准号: 9527477
• 负责人: Nagabhushana Prabhu
• 金额: $ 3万
• 依托单位: Purdue Research Foundation
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 1995
• 资助国家: 美国
• 起止时间: 1995-09-01 至 1997-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=9527477&HistoricalAwards=false
• 关键词: SGER; New; Approaches; Design; Nonlinear

9527477 Prabhu This research is funded as an SGER (Small Grant for Exploratory Research). The objective of the project is to enhance the design, development, and implementation of a computer-aided nonlinear discriminant for pattern recognition. The approach will employ techniques from algebraic geometry and dynamic systems. Discriminant design uses a database of feature vectors and partitions the feature space into disjoint regions corresponding to different classes of patterns. Based on the methodology, an algorithm will be developed and implemented. The performance of the algorithm will be benchmarked through a series of tests problems in weather forecasting and production systems for which quick pattern recognition and control are required. To rapidly recognize and act on objects, especially in situations involving automatic assembly and part sortation, are essential qualities for the design and implementation of effective automated production systems. The technique, apart from its nonlinear discriminant nature, is expected to yield some improvement in both speed and accuracy in pattern recognition in general. The development of fast algorithms capable of processing large amounts of data in very short time and with high accuracy is essential for large scale adoption of automatic pattern recognition techniques in a host of applications. The outcome of this research has the potential to advance the state-of-the-art in this critical field.

小行星9527477 这项研究是作为SGER（探索性研究小额赠款）资助的。该项目的目标是加强模式识别的计算机辅助非线性判别式的设计、开发和实现。该方法将采用代数几何和动态系统的技术。判别式设计使用特征向量的数据库，并将特征空间划分为对应于不同类型模式的不相交区域。基于该方法，将开发和实施一种算法。该算法的性能将通过一系列的测试问题，在天气预报和生产系统，快速模式识别和控制的要求。 快速识别并对对象采取行动，特别是在涉及自动装配和零件分拣的情况下，是设计和实施有效的自动化生产系统的基本素质。该技术，除了其非线性判别性质，预计将产生一些改进的速度和准确性在模式识别一般。开发能够在很短的时间内处理大量数据并具有高精度的快速算法对于在许多应用中大规模采用自动模式识别技术是必不可少的。这项研究的结果有可能推动这一关键领域的最新技术。