CISE Research Instrumentation: Specialized Equipment for Vision and Image Processing

CISE 研究仪器：视觉和图像处理专用设备

• 批准号: 9422288
• 负责人: Niels da Vitoria Lobo
• 金额: $ 3.92万
• 依托单位: The University of Central Florida Board of Trustees
• 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: 1995-05-01 至 1997-04-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=9422288&HistoricalAwards=false
• 关键词: CISE; Research; Instrumentation; Specialized; Equipment

9422288 Lobo This award is to purchase a real-time vision system to assist research in three projects: 1) Classify Human Age from Faces; 2) Visual Gesture Recognition; and 3) Multi-frame Order Statistics in Image Processing. In the first project, the research addresses the limited task of age classification of a facial image into a baby, young adult, and senior adult. It addresses the difficult issues of age classification of faces that are engaged in activity, i.e., not simply mugshots. The second project investigates performing gesture recognition based on motion information. The current library of gestures contains seven gestures: Left, Right, Up, Down, Rotate, Grab, and Stop. It is proposed to extend the method to 3-D gesture recognition. A 3-D hand model using Generalized Cylinders, will be employed and motion parameters will be estimated to compute 3-D hand trajectories. Gesture recognition becomes more realistic with a 3-D, rather than a 2-D, approach. The third project deals with the real-time performance of multi-frame order statistics involving median filtering. It is proposed to extend median and order statistic based filtering to 3-D by using spatio-temporal filtering (multiple successive images). It is also proposed to extend methods for suppression of narrow-band (periodic) interference to 2-D (for images).

小行星9422288 该奖项是购买一个实时视觉系统，以协助三个项目的研究：1）从面部分类人类年龄; 2）视觉手势识别;和3）图像处理中的多帧顺序统计。 在第一个项目中，该研究解决了将面部图像年龄分类为婴儿，年轻人和老年人的有限任务。 它解决了从事活动的面孔的年龄分类的困难问题，即，不仅仅是面部照片 第二个项目研究基于运动信息执行手势识别。 当前的手势库包含七个手势：向左、向右、向上、向下、旋转、抓取和停止。 建议将该方法扩展到3-D手势识别。 一个3-D的手模型，使用广义的手，将采用和运动参数估计计算3-D的手轨迹。 手势识别在3-D而不是2-D方法中变得更加真实。 第三个项目涉及多帧顺序统计量的实时性能，包括中值滤波。 提出了通过使用时空滤波（多个连续图像）将基于中值和顺序统计的滤波扩展到3-D。 还建议将抑制窄带（周期性）干扰的方法扩展到2-D（图像）。