SBIR Phase I: Filtering Objectionable Imagery

SBIR 第一阶段：过滤令人反感的图像

• 批准号: 0340384
• 负责人: Stuart Blundell
• 金额: $ 10万
• 依托单位: VISUAL LEARNING SYSTEMS INC
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2004
• 资助国家: 美国
• 起止时间: 2004-01-01 至 2004-06-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0340384&HistoricalAwards=false
• 关键词: SBIR; Phase; Filtering; Objectionable; Imagery

This Small Business Innovation Research Phase I project describes a new and innovative approach for filtering objectionable image content distributed on the Internet. As Internet usage in the workplace has increased, companies have had to deal with issues such as loss of productivity, unnecessary bandwidth usage and legal liability due to employees viewing objectionable image content. This project provides an automated, intelligent, object recognition solution that understands objectionable image content based on attributes beyond, for example, skin color. The research leverages the award-winning machine learning technology for object-recognition in imagery. The methodology is widely recognized within the industry as the first instance of an adaptive software agent technology capable of object recognition through both spectral and spatial processing of image attributes. This project will leverage this technology to develop a visual content filter. The system will not only be able to filter objectionable imagery, it will also provide solutions for filtering any image content, or inversely, providing a user-specific profile for finding imagery of a user's choice. As Internet usage in the workplace continues to increase, the research has strong benefits for government and private industry. The anticipated results are an adaptable commercial system that will continually improve the filtering of objectionable imagery from the database.

这个小企业创新研究第一阶段项目描述了一个新的和创新的方法来过滤不良的图像内容在互联网上分发。随着工作场所中互联网使用的增加，公司不得不处理诸如由于员工观看令人反感的图像内容而导致的生产力损失、不必要的带宽使用和法律的责任等问题。该项目提供了一种自动化、智能化的对象识别解决方案，可以根据肤色等属性来理解令人反感的图像内容。该研究利用屡获殊荣的机器学习技术进行图像中的对象识别。该方法被广泛认为是业内第一个实例的自适应软件代理技术能够通过光谱和空间处理的图像属性的对象识别。这个项目将利用这项技术来开发一个可视内容过滤器。 该系统将不仅能够过滤令人反感的图像，它还将提供用于过滤任何图像内容的解决方案，或者相反地，提供用于查找用户选择的图像的用户特定简档。 随着互联网在工作场所的使用不断增加，这项研究对政府和私营企业都有很大的好处。 预期的结果是建立一个适应性强的商业系统，不断改进从数据库中过滤不良图像的工作。