图像表示是模式识别研究中关键问题之一.奇异值分解(SVD, Singular Value Decomposition)是一种有效的图像表示方法,近年来已被广泛应用到计算机视觉、信号处理、模式识别和图像处理等领域.但是,奇异值分解在处理高维数据时的效率瓶颈以及无法同时考虑样本类别信息和固有几何结构信息的缺陷制约了奇异值分解的应用范围和应用研究的发展. 本项目针对奇异值分解存在的局限性,通过系统地研究奇异值分解在特征提取中的应用,拓展和推动奇异值分解的应用,具有重要的理论研究意义和实用价值.同时,将研究成果用于解决混纺纤维的纤维识别问题,对纺织品截面纤维进行准确的图像表示.其研究成果将为解决纺织品检验领域纤维自动识别与分析这一世界性难题带来创新性的突破.%Singular Value Decomposition (SVD)is an efficient method of image representation, which has been widely employed in computer vision and pattern recognition fields.However, due to the efficiency bottleneck when SVD dealing with high dimensional data, and the defects that sample category information and inherent geometric structure information can not be taken into consideration simultaneously, the application scope and application research of SVD are restricted.Aiming at the limitation existed in SVD, the task could systematically study the application of SVD in feature extraction, expand and promote the application of SVD, so it has important research meaning and practical value.Based on the above, the paper proposes a research project that applies the constrained SVD technology to automatic identifications of blended fibers.The accuracy, objectivity and testing efficiency will be improved by the inspection work presented in the paper, and a greater social and economic benefit can be brought to our country's textile exports, as well as to this worldwide technical difficulties in the commodity inspection field.
展开▼
