复杂场景中基于分数阶微积分的局部形状匹配方法研究

项目名称： 复杂场景中基于分数阶微积分的局部形状匹配方法研究

项目编号： No.61462065

项目类型： 地区科学基金项目

立项/批准年度： 2015

项目学科： 其他

项目作者： 张桂梅

作者单位： 南昌航空大学

项目金额： 46万元

中文摘要： 形状匹配是计算机视觉的核心问题之一，但由于投影变换带来的形状差异，物体的相互遮挡、运动物体的柔性变化及分割错误等造成目标轮廓不封闭等都是形状匹配面临的主要困难，而提取局部特征，进行局部形状匹配是解决上述问题的关键。本项目拟通过理论分析和实验分析，研究复杂场景中局部形状匹配的理论和方法。重点研究基于分数阶微积分的图像边缘检测方法；研究仿射或透视变换下局部遮挡目标、轮廓为不封闭开曲线目标、柔性变化目标的形状匹配方法。本项目拟提出的理论和方法对于自动导航、自动检测、医学诊断、环境监测和深空探测等都能提供比较直观的依据，具有重要的科学意义；本项目的研究成果在目标识别、图像检索、医学图像分析、机器人导航、传感器网络等众多领域有重要的实用价值和广阔的发展前景；该项目的研究具有重要的科学意义和广泛的应用前景。

中文关键词： 分数阶微积分；形状匹配；局部特征；复杂场景

英文摘要： Shape match is a key issue in computer vision, but deformations caused by iamgeing geometry/partial occlusion/flexible movement and open contour caused by segmentation fault are the difficult problems in shape match. Extraceting local featue and matching partial shape is an effective method to solve the difficulty. The project aims to investigate the methods for partial shape match based on fractional calculus in clutter scene by theory and experiments analysis. Several aspects of the research are as follows: the method to detect edge based on fractional calculus. The mehtod to extract local feature. On the basis of image preprocessing results, combined with local feature, we try to perform partial shape match, including partial occlusion/open countor/flexible movement shap match and recognition.The research of this project can provide more direct basis in various fields, such as robot navigating automatically, detecting object automatically, analyzing medical, environmental monitoring and deep space exploration and so on. So our research has important science significance. The research of this project has also been found wide applications in object recognintion、image retrieval、medical image analysis、robot navigation and sensor network. Therefore this project not only has important science significance, but also has wide application fields.

英文关键词： frational calculus;shap match;local feature;clutter scene