项目名称: 基于局部模式分析的特定目标检测方法研究
项目编号: No.61302154
项目类型: 青年科学基金项目
立项/批准年度: 2014
项目学科: 无线电电子学、电信技术
项目作者: 朱承飞
作者单位: 中国科学院自动化研究所
项目金额: 26万元
中文摘要: 特定目标检测在机器人自主导航、战场区域监测、视觉制导等方面具有重要的意义。由于特征匹配在复杂环境下适应能力差,目前通常采用模板匹配的方法。为减小模板匹配的搜索范围,部分研究者采用基于视觉注意机制的方法来提取ROI,但目前方法仍不能满足特定目标实时检测的需求。为此,本项目提出了一种基于局部模式分析的特定目标检测方法。该方法利用基准图像中目标和环境的局部模式分布建立目标视觉显著性模型,使得可以通过检索的方式快速得到实时图像的显著性图像,进而提取出ROI;利用基准图像目标区域典型局部模式的组合建立目标码模型,使得可以采用汉明距离匹配准则快速得到实时图像中目标的精确位置。本项目将对其中的目标显著性模型建立方法、基于目标显著性模型的ROI提取方法、目标码模型建立方法以及基于目标码模型的目标精确定位方法等方面展开深入地研究,以推动特定目标实时检测方法的进一步发展,具有重要的研究价值和应用前景。
中文关键词: 局部模式分析;目标视觉显著性模型;感兴趣区域;目标码模型;特定目标检
英文摘要: The technology of specified target detection has great significance in autonomous robot navigation, battlefield surveillance, vision-based guidance, and so on. Since the method of feature matching has low environmental adaptivity under clutter scenes, researchers usually employ template matching method to precisely localize the specified target currently. To reduce the searching range of template matching, some researchers ultilize visual attention mechanisms to get ROI (Region of Interest). However, these methods either ignore the specified target information which brings a risk of missing the interested target, or don't have a reasonable calculation model, thus can not meet the reqirement for real-time specified target detection.To solve these problems, this project proposes a specified target detection method based on local pattern analysis. The proposed method uses local patterns of the target as well as its environments in the reference image to build TVSM(Target Visual Saliency Model). Based on the acquired TVSM, VSM (Visual Saliency Map)for a real-time image can be fastly obtained by indexing, from which ROI can be extrated. As for the precise target localization stage, we concatenate typical local pattern codes of the target in the reference image to get OCM (Object Code Model). Thus, the matching can be
英文关键词: local pattern analysis;target visual saliency model;Region of interest;object-code model;specified object detection