基于词袋模型的多特征融合物体识别方法研究

项目名称： 基于词袋模型的多特征融合物体识别方法研究

项目编号： No.61272195

项目类型： 面上项目

立项/批准年度： 2013

项目学科： 自动化技术、计算机技术

项目作者： 李伟生

作者单位： 重庆邮电大学

项目金额： 85万元

中文摘要： 物体表示与识别是以图像或视频作为输入的实际应用系统中的核心问题和关键技术之一，在车辆辅助驾驶、智能视频监控、人机交互等领域都有着广泛的需求和应用。由于各种特征对于各类物体被正确识别的贡献不一样，单一特征在多类物体识别中难以达到良好的识别效果，如果将全局特征与局部特征有机地融合起来，物体识别系统的性能有望得到较大的提高。针对物体识别面临的难以对大规模物体统一建模的困难，本项目通过将多特征融合的方法引入到基于词袋模型的物体识别中，综合考虑词袋模型和多特征融合思想的优点，研究适合于复杂多样的物体及复杂场景下的智能化物体识别方法，主要研究内容包括：目标物体感兴趣区域获取、物体特征提取方法、基于视觉词典的图像表示方法、基于多特征融合的物体分类识别方法等，并将研究成果集成入物体识别系统进行分析和验证。通过本项目的研究有望丰富物体识别的理论和算法，为促进相关产业的发展提供理论和技术支持。

中文关键词： 物体识别；多特征融合；词袋模型；兴趣点；

英文摘要： Object representation and recognition is one of the core issues and key technologies in an actual application system which taken images or videos as inputs. It has a wide demand and application in the areas such as assisted driving, intellgent vedio surveillance and control, human-computer interaction, etc. Different features play different roles when we exactly recognize objects. According to single feature, it is difficult to achieve a good recognition rate during the object recognition. If we appropriately fuse the global feature and local feature, the performance of the object recognition system may be greatly improved. To solve the difficulties in an object recognition system that model a large scale objects, the proposal import the multi-feature fusion methods into the object recognition system based on bag-of-words model. Ultilizing the advantages of bag-of-words model and multi-feature fusion methods, this proposal is to study the intelligent object recognition methods under complicated and diversified object types and complex scenes. The major research contents include: obtaining the region of interest of an object, the feature extraction method, the image representation methods based visual vocabulary and the classfication and recognition methods based on multi-feature fusion, etc. These novel methods

英文关键词： Object recognition；Mutli-feature fusion；Bag-of-words model；Interest point；