학술논문
一种基于SURF与地理格网模型的增强现实方法 / An Augmented Reality Method Based on SURF and Geographic Grid Model
Document Type
Academic Journal
Author
Source
计算机与现代化 / Computer and Modernization. (6):47-53
Subject
Language
Chinese
ISSN
1006-2475
Abstract
图像识别与匹配是增强现实领域研究与应用的基础和关键,针对户外场景的广域性和随机性,以及目标纹理结构相似性等问题,提出一种基于SURF与地理格网模型的增强现实方法.该方法根据目标场景与地理位置的相关性,检测图像特征点并生成Location-SURF图像特征描述,基于地理格网模型构建空间四叉树索引,建成静态特征样本库.将视频帧、位置和角度信息生成特征图像,上传至服务端解析运算并与样本库训练匹配.选取宁波环球航运广场约0. 376 km2的区域,采集270余幅图像数据构建样本库并开展试验,通过现场图像的实时采集和计算,能够实现特征点的在线匹配,在此基础上通过调整点位距离比例的阈值,能够提升匹配的准确程度.基于该算法开发移动增强现实系统,运用四层技术架构实现了终端采集显示和服务端分析计算的并行化,达到真实场景与虚拟信息的融合显示效果.系统应用结果表明:该算法可以解决复杂环境下场景图像识别匹配率不高的问题,可快速地完成特征点的检测和提取,能够有效地进行样本训练和匹配,对户外移动增强现实进行了有益尝试并提供一种有效的途径.
Image recognition and matching is the foundation and key of research and application on the field of augmented reality. According to the wide area and randomness of outdoor scenes, similarity of target texture structure, an augmented reality method based on SURF and geographic grid model is proposed. The method according to the correlation between the target scene and the geographical location detects the image feature points to generate the Location-SURF image feature description, constructs the spa-tial four fork tree index based on the geographical grid, and builds the static feature sample database. Feature images are genera-ted by video frame, location and angle information, and uploaded to the server to parse operations, and trained to match with the sample database. Tests research which gathers 270 pieces of image data to construct a sample database is done in the Ningbo global navigation square about 0. 376 km2. Online feature points can be matched online through the real-time collection and cal-culation of spot image, and the accuracy of matching up can be improved by adjusting the proportion of distance threshold. A mo-bile augmented reality system is developed based on this method, which uses four layers of technology architecture to realize the parallelization of the terminal collection, display and server analysis and calculation, and achieves the fusion effect of the real scene and the virtual information. The application results show that: This method can solve the problem of low matching of image recognition in complex environment, can quickly complete the detection and extraction of feature points, can effectively complete sam-ple training and matching. As well as this method makes a useful attempt to provide an effective way in the outdoor mobile augmented reality.
Image recognition and matching is the foundation and key of research and application on the field of augmented reality. According to the wide area and randomness of outdoor scenes, similarity of target texture structure, an augmented reality method based on SURF and geographic grid model is proposed. The method according to the correlation between the target scene and the geographical location detects the image feature points to generate the Location-SURF image feature description, constructs the spa-tial four fork tree index based on the geographical grid, and builds the static feature sample database. Feature images are genera-ted by video frame, location and angle information, and uploaded to the server to parse operations, and trained to match with the sample database. Tests research which gathers 270 pieces of image data to construct a sample database is done in the Ningbo global navigation square about 0. 376 km2. Online feature points can be matched online through the real-time collection and cal-culation of spot image, and the accuracy of matching up can be improved by adjusting the proportion of distance threshold. A mo-bile augmented reality system is developed based on this method, which uses four layers of technology architecture to realize the parallelization of the terminal collection, display and server analysis and calculation, and achieves the fusion effect of the real scene and the virtual information. The application results show that: This method can solve the problem of low matching of image recognition in complex environment, can quickly complete the detection and extraction of feature points, can effectively complete sam-ple training and matching. As well as this method makes a useful attempt to provide an effective way in the outdoor mobile augmented reality.