应用于VR室内体验的跟踪识别系统

CASIA OpenIR > 毕业生 > 硕士学位论文

	应用于VR室内体验的跟踪识别系统
	徐文慧
	2017-05-31
学位类型	工程硕士
中文摘要	近年来，虚拟现实技术得到了快速发展。其中，定位跟踪技术作为人与虚拟场景之间信息交互的一个重要因素，是近年来虚拟现实技术发展的一个重要领域。在目前逐渐兴起的虚拟现实体验中，往往同时有多个用户同时参与虚拟场景体验，用户之间交错运动，用户需要的活动范围不断扩大，这对VR定位跟踪技术不断提出新的要求。目前的空间定位技术应用到VR中仍有很多不足，例如，有的不能满足对多个目标的定位跟踪，有的不易扩展到更大的范围，有的设备昂贵、成本太高。所以，为满足室内多目标同时参与的虚拟现实体验，改进目前VR中的定位跟踪方式，寻找实现多目标识别的低成本解决方案，对满足多人互动的VR体验具有重要意义。本文根据实际需求，研究并设计应用于虚拟现实室内体验的多目标定位跟踪识别系统。整个系统用到的技术主要有室内红外定位跟踪识别技术、网络通信技术、虚拟场景显示技术。该系统选用低成本的红外摄像头和红外标记点，采用主动式红外光学定位技术实现目标的定位。在红外光学定位技术的实现上，本文比较了极线约束法和求异面直线最小距离法，最终选择精度更高、更简单的极线约束法，实现对同一标记点不同像素坐标的匹配。在像素坐标匹配的过程中，我们借助第三个相机拍摄的图像，利用重投影判别误匹配的标记点，提高了标记点匹配的准确率。我们通过比较最小二乘法和投影射线相交法，选择精度更高的最小二乘法对红外标记点的世界坐标进行重建。之后，本文提出了通过距离向量识别每个红外标记点的方法。在红外标记点模块的构建上，我们选择用4个红外LED组成红外标记点模块，实现对目标的定位和识别。在标记点识别的过程中，本文通过设置标记量，并对匹配距离前后的元素都进行比较，得到最匹配的结果。之后通过排错处理和消重叠处理进一步保证红外标记点识别的可靠性。在目标的跟踪上，本文采用世界坐标系下的卡尔曼滤波对红外标记点的世界坐标进行跟踪处理。在网络通信模块，本文采用UDP协议，实现服务器与多个相机处理器的快速通信，并通过两个事件控制服务器端多线程通信的同步机制，满足系统的实时性需求。本文采用TCP协议实现服务器与客户端的可靠传输，同样采用事件机制实现服务器端主线程和通信线程的同步。最后，系统借助Unity3d引擎和虚拟现实头盔显示器实现虚拟场景的交互体验。经过试验验证，该系统采用的定位方式和识别方法可以应用到虚拟现实中，定位精度、数据的稳定性和系统的响应速度可以满足虚拟现实体验的要求。系统目前已经测试了对3个目标的定位、跟踪和识别。采用红外相机的曝光时间和帧率对系统的性能有较大的影响。采用具有更低曝光时间、更高帧率的相机，能够降低系统的响应时间，提高系统的定位精度和识别准确率。但是受相机采集图像不严格同步的影响，目前采用的红外相机构成的系统仍然不能满足较快速度下运动目标的定位跟踪和识别。
英文摘要	In recent years, virtual reality (VR) technology has been rapidly developed. Among all the VR technology, positioning and tracking technology which is an important factor in the interaction between users and virtual scene, is an important field in the development of VR technology in recent years. In the current VR experience, there are often a number of users to participate in the virtual scene at the same time. The users move staggered, and the scope of activities need to expand, which put forward the new demand to VR positioning and tracking technology. There are many deficiencies in the existing technology. For instance, some can not meet the positioning of multiple users, some are difficult to expand to a larger scope, some use expensive equipment, and the cost is too high. Therefore, improving the current methods in the positioning and tracking, and finding a multi-target identification solution are very important to meet the multi-target VR experience’s interactive. Based on the actual demand, this paper designs a multi-target location tracking and recognition system for VR indoor experience. The technology used in this system are infrared position tracking technology, network communication technology, and virtual scene display technology. The system uses low-cost infrared camera, infrared markers, and active infrared optical positioning technology to achieve the goal of positioning. On the basis of infrared optical positioning technique, this paper compares the epipolar constraint method and the method of finding the distance of different lines in different plane, and finally chooses the epipolar constraint method which is simpler and more accurate to realize the matching of different pixel coordinates of the same marker. In the process of pixel coordinates matching, we use the image taken by the third camera to identify the incorrect matching with the re-projection technology to improve the accuracy of the matching. In this paper, via comparing the least squares method and the difference line intersection method, the former method is chosen to reconstruct the world coordinates of the infrared markers. After that, this paper presents a method to identify each infrared markers by the distance vector of the infrared marker. In the construction of infrared markers module, we use four infrared LED to composition the infrared markers module, and to achieve the positioning and identification of the users. In the process of markers’ recognition, this paper set a variable value, and compare the distance before and after the matching distance to get the most matching results. And then through the troubleshooting and elimination of overlapping to further ensure the reliability of the recognition. In the tracking of the targets, this paper uses the Kalman filter under the world coordinate system to track the world coordinates of the infrared markers. This system uses UDP protocol to realize the rapid communication between server and multiple camera processors. And for meeting the real-time requirements of the system, we control the synchronization mechanism of the server’s multi-thread by two events. In this paper, the TCP protocol is used to realize the reliable transmission of the server and the client. The event mechanism is used to synchronize the server main thread and the communication thread. Finally, the system uses the Unity3d engine and VR helmet display to achieve virtual scene interactive experience. Based on the experimental results, the positioning method and the recognition method can be applied to the virtual reality. The positioning accuracy and the stability of the data and the response speed of the system can meet the requirements of the VR experience. The system has now tested the positioning, tracking and identification of three targets. The exposure time and frame rate of the infrared camera have a great influence on the performance of the system. Using a camera with lower exposure time and higher frame rate can reduce the response time of the system, improve the positioning accuracy and recognition accuracy of the system. However, by the effect that camera capturing the image is not strictly synchronized, the infrared camera used by the system still can not meet the positioning and identification of the target with faster speed.
关键词	虚拟现实、定位、跟踪、识别
文献类型	学位论文
条目标识符	http://ir.ia.ac.cn/handle/173211/14633
专题	毕业生_硕士学位论文
推荐引用方式 GB/T 7714	徐文慧. 应用于VR室内体验的跟踪识别系统[D]. 北京. 中国科学院研究生院,2017.

条目包含的文件
文件名称/大小	文献类型	版本类型	开放类型	使用许可
毕业论文最终版.pdf（2664KB）	学位论文		限制开放	CC BY-NC-SA