基于SOPC的投影—摄像互动系统的研究与设计

发布时间：2018-09-04 08:51

【摘要】：人机交互就是通过输入输出设备,实现人与计算机之间的一种有效互动方式。该领域技术的快速发展使得人机互动更加舒适,更加自然,并应运而生多种人机交互模式,譬如数据、触控设备、声波、人体动作等模式。本文所涉及的投影-摄像系统就是采用投影仪、摄像机作为输入输出设备,结合人体动作作为交互模式的人机互动系统。它是一种新型的多媒体互动平台,采用先进的投影显示技术、计算机视觉技术、模式识别技术和人工智能技术来营造一种奇幻动感的交互体验。目前投影-摄像系统的人机交互中存在手势跟踪分割不理想、手型变化大、特征难以提取以及运算量大无法满足系统的实时性要求等技术难题。本文主要针对以上技术难题,将人们在触摸屏上的交互控制习惯嫁接到摄像-投影系统交互中,提出采用交互手势遮挡投影光线在投影屏幕上出现的手势阴影作为交互控制器,并构建具有缩放、拉伸、双击、书写等交互指令识别功能的投影-摄像人机交互SOPC系统。首先,对摄像机系统的光度学模型进行研究分析。充分考虑摄像头的一阶径向畸变和切向畸变对图像精度的影响,实现摄像机的基本标定,并获取摄像机的内外参数。结合摄像头的畸变原理,对投影仪与摄像机系统采用射影变换的算法进行几何映射。然后,运用混合高斯模型定位移动物体,结合交互人在投影仪前方交互过程中遮挡投射光线在投影屏幕上出现的手势阴影作为交互指令,采用大律法检测阴影初步定位手势投影。结合多颜色空间的混合模型检测手势肤色,并根据真实交互手势必定位于阴影周围的投射特征,最终定位跟踪交互手势。接着,根据手指指尖具有强烈的轴对称特征,采用径向对称变换算法初步定位指尖。结合指尖半圆与支撑指尖的手指两平行线的交点与指尖所构成的两直线之间在指尖处具有强约束夹角的特性定位手指指尖。同时根据双击交互手势的动作特性,采用0TW匹配算法识别手势。最后,采用以Xilinx公司研发的Virtex-4FX为主处理器的实验箱构建基于SOPC的摄像-投影互动系统的实现平台。采用硬件的方式对CCD摄像头进行视频采集,并根据PAL制式视频流格式对视频数据进行解码。采用乒乓算法减少内部存储资源消耗,提高视频采集速率,并构建DTW算法处理IP核,提高系统的实时性。
[Abstract]:Human-computer interaction is an effective way of interaction between people and computers through input and output devices. The rapid development of technology in this field makes man-machine interaction more comfortable and natural, and many human-computer interaction modes emerge as the times require, such as data, touch devices, sound waves, human body action and so on. The project-camera system in this paper is a human-computer interactive system which uses projector, camera as input and output device and human body action as interactive mode. It is a new kind of multimedia interactive platform, which uses advanced projection display technology, computer vision technology, pattern recognition technology and artificial intelligence technology to create a kind of fantastic interactive experience. At present, in the human-computer interaction of projection-camera system, there are some technical problems, such as: gesture tracking and segmentation is not ideal, hand shape changes greatly, feature extraction is difficult, and the computation is too large to meet the real-time requirements of the system. Aiming at the above technical problems, this paper grafted the interactive control habit of people on the touch screen into the camera-projection system interaction, and proposed to use the interactive gesture shading shadow on the projection screen as the interactive controller. A project-camera human-computer interactive SOPC system with the functions of zooming, stretching, double-clicking, writing and other interactive instruction recognition functions is constructed. Firstly, the photometric model of camera system is studied and analyzed. Considering the influence of the first order radial distortion and tangential distortion of the camera on the image accuracy, the basic calibration of the camera is realized, and the internal and external parameters of the camera are obtained. Based on the distortion principle of camera, the projector and camera system are mapped by projective transformation algorithm. Then, the mixed Gao Si model is used to locate the moving object, and the interactive instructions are combined with the gesture shadow of the projecting light that appears on the projection screen during the interaction of the interactive person in front of the projector. Using the Great Law to detect the shadow of the initial positioning gesture projection. The color of gesture is detected by a mixed model of multi-color space, and the real interactive gesture is located around the shadow according to the projection feature of the real interactive gesture. Finally, the interactive gesture is located and tracked. Then, according to the strong axisymmetric characteristic of finger tip, the radial symmetry transform algorithm is used to locate the finger tip. The finger tip is located at the finger tip by combining the intersection of the two parallel lines of the finger semicircle with the finger tip and the two straight lines formed by the finger tip with strong constraint angle at the finger tip. At the same time, according to the action characteristics of double-click interactive gesture, the 0TW matching algorithm is used to recognize the gesture. Finally, the experimental box of Virtex-4FX main processor developed by Xilinx Company is used to construct the realization platform of video-projection interactive system based on SOPC. The CCD camera is captured by hardware, and the video data is decoded according to the PAL format. The ping-pong algorithm is used to reduce the consumption of internal storage resources, to improve the video acquisition rate, and to construct a DTW algorithm to deal with the IP core to improve the real-time performance of the system.
【学位授予单位】：广东工业大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP11;TP391.41;TB85

【相似文献】