基于相位编码结构光的深度获取研究

发布时间：2018-02-04 01:39

本文关键词： 深度获取结构光相位编码双目匹配　出处：《西安电子科技大学》2014年硕士论文　论文类型：学位论文

【摘要】：在高度自动化、高度信息化的当今世界,三维信息的应用场景已经越来越广泛,并开始深刻地改变着人们的生活,如无人汽车、体感游戏仪和3D打印机的实现及应用。如今,针对各种场景的深度信息获取技术已成为现阶段的研究重点及难点。本文在对现有深度获取方法进行研究的基础上,提出了两种基于相位编码结构光的深度获取技术。第一种方法中使用了单帧混频相位编码结构光,可用来实现动态场景下的稠密深度获取。且与前人方法不同,该方法利用颜色及强度的变化来对双频相位进行编码,避免了傅里叶频谱分离的问题,使得分频变得更加容易。同时该方法仅采用一幅模板,即实现了基于余数定理的相位快速展开,可应用于动态场景之中。此外,为增加方法鲁棒性,在进行相位展开过程中,该方法充分利用了邻域信息,以克服由余数定理引起的测量敏感性。该方法在使用单幅模版、不增加设备复杂性的情况下,有效地提高了相位展开的速度,实现了适用于动态场景的高精度、高分辨率的深度信息获取。经过实际实验验证,该方法对于表面深度多变物体、复杂物体边缘,本算法均有较强的鲁棒性,且对彩色场景、动态场景也能实现精确深度测量。实验结果表明,相对于Kinect和ToF两种商用深度获取设备来说,该方法有着更高的精度。第二种方法采用了基于相位编码的主被动结合式深度测量。该方法基于经典相移(Phase-Shift)三维轮廓测量法,进行主动结构光投射,解码时,则利用场景自身特征进行被动双目匹配和相移模板进行结构光主动测距,进行主被动联合解码。该方法解决了传统三帧三维轮廓测量法中出现的无法对孤立相位进行正确展开,从而无法正确估计场景深度的问题;同时通过联合利用场景特征及变形的模板特征,增加了深度测量的鲁棒性,降低了相位展开的时间复杂度,大大加速了高精度结构光测距过程。经实验验证,该方法测量精度高,鲁棒性强,对物体颜色信息不敏感,可广泛用于静态及慢速运动场景,如室内建模等。且只要投影及拍摄设备的帧率达到一定要求,该方法也可适应动态场景的需要。
[Abstract]:In the highly automated and highly information-based world, the application of three-dimensional information scene has become more and more extensive, and has begun to profoundly change people's lives, such as unmanned cars. The realization and application of somatosensory gameplay and 3D printer. Depth information acquisition technology for various scenes has become the focus and difficulty of the research at this stage. Based on the research on the existing depth acquisition methods. Two depth acquisition techniques based on phase-coded structured light are proposed. In the first method, single-frame mixed-phase coded structured light is used to achieve dense depth acquisition in dynamic scenes, which is different from previous methods. The method uses the change of color and intensity to encode the dual-frequency phase, which avoids the problem of Fourier spectrum separation and makes the frequency division easier. At the same time, the method uses only one template. In addition, in order to increase the robustness of the method, the method makes full use of the neighborhood information in the phase unwrapping process. In order to overcome the sensitivity caused by residue theorem, this method can effectively improve the speed of phase unwrapping and realize the high accuracy of dynamic scene by using single template without increasing the complexity of equipment. High resolution depth information acquisition. The experimental results show that the method is robust to the surface depth and complex object edge, and is robust to the color scene. Dynamic scene can also achieve accurate depth measurement. Experimental results show that compared with Kinect and ToF commercial depth acquisition equipment. The method has higher accuracy. The second method is based on phase coding combined with active and passive depth measurement. This method is based on the classical phase shift Phase-Shift 3D profile measurement method. For active structured light projection and decoding, passive binocular matching and phase-shifting template are used to carry out active structural light ranging. This method solves the problem that the isolated phase can not be expanded correctly and the depth of the scene can not be estimated correctly in the traditional three-frame 3D contour measurement method. At the same time, the robustness of depth measurement is increased, the time complexity of phase unwrapping is reduced, and the high precision structured light ranging process is greatly accelerated by using scene features and deformed template features. This method has the advantages of high measurement accuracy, strong robustness and insensitivity to object color information. It can be widely used in static and slow moving scenes, such as indoor modeling, and the frame rate of projection and shooting equipment can meet certain requirements. The method can also meet the needs of dynamic scenes.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN911.2

【相似文献】