A Study Of Depth Acquisition Based On Phase Coding

In this highly automated and information-oriented society, the 3D information is more widespread used in lots of areas such as in driverless cars, motion sensing, 3D printing and so on. It has begun to change people’s life deeply. Now, the depth acquisition for various scenarios has been an important and difficult research topic. We carried out research into the existing depth acquisition technology and propose two novel phase-coding based methods in this paper.The first uses a dual-frequency phase-coding structured light with a single shot, which can be used in dynamic scenes for real-time and dense depth acquisition. Different from the former methods, we avoid the Fourier spectra separation by using both color and intensity information to encode the phase of one pixel. And a number-theoretical approach is employed to acquire the absolute phase differences conveniently and speedily. With the only one pattern, this method can be used in dynamic scenes. Meanwhile, we raise a procedure to overcome the measurement sensitivity caused by the number theory. According to the practical experiment results, this method is robust to objects with changeable depth and sharp edges. And even in the colorful or dynamic scenes, depth can be also acquired. Meanwhile, the results shows the higher accuracy of the proposed method compared with Kinect and To F.The second method in this paper is also based on the phase-coding and combines the active and passive depth acquisition technique. We adopt the classical three-frame phase shifting patterns to illuminate the scenes in this method. The decoding procedure depends not only on the active phase based structured light but also on the passive stereo matching based on the feature of the scenes. With this method, we solved the depth sensing problem for the isolated points or areas by the accurate phase unwrapping which is difficult for the traditional phase shifting method. Meanwhile, owing to the combining use of the scene feature and the information contained in the deforming patterns, this method is very robust and reduces the time complexity in phase unwrapping, accelerates the depth sensing procedure. The high accuracy and strong robustness has been verified by the experimental results. This method can be widely used in complex static scenes or objects with slow movement. If the projection and camera capturing frequency meets requirements, it can also be used in dynamic scenes.