1-Bit 3-Dimensional Imaging

1 位 3 维成像

• 批准号: 514479946
• 负责人: Professor Dr. Bhaskar Choubey
• 金额: --
• 依托单位: Lehrstuhl für Hochfrequenzsysteme in der Kommunikationstechnik (IHCT)
• 依托单位国家: 德国
• 项目类别: Research Grants
• 资助国家: 德国
• 项目状态: 未结题
• 来源: https://gepris.dfg.de/gepris/projekt/514479946?language=en
• 关键词: Bit; Dimensional; Imaging

Time-of-Flight (ToF) cameras are 3D imaging systems able to capture the geometry of a scene. Typically, little attention is paid to the quantization scheme and uniform quantization of the measurements with enough number of bits is assumed. In this project we aim to follow up on recent results showing that reliable multiple-path 3D imaging can be obtained from one-bit-quantized data. The first one-bit ToF pixel will be designed, fabricated and evaluated. The ability to obtain reliable 3D images from one-bit data enables a massive reduction of the data to be captured and transmitted by the camera. This reduces the hardware requirements upstream in terms of memory and bandwidth. Furthermore, one-bit ToF imaging deals with a new form of inverse problem, in which the measurements have undergone a non-linear transformation. The reduction of the data flow per pixel also allows for larger ToF pixel arrays, paving the way for Megapixel Tof imaging. The binary nature of the measurements also avoids the need for ADC banks, freeing chip space and further helping on this front. Despite to date all existing ToF cameras rely on uniform quantization, preliminary work of Dr. Heredia Conde has demonstrated the feasibility of the proposed concept by means of realistic simulations. These first results suggest that such a one-bit ToF pixel would exhibit superior performance to alternative sensing schemes relying on different tradeoffs between number of measurements and bit depth for the same bit budget. In this project we will design and construct the first one-bit ToF camera. This hardware novelty will be coupled to an algorithmic counterpart, able to estimate the geometry of the observed scene from the measurements. The result will be an unprecedented time-resolved computational imaging system, in which the information is first encoded into lightweight streams of one-bit data and then decoded into one or several depth images. The necessary tasks for the development of a functional one-bit 3D imaging system can be summarized as follows. First, the architecture of the one-bit pixel will be designed and simulated. Then a single pixel or a small array will be fabricated for testing. The fabricated device will be tested for operability. Building upon the insights gained from this first device, a complete array of ToF pixels will be fabricated and integrated with a modulated NIR illumination system and control circuitry and the system will be tested. The necessary algorithms for reconstructing the 3D geometry of the scene from the obtained one-bit data will be developed. This includes the disentanglement of multiple return paths per pixel. Finally, the overall performance of the system as 3D imaging sensor is to be systematically evaluated and compared to the state of the art and to the results obtained from simulations.

飞行时间(ToF)相机是一种3D成像系统，能够捕捉场景的几何形状。通常，很少关注量化方案，并且假设具有足够比特数的测量的均匀量化。在这个项目中，我们的目标是跟进最近的结果，这些结果表明可以从一位量化的数据中获得可靠的多路径3D成像。第一个1位ToF像素将被设计、制造和评估。从一位数据获得可靠的3D图像的能力使相机捕获和传输的数据量大大减少。这降低了上游在内存和带宽方面的硬件要求。此外，一位ToF成像处理了一种新形式的逆问题，其中的测量值经历了非线性变换。每像素数据流的减少还允许更大的TOF像素阵列，为百万像素TOF成像铺平了道路。测量的二进制性质还避免了对ADC组的需要，从而释放了芯片空间，并在这方面提供了进一步的帮助。尽管到目前为止所有现有的TOF相机都依赖于均匀量化，但Heredia Conde博士的初步工作已经通过真实的模拟证明了所提出的概念的可行性。这些第一个结果表明，对于相同的比特预算，这种1比特ToF像素将表现出比依赖于相同比特预算的测量数量和比特深度之间的不同权衡的替代感测方案更好的性能。在这个项目中，我们将设计和建造第一个一位ToF相机。这种硬件的新颖性将与算法相对应，能够根据测量结果估计观察到的场景的几何形状。其结果将是一种史无前例的时间分辨计算成像系统，其中信息首先被编码成轻量级的一位数据流，然后解码成一个或几个深度图像。开发功能强大的一位3D成像系统的必要任务可以概括如下。首先，对一位像素的体系结构进行设计和仿真。然后，将制造单个像素或小阵列用于测试。制造的设备将接受可操作性测试。基于从第一个设备获得的见解，将制造一个完整的ToF像素阵列，并将其与调制近红外照明系统和控制电路集成在一起，并将对该系统进行测试。将开发从所获得的一位数据重建场景的3D几何图形所需的算法。这包括每个像素多个返回路径的解缠。最后，对系统作为3D成像传感器的整体性能进行了系统的评估，并与现有技术和仿真结果进行了比较。