WiMi Hologram Cloud Inc (WIMI)

[StockpickerCAPC]

WIMI develops a dedicated computer system chip SoC-FPGA for real-time single-pixel holographic imaging.

Source
https://t.cj.sina.com.cn/articles/view/1765776051/693f9ab30200110bw?from=tech

March 2, 2023

Holographic single-pixel imaging is a computationally intensive technique that requires compact and efficient devices for specific applications. Embedded computers may be a potential solution, but they are not suitable for reconstruction calculations due to their low computational performance. Therefore, there is a need to realize a small computer with high computing performance in a single large scale integration (LSI) chip. WIMI has developed a dedicated system chip for single-pixel holographic imaging using Field-Programmable Gate Array (FPGA, Field-Programmable Gate Array).

FPGA is an LSI chip that can freely rewrite logic circuits on site. FPGAs can perform high-performance computing by designing application-specific circuits. WIMI hologram implements iterative calculation method, Hadamard transform and differential ghost imaging (DGI) on FPGA, and the whole process from reconstruction calculation to display is completed on FPGA, which speeds up image reconstruction.

It is reported that in this research and development project, WIMI has developed a special computer system chip (SoC, System on a Chip) FPGA for real-time single-pixel holographic imaging. SoC FPGA is a type of LSI in which an embedded CPU and FPGA are implemented on a single-chip system. It has higher computing performance than an embedded CPU alone, more flexibility than an FPGA alone, and can be much smaller than a computer. Furthermore, the choice of a reconstruction algorithm that should be implemented as a computational circuit is important for designing a computer dedicated to single-pixel imaging. Although FPGA has high computing performance, its hardware resources are limited, and it is not good at complex calculations such as division and square root calculations. The optimization method and deep learning in the algorithm can obtain high-quality reconstruction in single-pixel imaging, and the optimization method has the problem of computational load due to the iterative method.

WIMI Holographic SoC FPGA test process for real-time single-pixel holographic imaging:
The camera lens forms an image of the target object on the DMD. The image of the target object is modulated by encoding the mask pattern displayed on the DMD. The modulated light is collected by a lens and measured by a single-element detector before being converted to a digital signal. In addition, a dedicated computer reconstructs an image of the target object based on the light intensity.
The FPGA partially reconstructs the image, while the embedded CPU on the WIMI Holographic real-time single-pixel holographic imaging dedicated computer system chip SoC-FPGA generates and initializes the drawing on the holographic display.

The object light is formed on the DMD by the camera lens. A coded mask pattern is displayed on the DMD by which the subject light is modulated. The modulated light is collected by a lens and measured as light intensity by a single-element detector. The obtained light intensity is converted from an analog intensity signal to a digital signal by an analog-to-digital converter. Repeat this when switching encoding masking modes. The receive circuitry in the FPGA saves the converted signal in the FPGA's internal memory when the sync signal is asserted, which is generated when the DMD switches to a new encoding mask mode. After the receiving circuit saves the signal for a specified number of times, the reconstruction circuit starts to calculate the holographic image of the target object. Then, the embedded CPU on the SoC FPGA chip receives the reconstruction result and displays it on the dedicated display panel to realize real-time observation of the holographic image of the target object on the dedicated holographic display panel.

In order to improve computing efficiency, WIMI adopts a ghost imaging related algorithm in the SoC-FPGA of single-pixel holographic imaging, which is suitable for FPGA. The algorithm has low memory usage and simple calculation form. The algorithm introduces an encoding mask pattern optimization. This ghosting algorithm improves image quality, but has higher memory requirements.
Specifically, the implementation of the ghost imaging algorithm requires the use of two spatially separated beams:
- a reference beam and an object beam. It is an imaging method based on cross-correlation or cross-correlation-like techniques, which enables image reconstruction by using single-photon detectors. The basic principle of the algorithm is to take a cross-correlation measurement between two spatially separated beams and then use a computer algorithm to reconstruct the target image. For example, a reference beam passes through a random perturbation device, which produces a random pattern of light intensity. These light intensity patterns are transmitted into the object beam, where they are detected by single-photon detectors after passing the object. The light intensity values ??measured by the single-photon detector are recorded and cross-correlated with the light intensity pattern of the reference beam. The information of the target image can be obtained by averaging multiple cross-correlation measurements.

The ghost imaging algorithm has some unique advantages in imaging, such as the ability to realize three-dimensional holographic imaging without an objective lens, suitable for imaging at low light levels, and suitable for multiple imaging modes such as transmission and reflection imaging.

WIMI Holographic's real-time single-pixel holographic imaging dedicated computer system chip SoC-FPGA can obtain higher image quality than traditional holographic imaging technology, and through SoC-FPGA integrated structure optimization and algorithm optimization can achieve size, image quality and speed Real-time holographic imaging is realized through the improvement. And because the SoC-FPGA dedicated to real-time single-pixel holographic imaging is very compact compared with typical computer servers, it can extend the application of single-pixel imaging to the Internet of Things and outdoor applications. Dedicated specific applications also include the implementation of satellite topographical surveys, and can also be used for object tracking to build car navigation IoT systems.