Industry Focus

According to foreign media reports, holographic digital twin technology supplier MicroCloud Hologram (HOLO) announced the development of a point cloud denoising algorithm for real-time 3D holographic reconstruction of single-photon lidar data. The algorithm is independently developed by MicroCloud Hologram, which will further improve the company's intellectual property protection system, maintain technological leadership, and enhance core competitiveness.

Although 3D holographic lidar point cloud imaging continues to develop rapidly, the currently available computational imaging algorithms are often too slow, not detailed enough, and require extremely high computing power, and even the scene depth estimation algorithm based on CNN (convolutional neural network) is difficult to meet the real-time requirements after training. Holo proposes a new algorithm structure to meet the requirements of speed, robustness and scalability. The algorithm applies point cloud denoising tools to computer graphics, which can effectively model the target surface as a 2D manifold embedded in 3D space.

The algorithm can incorporate information about the observation model, such as Poisson noise, the presence of bad pixels, compressed sensing, and more. The algorithm also uses computer graphics flow modeling tools that can process tens of frames per second with massively parallel denoisers. Holo's algorithm consists of three main steps: depth update, intensity update, and background update.

Depth update: Take a gradient step on the depth variable and denoise the point cloud using the point set surface algorithm. The update operates in a coordinate system in 3D holographic space. Under the control of the kernel, adaptive is performed on a smooth and continuous surface. In contrast to traditional depth image denoising, HOLO's point cloud denoising can handle any number of surfaces per pixel, regardless of format. In addition, all 3D points are processed in parallel, which significantly reduces calculation time.

Intensity update: Reduce noise by taking gradient steps against the coordinates of individual pixels in 3D holographic space. In this way, it is sufficient to consider the correlation between points within the same surface. The nearest low-pass filter is used for each point. This step considers only local correlations and processes all points in parallel. After the denoising step, points below a given intensity threshold, the minimum allowable reflectivity, are removed.

Background update: This depends on the characteristics of the lidar system, similar to the intensity and depth update. In a double-Bragg-grating scanning system, the laser source and single-photon detector axes are different, and the background count is not necessarily spatially related. Therefore, no spatial regularization is applied to the background, in which case the noise reduction operation is reduced to a constant equation. Background detection is similar to passive images in single Bragg raster scanning systems and LiDAR arrays. In this case, spatial regularization helps improve the estimates, so ready-made image denoising algorithms with low computational complexity can be used.

HOLO real-time 3D holographic reconstruction based on single-photon data uses a new computing framework. By combining statistical models with highly scalable computational tools for computer imaging techniques, the framework allows for 3D reconstruction of complex outdoor scenes with a processing time of about 10-20 milliseconds. The algorithm developed by Holo can process every pixel face, allowing object detection and imaging in complex scenes. It also enables stable real-time target reconstruction of complex moving scenes, paving the way for video-rate single-photon LiDAR technology for 3D holographic imaging applications.

3D holographic scene reconstruction is widely used, such as autonomous navigation and environmental detection, and there are several subdivisions, such as RGB-D sensors for emissivity imaging, stereoscopic imaging, or full-waveform LiDAR 3D holographic imaging. Compared with traditional solutions, Holo's single-photon LiDAR technology solution has several outstanding advantages. The technology uses a secure laser light source with low power consumption and high sensitivity to reconstruct high-score 3D holographic images in highly scattered underwater or polar fog environments.