Compressive Super-Pixel LiDAR for High-Framerate 3D Depth Imaging

We propose a new sampling and reconstruction framework for full frame depth imaging using synchronised, programmable laser diode and photon detector arrays. By adopting a measurement scheme that probes the environment with sparse, pseudo-random patterns, our method enables eye safe LiDAR operation, while guaranteeing fast reconstruction of depth images with a high signal-to-noise ratio (SNR). Building up on the observation that certain quantities derived from the photon count histograms are sparse in either the Li-norm or have small total variation (TV), reconstruction is performed via compressed sensing (CS) and takes approximately 30 s per frame. To speed up reconstruction, we further introduce a checkerboard tiling approach (CB-CS) that reduces the processing time to milliseconds per tile, with comparable or even less reconstruction error. Although in our experiments we reconstruct tiles sequentially at a frame rate of similar to 4 Hz, this process is highly parallelizable and has the potential to achieve 1 kHz frame rates.