Real-time Dense Lidar from Gated Images

written by: Tobias Gruber, Frank Julca-Aguilar, Mario Bijelic, Werner Ritter, Klaus Dietmayer, Felix Heide

We present an imaging framework which converts threeimages from a gated camera into high-resolution depthmaps with depth resolution comparable to pulsed lidar mea-surements. Existing scanning lidar systems achieve lowspatial resolution at large ranges due to mechanically-limited angular sampling rates, restricting scene under-standing tasks to close-range clusters with dense sam-pling.In addition, today’s lidar detector technologies,short-pulsed laser sources and scanning mechanics resultin high cost, power consumption and large form-factors.We depart from point scanning and propose a learned ar-chitecture that recovers high-fidelity dense depth from threetemporally gated images, acquired with a flash source anda high-resolution CMOS sensor. The proposed architectureexploits semantic context across gated slices, and is trainedon a synthetic discriminator loss without the need of densedepth labels. The method is real-time and essentially turnsa gated camera into a low-cost dense flash lidar which wevalidate on a wide range of outdoor driving captures and insimulations

You should definitely visit arXiv and take a look on our preprint: