Design and implementation of a visual sensor - based depth estimation module in a system-on-chip architecture
Botero Gallego, Jorge Iván · 2017
In this document, an image-based depth estimation module is proposed. It uses a visual sensor, which in this case, is a stereoscopic camera and a semi-global block matching (SGBM) technique to compute the disparity in the scene. After selecting and implementing the previously mentioned method, it is then adapted to a SoC (system-on-chip) architecture since it is meant to be embedded into a drone. An initial calibration and rectification procedure is necessary to assure that all captured images are consistently aligned between both left and right cameras.