Functionally Layered Video Coding Based on JP2K for Robot Vision Network

Abstract

We propose video coding that extracts minimum information for robots to reduce data transmission in a robot vision network where a ceiling map is generated from ceiling video from an indoor search robot and shared among search robots. Video signals are conventionally decomposed into several components and only visually significant components are transmitted. We propose that only required components for ceiling map generation and auto localization be designated as the basic layer and that other components be assigned as the enhancement layer. Only the basic layer is transmitted and received in regular communication between the robots. The enhancement layer is additionally transmitted to enable users to browse images. Experiments verified that transmission data volume was reduced by 15%, enabling more users and robots to link up in networks whose communication channel transmission is limited.