Mesh Generation Technique and Object Identification for Robotic/Artificial Intelligence

Abstract

This technique of mesh generation is based on advance and researched quad tree approach which makes use of mathematical technique of variance for selecting quad size and further triangulate the quad for final mesh of image and later filtering of vertices is done as per mapping based on robotic model. Object identification based on Bayesian statistical and probability theorem is used to estimate the foreground object for getting selective object within image for mesh generation. This paper explains estimation algorithms for object identification by detecting background and foreground objects in image obtained from raw video frame@30fps supporting sampling format 4:2:0. This algorithm is implemented tested/verified on and written for android based ARM system and x86 for demo and quality propose.Video frame is live captured in .mp4 file format using aac/avc (H264) audio and video codec. Video is decoded and sub sampled and scaled using ffmeg framework to desired frame size and frame format for Video processing using Open source based framework integrated into propriety applications. This algorithm can be applied for various application including application in defense/artificial intelligence and medical imaging