面向小型机器人的超大视场红外立体视觉可行性分析

Analysis on Feasibility of Infrared Ultra-wide FOV Binocular Vision for Small Robots

摘要:

小型机器人传统视觉方法对环境适应性差。提出了一种基于双目超大视场红外相机的环境感知方法。利用大视场镜头高阶奇次多项式模型,建立了超大视场红外双目立体成像的水平和垂直视差数字模型。以视差模型为基础建立超大视场红外双目视觉模型,研究了超大视场红外双目系统立体视觉范围和阈值。搭建了视场为170°×128°的超大视场红外立体视觉系统,分析了立体视觉范围及阈值应用于小型机器人视觉的可行性。同时,针对照度不均、雾霾等条件下的场景开展超大视场红外双目立体视觉实验研究,构建了双目图像标准视差图,结果表明,超大视场红外双目立体视觉系统对复杂场景具有良好的适应性,基本能够满足小型机器人视觉系统需求。

Abstract:

To overcome such defects as poor environmental adaptability of traditional machine vision, a kind of environmental perception method based on binocular infrared ultra-wide field of view (FOV) camera was proposed. Horizontal and vertical models for ultra-wide FOV binocular infrared imaging were established by using high order polynomial model. Stereo vision range and threshold were researched on the basis of parallax model. Infrared ultra-wide FOV binocular stereo vision system with the vision field of 170°×128° was established. Stereo vision experiments were carried out under different conditions, such as uneven illumination, smog, etc. Subjective and objective evaluation shows that infrared ultra-wide FOV binocular stereo vision system has good applicability to complex scenes and is able to realize 3D depth perception of the infrared ultra-wide FOV, which basically meets the requirements of small robot vision system.