Research on Active Vision based on a Human Visual System

基于人类视觉系统的主动视觉研究

• 批准号: 06650465
• 负责人: MIYAZAKI Fumio
• 金额: $ 1.41万
• 依托单位: Osaka University
• 依托单位国家: 日本
• 项目类别: Grant-in-Aid for General Scientific Research (C)
• 财政年份: 1994
• 资助国家: 日本
• 起止时间: 1994 至 1995
• 项目状态: 已结题
• 来源: https://kaken.nii.ac.jp/en/grant/KAKENHI-PROJECT-06650465/
• 关键词: Active Vision; Head-eye System; Gyro-sensor; VOR (Vestibulo Ocular Reflex); Sensor Fusion; LPM (Log Polar Mapping); Object Tracking; Zero Disparity Filter; 視差フィルタ; VOR; 協調制御; トラッキング; パース-ト運動; バ-ジェンス運動; 前庭性動眼反射

In this research, we investigated the active vision which integrates the vision and motion of the eye to realize a vision system for autonomus mobile robots. This is motivated from insects such as bees and flys which performs special pattern of movement to simplify the visual processing.In 1994, we investigated the structure of human visaul system and movements of the eyes to decide the model of the artificial retina and control system. We found that there are several kinds of eye movements such as saccade, pusuit, vergence, and VOR (Vestibulo Ocular Reflex) in human visual system to simplify the visual processing. Furthermore, we developed the head-eye system which has two eyes and Gyro-sensors mounted on a head. The system has a performance as good as human eye movements and useful for various research such as mobile robot application and a hand-eye system because it is very small and right (about 1.5kg).In 1995, we realized a function like human VOR which rotates both eyes to the opposite direction to the head rotation to compensate for the disturbance to the head. This function is important for mobile robot to maintain fixation while moving a rough terrain. We realized a similar ability as a human visual system by considering both dynamics of motor and sensor and kinematics of the head eye system. Furthermore, we proposed a tracking method using LPM (Log Polar Mapping) which is similar to the human retinal structure. The method extract the target object by using the several disparity filters and realize tracking by comparing the ouput of disparity filters.

在本研究中，我们探讨整合眼睛视觉与运动的主动视觉，以实现一个可携式移动的机器人视觉系统。1994年，我们研究了人类visaul系统的结构和眼睛的运动，确定了人工视网膜和控制系统的模型。我们发现，在人类视觉系统中，存在着几种眼动，如扫视、聚散和前庭眼反射（VOR），以简化视觉处理过程。此外，我们开发了头眼系统，它有两个眼睛和陀螺仪传感器安装在头上。该系统具有与人眼运动一样好的性能，并且对于诸如移动的机器人应用和手眼系统的各种研究是有用的，因为它非常小和正确（约1.5kg）。该功能对于移动的机器人在崎岖地形上移动时保持固定是很重要的。我们实现了类似的能力，作为一个人类的视觉系统，同时考虑电机和传感器的动力学和头眼系统的运动学。此外，我们提出了一种使用LPM（对数极坐标映射），这是类似于人类视网膜结构的跟踪方法。该方法利用多个视差滤波器提取目标对象，通过比较视差滤波器的输出来实现跟踪。