权益分类	功能权益	普通用户	{{item.name}}会员
{{category.name}}	{{benefitItem.name}}

DEVELOPMENT OF A GUIDE DOG ROBOT

导盲犬机器人的开发

基本信息

批准号：
03650343
负责人：
MORI Hideo
金额：
$ 1.15万
依托单位：
YAMANASHI 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：
1991
资助国家：
日本
起止时间：
1991 至 1993
项目状态：
已结题

项目摘要

We have developed a Guide Dog Robot "HITOMI" and some new robot vision methods.1. Robot Mechanics : "HITOMI" is 1050mm(H)X540mm(W)X840nn(D), and its undercarriage is based on SUZUKI's motor wheelchair. It has a video camera with platform, 68040 based one board vision system, sonar range sensors and optical fiber gyroscope as multi-sensors, and V25 single board computer as locomotion control system, an command bar as human interface and 40 AH battery as power supply. It can move about 2 hours on even flat road.2. Road Following : The robot can detect lane mark or Braille block of a road or sidewalk scene in 70-100msec and can move 30cm/sec.3. Vehicle Detection : A vision based real time vehicle detection method is proposed and implemented on 68040 vision system. The detection time is depend on the distance of the vehicle, 10 msec for 30 meters and 70 msec for 5 meters. Its successful detection rate is 92-97% for fine day and 98% for cloudy day. Almost errors are caused by misunderstanding of shadow as a vehicle. Vehicle missing rate is less than 1% in all weather.4. Pedestrian detection : A real time pedestrian detection method "Rhythm model" is proposed and implemented on 4DSPs based paralle vision system. It can distinguish pedestrians from other moving objects such as bicycle and dog.Its successful distinguish rate is 76-86%. The system is not yet installed on the robot.5. Guide Map System : The digital map system for the blind guiding is a improvement of ZENRIN's 1/2,000scale JYU'UTAKU map and is now on development.

我们开发了一个导盲犬机器人“HITOMI”和一些新的机器人视觉方法。机器人机械：“HITOMI”为1050 mm（H）X540 mm（W）X840 nn（D），其底盘基于铃木的电动轮椅。该机器人由带平台的摄像机、基于68040的单板视觉系统、声纳测距传感器和光纤陀螺等多传感器组成，V25单板计算机作为运动控制系统，命令栏作为人机界面，40 AH电池作为电源。它可以移动约2小时，即使平坦的道路。2.道路跟踪：机器人可以检测车道标记或盲文块的道路或人行道场景在70- 100毫秒，可以移动30厘米/秒。3.车辆检测：提出了一种基于视觉的真实的实时车辆检测方法，并在68040视觉系统上实现。检测时间取决于车辆的距离，30米为10毫秒，5米为70毫秒。晴天探测成功率为92-97%，阴天探测成功率为98%。几乎所有的错误都是由于将阴影误解为一种载体而引起的。全天候车辆丢失率小于1%。行人检测：提出了一种真实的行人检测方法“节奏模型”，并在基于4DSP的双目视觉系统上实现。该系统能将行人与自行车、狗等其他运动物体区分开来，识别成功率为76- 86%。该系统尚未安装在机器人上。5.导盲地图系统：导盲数字地图系统是对ZENRIN的1/2,000比例尺JYU'UTAKU地图的改进，目前正在开发中。