跟隨前導者軌跡行進的自走車;Automatic vehicle following the guide’s trajectory

NCU Institutional Repository > 資訊電機學院 > 資訊工程研究所 > 博碩士論文 > Item 987654321/65689

jsp.display-item.identifier=請使用永久網址來引用或連結此文件: http://ir.lib.ncu.edu.tw/handle/987654321/65689

题名:	跟隨前導者軌跡行進的自走車;Automatic vehicle following the guide’s trajectory
作者:	牛建焜;Niu,Chien-Kun
贡献者:	資訊工程學系
关键词:	自走車;前導者;軌跡;單一相機;aoutomatic;vehicle;trajectory;camera
日期:	2014-07-28
上传时间:	2014-10-15 17:08:18 (UTC+8)
出版者:	國立中央大學
摘要:	透過一個可以自動跟隨前導者軌跡行進的自走車，不論對於廣大地域的導覽、貨品的搬運，或是由於身體障礙需要協助行動的障礙者，都可大大的提高便利性。本研究藉由一個自主控制的系統，可減少許多使用者的負擔；而本系統除了自走車本身外，僅需要一台單眼相機與電腦，就可以完成跟隨前導者軌跡行進的動作。整個系統共分為四部分：一是前導者偵測，二是前導者追蹤，三是前導者距離與方向的估計，四是輪椅控制。第一部分的前導者偵測是以前導者的方向梯度累積圖 (Histogram of Oriented Gradients) 特徵，經由支援向量機 (Support Vector Machine) 分類器判斷出可能是前導者的物件。第二部分的前導者追蹤，則是透過比對物件的色彩分佈圖 (histogram of color) 確認多物件中最有可能的前導者。第三部分則僅透過一台光學單眼 PTZ 相機，判斷前導者與自走車之間的相對關係，再加上已知相機的架設高度以求得前導者的距離與左右方向。第四部分的輪椅控制是依據前導者與自走車的相對關係，決定自走車的速度與方向。操控自走車以盡量維持前導者在自走車前方 2 公尺的位置，這是前導者可以與身體障礙者溝通，彼此之間又有緩衝的空間，不會讓輪椅撞上前導者的距離。跟隨前導者軌跡行進的自走車是伍氏科技的電動輪椅 Mambo 513；使用 1 部 Logitech 的 QuickCam® Sphere AF PTZ 彩色相機偵測前方環境。在 Intel Core™i7-3740QM 2.70GHz 及 4GB RAM 的個人電腦上執行，可達每秒 6 至 15 張影像的處理速度。使用多種環境狀況的 3580 張影像實驗，前導者偵測率為 88%，前導者追蹤的正確率為 91%，前導者與自走車距離的錯誤率則有 12%。本研究僅使用一台 PTZ 相機，做到軌跡重現的目標。 ;We can greatly improve convenience for a vast area of navigation, transportation of goods, or the disable through an automatic vehicle following the guide’s trajectory. This study can reduce the burden of many users by an autonomous system control. In addition to automatic vehicle, the system requires only a monocular camera and a computer to complete the automatic vehicle following guide’s trajectory. The whole system is divided into four parts, including guide detection, guide tracing, estimation of the distance and direction of the guide and automatic vehicle control. The first part is the guide detection by the features of Histogram of Oriented Gradients and then determining the possible guide objects via Support Vector Machine. The second part is guide tracking which identifies the most likely guide by matching the histogram of color. In the third part, we determine the relationship between the guide and the automatic vehicle and compute the distance and direction of the guide with the known height of the camera. The fourth part is the automatic vehicle control, based on the relative relationship between guide and automatic vehicle we determine automatic vehicle’s speed and direction. The controlled automatic vehicle keeps the guide two meters ahead in order to spare a buffer that allows the communication and avoid collision of the guide and the vehicle. The automatic vehicle in our study is Wu′s Tech electric wheelchair Mambo 513; using a Logitech’s QuickCam ® Sphere AF PTZ color camera detects in front of the environment. On Intel® Core™ i7-3740QM 2.70GHz PC 4GB RAM and executed up to 6-15 per second processing speed of the frame. Using a variety of environmental conditions in 3580 frames, the correction rate of the guide detection is 88%, the correction rate of guide tracing is 91%, and the error rate for distance estimation between guide and automatic vehicle is 12%. In this study, we get the trajectory reproduce goal by only one PTZ camera.
显示于类别:	[資訊工程研究所] 博碩士論文

文件中的档案:

档案	描述	大小	格式	浏览次数
index.html		0Kb	HTML	400	检视/开启

在NCUIR中所有的数据项都受到原著作权保护.

社群 sharing

数据加载中.....