輪椅機器人視覺定位與伴隨者辨識

Abstract

在移動式服務機器人中，機器人與人之間的互動是很重要的一個議題，要達到機器人與人互動的功能，機器人必須要能夠偵測與辨識出環境中是否有所要互動之對象。在本論文中，我們提出結合多重感測器的資訊去估測出目前伴隨者的所在位置，首先視覺同時定位與建立地圖(visual simultaneous localization and map building, VSLAM)是利用Microsoft Kinect所提供的彩色與深度影像資訊做為依據並搭配加速穩健性特徵(Speeded Up Robust Feature, SURF)演算法與擴展卡曼濾波器(extended Kalman filter, EKF)去估測出機器人的位置。而基於Kinect所提供的彩色與深度影像資訊，利用SURF演算法辨識出伴隨者，並藉由VSLAM所得的機器人位置推估出伴隨者的所在位置。本論文以實驗結果成功顯示出我們的系統可以即時偵測出所要跟隨的伴隨者，也可精準定位出機器人與伴隨者的位置。

Human-robot interaction is an important issue for the mobile service robots. To provide such a task and service, the mobile robot need to detect and recognize people in the surroundings. In this paper, we propose to utilize the multisensory data fusion to estimate the human position. First, the visual simultaneous localization and map building (VSLAM) is achieved by using the Microsoft Kinect and inertial sensors with the speeded up robust features (SURF) algorithm and the extended Kalman filter (EKF) to estimate the robot location recursively. Accordantly, based on the observed color image with depth measured from the Kinect sensor, a SURF algorithm is used for identity recognition and the accompanist’s position is evaluated with the robot position which is provided by VSLAM. The experimental results demonstrated the performance of the system, which could be able to recognize immediately and estimate accurately the localization with accompanist.