Purpose  In the industry, can reduce manpower and improve efficiency  can contribute to the restoration of some functions lost by handicapped people  medical assistance  Do not need any one to help in 24 hours

Motivation  The handicapped people doesn’t accept an assistive arm onto the wheelchair, because they feel like being assimilated to a robot.  Manus® arm Handicapped people takes the control of the mobile robot and move the platform wherever he wants and can also use the manipulator.

GLOBAL DESCRIPTION

 It is made of two parts : ○ 1. the wheelchair equipped with a standard PC and an interface screen and a joystick for control. ○ 2. the mobile robot equipped with two omnidirectional vision sensors and a Manus® arm.

 This robotic assistance will be used in: ○ Automatic mode: track and follow a wheelchair autonomously without holding up the wheelchair. ○ Remote mode: Standing close to the wheelchair in order to be manually controlled. (this mode the arm can be used )

THE OMNIDIRECTIONAL VISION

 What is omnidirectional vision ? A : Use only one camera, with a rotation motion, in order to sweep a large space. Disadvantage : the camera’s movement takes time.  extend two classes of omnidirectional vision systems : 1. the systems made of a mirror and a camera are called 『 catadioptric systems 』.

2. a classical camera with a fish-eye lens, such mountings are called 『 dioptric systems 』.

THE AUTOMATIC TRACKING

 The main goal is to enable the mobile robot to track a wheelchair with no mark-up and no particular equipment onto the wheelchair.  The omnidirectional vision is used in two different ways : ○ 1. The free space map ○ 2. The tracking

The free space map  the two omnidirectional sensors are used as goniometric sensors and compute distances of the different verticals parts called natural landmarks

Correspondences between angles from the left sensor to the right sensor for objects situated at different distances from the mobile robot.

The tracking  As the wheelchair is not equipped with any particular mark-up, we have to track it as it is, so we use the CAMShift algorithm which perform a tracking by using an image of the object to track.  What is CAMShift? A: Continuously Adaptive Mean Shift

CAMShift  Architecture ○ CAMShift 是 Mean Shift 的改良版 ○ CAMShift 會更新每次搜索的範圍, 更適合在連續 的影像中作物件追蹤  Step ○ 轉換 HSV 色彩空間 ○ Back Projection ○ Mean Shift ○ CAMShift

CAMShift ’s Step  轉換成 HSV 色彩空間 ○ 取得色彩直方圖

CAMShift ’s Step  Back Projection

CAMShift ’s Step  Mean Shift ○ 計算中心點 ○ 旋轉的角度

CAMShift ’s step  CAMShift ○ 將新得到的搜尋視窗，當 作第三步驟的初始搜尋視 窗。如此重複的執行，就 可以達到連續追蹤物件的 效果。

CAMShift ’s defect  背景過度複雜或是 有類似顏色的時候 效果會降低。  物體本身就很複雜， 導致辨識困難或誤 判。

The tracking _ Initialization of the CAMShift  In order to init the CAMShift, we need images of the wheelchair. ○ compute the subtraction between the background and the current image.

RESULTS

 CAMShift : PGw&feature=related  Manus® arm : PCcY