Research On 3D Fingertip Detection For Intelligent Wheelchair

With the aging of population and the number of people with disabilities increasing,there is an increasing demand for maneuvering tools,intelligent wheelchair has become a hot spot for social and research institutions.As a representative intelligent mobile service robot,intelligent wheelchair has the function of rocker control,limb control,navigation,obstacle avoidance,rehabilitation and so on.As the traditional wheelchair interaction,rocker control has a good control performance,but for the physically impulsive person or Myasthenia gravis,the joystick control method has been greatly constrained.In order to control wheelchair conveniently and naturally,this paper presents a contactless interaction method using three-dimensional fingertips information to control wheelchair.In this paper,three-dimensional fingertips information is used for intelligent wheelchair humancomputer interaction system to control wheelchair without touch based on the identification of the fingertips.First,the hand area is detected that using the priori spatial range information combined with skin color characteristics in the point cloud.After the growth of the color area,the fast component detection is realized by using the RGB component mean of the point cloud cluster.Then,the three-dimensional convex hull algorithm is used to extract the contours of the hand,and the candidate fingertips are generated.On the basis of the candidate fingertips,the fingertips are extracted in real time by three-dimensional K-curvature.After that,the fingertip information is used for intelligent wheelchair interactive control to achieve the mode settings,motion control and other major functions.In this paper,we design two sets of experiments: First,verify the accuracy and robustness of fingertip extraction by comparing the traditional method;second,verify the fingertip control effect.The former test fingertips detection effect for different hand-shaped in different backgrounds,lighting and other conditions.The experimental results show that the commonly used hand-1 recognition rate can reach 94%,better than the traditional method.The latter captures the complete movement of the fingertip point in the control process,determines the gesture movement through the fingertip point motion trajectory,and sends the control instruction by the gesture action to realize the wheelchair real-time control.