Xiaopeng Chen

Dynamic model based ball trajectory prediction for a robot ping-pong player

In this paper, a novel ball trajectory prediction method is proposed for a robot ping-pong player. The method is based on the dynamic model in the sky and the bouncing dynamic model between the ball and the table. First, the ambiguity of the dynamic model in the sky is removed. Then, the bouncing model between the ball and the table is proposed to get the relationship of the emergence speed and the incidence speed. The corresponding coefficients of models are trained according to the 3D Cartesian coordinates of the ball calculated from a high speed stereo vision system. With known dynamic models, the trajectory can be predicted if several initial trajectory points of the ball from the vision system are given. Experiments show that the trajectory is able to be predicted accurately in several milliseconds. The result of our method is better than that by LWR based approach.

Rapid and precise object detection based on color histograms and adaptive bandwidth mean shift

Speed and precision are important for object detection algorithms. In this paper, a novel object detection algorithm based on color histogram and adaptive bandwidth mean shift is proposed. The algorithm is capable of detecting objects rapidly and precisely. It is composed of two stages: a rough detection stage and a precise detection stage. At the rough detection stage, histogram back projection and thresholding are applied to fast object identification and rough global localization. At the precise detection stage, the precise position, size and orientation are derived under the adaptive bandwidth mean shift framework. Experiments verify that the algorithm is able to detect the size, position and orientation of general objects rapidly and precisely.

Ping-pong trajectory perception and prediction by a PC based High speed four-camera vision system

A high speed vision system is vital for a robot player to play table tennis successfully. The three dimensional ping-pong trajectory should be perceived in high precision at a high frame rate. Whats more, with the beginning parts of the trajectory, the following trajectory should be predicted to save enough time for successful planning and control.

Bouncing model for the table tennis trajectory prediction and the strategy of hitting the ball

The bouncing model plays an important role in the trajectory prediction process when the humanoid robots play table tennis. A nonlinear bouncing model based on Momentum Theorem and Moment of Momentum Theorem considering spin of the ball is established to describe the collision between the ball and the table. With the constructed bouncing model, the hitting strategy is discussed to hit the ball the way we expect. The experiments show that with the existing aerodynamic model, we can predict the table tennis trajectory within the error range we expect.

A novel artificial landmark for monocular global visual localization of indoor robots

To obtain the pose of a monocular camera installed on the robot relative to the static artificial landmarks in a prepared room is one of many good methods for indoor robot localization. However, the point correspondence problem caused by rotation of cameras and the point detection precision problem are not paid enough attention to. In this paper, we proposed a novel coplanar landmark and a method to calculate the camera pose. The landmark is composed of two carefully placed polygons to contain point correspondence information. In this way, the camera can rotate arbitrarily relative to the landmark. Whats more, there are many parallel line pairs and perpendicular line pairs to impose more constraints on the camera pose. In our method, the SUSAN detector is used for edge detection. The line segments of the two polygons are then detected by RANSAC algorithm. After sorting the lines, the corners of each polygon are extracted by intersection of neighboring lines instead of detecting corners directly to improve precision. Once the intrinsic camera parameters and the coordinates of line segments and corners in the world frame and the image frame are derived, the camera pose can be obtained easily from the homography between the world frame and the image frame. Through the analysis of the experimental data, it can be concluded that the robustness and accuracy are much improved, and the rate is high enough for real-time indoor robot localization.

Adaptive Bandwidth Mean Shift Object Tracking

In this paper, a novel adaptive bandwidth mean shift algorithm toward 2D object tracking is proposed. It can simultaneously tracks the scale and orientation besides position in real time. The feature histogram weighted by a kernel with adaptive bandwidth is used for representing the target and the candidate target. The similarity of the target model and the candidate model is measured by the Bhattacharyya coefficient. A two step method is used iteratively to find the most probable target position, scale and orientation. The first step is to find the position using a mean shift iteration, the second step is to find the bandwidth which best describes the region of the object. Its convergence is proved theoretically. Experiments show that it can successfully track the position, scale and orientation in real time.

System design of an Anthropomorphic arm robot for dynamic interaction task

This paper investigates the system design of a 7-DOFs Anthropomorphic arm robot for fast dynamic interaction task, taking the ping-pong rally task against human as an example. The robot system includes the arm mechanism, real-time stereo vision sub-system, ball trajectory prediction, racket motion trajectory planning, and distributed joint controllers. The algorithms for ball identification, ball trajectory prediction, and hitting planning are presented. The validity of the system is demonstrated via rally experiments.

Adaptive Bandwidth Mean Shift Object Detection

In this paper, a novel adaptive bandwidth mean shift algorithm toward 2D object detection (ABMSOD) is proposed. It can not only identify whether an object of certain classes exists or not, but also get the scale and orientation besides position very fast. The feature histogram weighted by a kernel with adaptive bandwidth is used for representing the target object model and the candidate object model. Features such as color, texture, gradient and so on can be used. A single piece of image is enough to build a model by calculating the weighted feature histogram of the object in the image. There is no exhaustive training. The similarity of the target model and the candidate model is measured by the Bhattacharyya coefficient. After gathering the models of targets, the algorithm can be used for object detection. In the first step, the algorithm searches the whole image to find the rough positions of possible candidate objects. If the similarities are all below a certain threshold, it reports no object existence. If the similarities are above the threshold, the second step or the adaptive bandwidth mean shift search step is executed to find the best position, orientation and scale of these objects. Experiments show that it successfully detects the position, scale and orientation of objects.

Vision guided path planning system for vehicles using infrared landmark

Infrared-landmark based vision navigation system is proposed to reliably guide drivers from starting point to goal following a feasible shortest path in a known indoor or outdoor environment no matter whether it is day or night. Whats more, this system is capable of warning drivers about the dangerous zones “like no right turn, no U-turn, damaged road ahead etc” conditions when no visible light is available for human beings, mostly at night. The system consists of Basler camera, designed Infra-Red landmark, our developed software, IR filter and an industrial computer. Each infrared landmark contains a unique ID. Address information and important notes are attached to each ID. Once a landmark is efficiently detected and its ID is recognized, location information related to the corresponding ID is presented. And then Dijkstras algorithm is efficiently utilized to find a possible shortest path to destination. After that, path is followed with the aid of vision system. Experimental results demonstrate suitability, effectiveness and feasibility in intelligent transportation system.

Kinematic analysis and solution of the natural posture of a 7DOF humanoid manipulator

Biped walking control and motion planning of humanoid robots have been researched for a long time and have made great progress. In order to serve people, the humanoids arm should appear natural, have large working space and be easy to avoid obstacles In this paper a 7DOF manipulator with the same joint configuration as humans arm and its planning algorithm is proposed. The arm imitates the posture of humans arm to give the optimal solution in the joint space to make our manipulator look natural. The redundancy is solved and the manipulators natural posture is derived conveniently and easily. Experiments show the feasibility of the method.