A Single-Chip Stereo Vision System For Motion Parameter Measurement Of High-Speed Ball

Detecting and tracking high-speed ball by computer vision is a research hotspot of sports electronic systems in recent years.In order to meet the needs of tracking and detecting high-speed ball,the systems requires characteristic that having a high framerate image sensor for capturing a trajectory without motion blur,having a high-speed image acquisition and processing capabilities to improve real-time performance,having the ability to reconstruct the three-dimensional pose to calculate the motion parameters of the ball.In this study,a single-chip binocular vision measurement system was designed and implemented in view of the above requirements.This paper using ZYNQ as main-control chip,which has a hybrid architecture contains FPGA(Field-Programmable Gate Array)and ARM(Acorn RISC Machine).The logic part of ZYNQ is responsible for collecting the LVDS(Low-Voltage Differential Signaling)signal,ARM is used to track and measure the motion parameters of the target.The binocular stereo cameras were calibrated by volume template in this paper,when the group of 3-D target points coordinates and their corresponding image coordinates are known,they can be used in a linear algorithm to work out the transform matrix between image coordinate system and the world coordinate system.Based on the feature of the ball,the paper proposed a method to locate position and detect the motion state of the ball with p rate threshold.According to the motion characteristics of the high-speed ball,the intelligent ROI(Region of Interest)ball tracking algorithm based on Kalman filter is designed.Due to the high frame rate and the low image resolution in the actual work,especially the short exposure time,caused the less of target pixel and darker image,this paper designed a stripe mark pattern which is easy to be identified,and a fast,adaptive,Chan-Vese model based image segmentation algorithm was also proposed in order to detect and extract pattern feature.The experimental results show that the proposed algorithm and surface markers can be more accurate to identify the rotation of the sphere.Main works in this paper are concluded:1)In this paper,a new stripe mark is used to measure the ball’s spin,which is based on the motion characteristics of the ball and the distribution of the spherical pattern and the shooting interval of the high-speed binocular camera.2)The adaptive image segmentation algorithm based on Chan-Vese model is designed and implemented to get more precise spherical tags extraction effect.3)An intelligent region of interest ROI Kalman filter sphere tracking algorithm is proposed to improve the real-time performance and the reliability in practical application.4)A simple and easy calibration process is proposed.Compared with the calibration of Zhang,the calibration process is simplified on the basis of ensuring the accuracy.Only one calibration plate is placed in the whole process,which solves the difference of calibration caused by the display plate.5)This Paper collecting LVDS signal through the HDMI interface by FPGA,with no professional-grade image codec board,improve the efficiency of the collection and transmission system.The experimental results indicate that the overall processing time from image acquisition to generating motion parameters is less than 300 ms,achieve the goal of real-time motion parameters measurement of high-speed ball.