基于DM642的金字塔Lucas-Kanade光流法计算速度信息

通过垂直拍摄地面的摄像头连续抓拍两帧图像,从中计算出横纵向速度信息.该设计以TMS320DM642为核心,结合TI的VLIB视频处理库,利用VLIB中的Harris角点提取算法对特征点进行提取,并利用金字塔Lucas-Kanade光流法实现对特征点的大位移跟踪,对跟踪出来的横纵向位移信息进行筛选并利用帧率与位移之间的关系计算出速度信息.然后,将该设计与在PC机上用OpenCV实现的金字塔Lucas-Kanade光流法和SURF特征点跟踪匹配法进行比较,其结果表明该设计简易可行且具有实时性好的优点.最后在此基础上简要介绍了此设计的应用前景并对设计进行了总结.

Speed Calculation with Pyramid Lucas-Kanade Optical Flow Based on DM642

By using a camera which is shooting the ground vertically to capture two continuous images,the horizontal speed and the vertical speed are calculated.TMS320DM642 is taken as the core,and then the vision library(VLIB) of Texas Instruments(TI) is utilized to realize complex algorithm.By using the Harris corner detection of VLIB,the feature points are extracted.And then,by utilizing Lucas-Kanade optical flow with image pyramid,tracking the feature points that have large displacement is realized.The data of horizontal and vertical displacement are filtered to get the true displacement of the pixels.Then the relationship between the frame rate and displacement is used to get the speed.The result of this design has been compared with the results of the method of Lucas-Kanade optical flow with image pyramid and the method of tracking SURF feature points using OpenCV in PC.Comparison shows that this design is feasible and has good real-time performance.Finally,the application prospect of this design is introduced and some conclusions about this design have been made.