基于景深和LK光流法的视频测速算法  

Video velocity measurement algorithm based on depth of field and LK optical flow

作　　者：周登浩 武斌 ZHOU Denghao;WU Bin(School of Computer and Information Engineering,Tianjin Chengjian University,Tianjin 300384,China)

机构地区：[1]天津城建大学计算机与信息工程学院,天津300384

出　　处：《传感器与微系统》2021年第8期116-120,共5页Transducer and Microsystem Technologies

基　　金：天津市科技计划项目(18JCTPJC60000)。

摘　　要：根据景深图虚化背景影响的思想,提出一种基于景深和LK光流法的视频测速算法。通过多尺度自适应小波算法沿水平、垂直及对角线4个方向对图像进行分解;引入边缘图和李氏指数,优化由显著特征区域高频小波系数与深度值关系生成的景深图;最后用LK光流法对景深图进行ORB角点检测,并结合SUSAN算法剔除异常角点绘制角点分布图,由角点坐标在相邻帧图像中的位移计算目标的运动速度和角度。实验结果证明:该算法能有效抑制背景影响、减少计算量,提高运动目标在复杂背景与动态背景下的测速精度和实时性。According to the idea of depth of field map blurs background,a video speed measurement algorithm based on depth of field and LK optical flow method is proposed.The multi-scale adaptive wavelet algorithm is used to decompose the image in four directions:horizontal,vertical and diagonal;the edge map and the Li’s index are introduced to optimize the depth of field map generated by the relationship between the high-frequency wavelet coefficients and the depth values of the salient feature regions;The LK optical flow method is used to detect the ORB corner point of the depth of field map,and the SUSAN algorithm is used to eliminate the abnormal corner points and draw the corner point distribution map.The movement speed and angle of the target are calculated from the displacement of the corner point coordinates in the adjacent frame image.The experimental results show that the algorithm can effectively suppress background influence,reduce the amount of calculation,and improve the precision and real-time performance of moving targets in complex background and dynamic background.

关 键 词：视频测速 景深图 LK光流法 ORB角点检测 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]