基于SURF特征的运动视频图像边缘跟踪方法研究

Edge Tracking Method of Motion Video Images Based on SURF Features

传统的运动视频图像边缘跟踪方法存在图像边缘特征匹配时间过长,实时性差的问题,提出并设计基于SURF特征的运动视频图像边缘跟踪方法.根据盒子滤波器大小的不同建立尺度空间,将原始图像转换为积分图像,利用积分图像达到加速的目的,使用构建Hessian矩阵检测图像边缘特征点,利用图像边缘特征点生成特征描述子,通过计算特征描述子之间的欧式距离完成图像之间的边缘匹配,实现图像边缘跟踪.测试结果表明:在测试数据集相同的情况下,与传统两种跟踪方法相比,设计的基于SURF的运动视频图像边缘跟踪方法特征匹配所耗时间极短,实时性更好,该方法适合应用在运动视频图像边缘跟踪中.

The traditional edge tracking method of moving video image has long matching time and poor realtime.A SURF feature is proposed and designed to alleviate these two problems.Depending on the size of box fil⁃ter to establish scale space,to convert the original image as the integral image,the integral image for the pur⁃pose of accelerating and the construction of the Hessian matrix were used to detect the image edge feature point,using the image edge feature points generated character description.By calculating the Euclidean distance be⁃tween feature descriptors,the edge matching between images was accomplished and the edge tracking was real⁃ized.The test results show that under the same test data set,compared with the traditional two tracking methods,the feature matching time of the designed SURF based motion video image edge tracking method is significantly shortened with better real-time performance.This method is suitable for motion video image edge tracking.