基于改进Sobel算子和动态区域的铁轨稳像算法  

Rail Video Stabilization Algorithm Based on Improved Sobel and Dynamic Region

作　　者：石泓[1] 任智姣 SHI Hong REN Zhi-jiao(College of Computer Science,Sichuan University,Chengdu 610065)

机构地区：[1]四川大学计算机学院,成都610065

出　　处：《现代计算机》2016年第18期3-5,9,共4页Modern Computer

基　　金：国家自然科学基金(No.61071162)

摘　　要：由于列车运行时产生的震动,导致拍摄的视频出现抖动现象。因铁轨具有特定直线角度和实际运行情况,在动态区域中用改进的Sobel算子结合LSD算法来提取直线。在仿射模型下构造特征三角形,利用最小二乘法计算全局运动矢量,最后用Kalman滤波得出运动补偿矢量,对视频帧进行运动补偿。实验表明,该方法在机载视频稳像上具有比较好的运行时间和最终效果。Due to the vibration generated by the train running, leading to the emergence of a video shot jitter phenomenon. Because the rail has a specific line angle and the actual situation, the linear is extracted by LSD in dynamic region with the improved Sobel operator. In the affine model, constructs the characteristic triangle, and calculates the global motion vector by the least square method. Finally, the motion compensation vector is obtained by the Kalman filter, and the motion compensation is performed on the video frame. Experimental results show that the proposed method has a better running time and the final effect on the vehicle video stabilization.

关 键 词：铁轨稳像 SOBEL LSD直线检测 KALMAN滤波 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]