基于改进ORB特征检测的乡村景区全景标识系统视频拼接研究  

Research on video Mosaic of rural scenic panorama identification system based on improved ORB feature detection

作　　者：刘晓华[1] 李航 LIU Xiaohua;LI Hang(Yan’an University,Yan’an Shaanxi 716000,China)

机构地区：[1]延安大学,陕西延安716000

出　　处：《自动化与仪器仪表》2024年第10期10-13,共4页Automation & Instrumentation

基　　金：陕西省科技厅一般项目课题《基于数字技术的延安红色工业遗产保护与旅游利用研究》(2022GY-346);延安大学“研究阐释党的二十大精神”专题项目(2022YDZX20)。

摘　　要：为了提高乡村景区全景标识系统视频拼接的效果,结合Hessian矩阵法和灰度质心法,提出一种改进ORB特征检测的视频拼接方法。该方法一方面使用Hessian矩阵法替换原有ORB(Oriented FAST and Rotated BRIEF)算法中的FAST特征提取方法,并在此基础上使用灰度质心法对Hessian检测出的特征点确定方向,使改进后的ORB算法在保留运行效率和旋转不变性的前提下,具有尺度不变性;另一方面,针对改进ORB算法完成特征提取与匹配后的图像,在进行视频拼接时,通动态规划算法寻找最佳缝合线,并利用泊松融合算法使拼接视频更平滑。仿真结果表明,相较于原始ORB算法,在相近的运行时间内,改进ORB算法所提取的特征点误差有了明显下降,检测精度更高。相较于SURF算法和SIFT算法,改进ORB算法特征点提取误差相对较低,检测速度大幅提升,总体性能更好。采用改进ORB特征检测方法拼接的视频图像,相较于多分辨样条法、加权平均法等经典图像融合算法,无论是在空间频率、互信息量、信息熵、平均梯度还是相关度上评价结果都更优越,在解决拼接裂痕或拖尾现象等常见视频拼接的情况下,图像融合自然,更满足乡村旧址文化展示与乡村旅游文化传播需求。In order to improve the video splice effect of rural scenic panoramic sign system,a video splice method with improved ORB feature detection was proposed by combining Hessian matrix method and gray-scale prime center method.On the one hand,the Hessian matrix method is used to replace the FAST feature extraction method in the original ORB algorithm,and on the basis of this,the gray prime-center method is used to determine the direction of the feature points detected by Hessian,so that the improved ORB algorithm has scale invariance while maintaining the operating efficiency and rotation invariance.On the other hand,for the image after feature extraction and matching with the improved ORB algorithm,the dynamic programming algorithm is used to find the best suture line during video stitching,and Poisson fusion algorithm is used to make video stitching smoother.The simulation results show that compared with the original ORB algorithm,the error of feature points extracted by the improved ORB algorithm is significantly reduced and the detection accuracy is higher during the similar running time.Compared with SURF algorithm and SIFT algorithm,the improved ORB algorithm has relatively low error in feature point extraction,greatly improved detection speed,and better overall performance.Compared with classical image fusion algorithms such as multi-resolution spline method and weighted average method,video images splicing with improved ORB feature detection method have better evaluation results in terms of spatial frequency,mutual information,information entropy,average gradient and relevance.In the case of solving common video splicing phenomena such as splicing cracks or trailing,image fusion is natural.More to meet the needs of rural old site cultural display and rural tourism cultural communication.

关 键 词：乡村旧址 全景标识 特征检测 ORB算法 旅游文化传播 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]