现有的车载全景影像系统普遍存在低实时性和图像质量不佳的问题,针对其中消耗时间和硬件算力最多的视角转换进行重点优化,阐述一种面向车载异构平台的高性能视角转换算法。首先结合有效线段检测和角点检测设计一种改进的最优控制点检测算法,甄选出精确的角点坐标用于透视变换矩阵的求取;然后采用基于局部的双矩阵视角转换算法,得出效果优良的俯视图;最后基于开放运算语言在车载异构平台上实现整套算法。实验结果表明,该算法在硬件友好的基础上缩减了计算耗时,有效提升了图像转换质量和车载全景影像系统的实用性。
Abstract
The existing vehicle-mounted panoramic image system generally has the problems of low real-time performance and poor image quality.Focusing on the perspective conversion that consume the most time and hardware computing power,optimize and elaborate a high-performance perspective conversion algorithm for vehicle-mounted heterogeneous platforms.Firstly,combine effective line detection and corner detection,design an improved optimal control point detection algorithm,select the precise corner coordinates for the perspective transformation matrix.Then use the local-based dual-matrix perspective conversion algorithm to get the effect excellent top view.Finally,implement the high-performance perspective conversion algorithm on the vehicle-mounted heterogeneous platform based on OpenCL.The experiment results show that the algorithm is based on hardware-friendly,reducing computational time,and effectively improving the quality of image conversion and the practicability of the vehicle-mounted panoramic imaging system.
关键词
视角转换 /
有效线段检测 /
角点检测 /
透视变换矩阵 /
OpenCL
Key words
perspective conversion /
effective line detection /
corner detection /
perspective transformation matrix /
OpenCL
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] Gao J,Kim S J.Constructing image panoramas using dual-homography warping[C]//IEEE Conference on Computer Vision and Pattern Recognition,2011:97-105.
[2] Zhang F,Liu F.Parallax-tolerant Image Stitching[C]//IEEE Conference on Computer Vision and Pattern Recognition,2014:1354-1361.
[3] 谭兆一,陈白帆,苏苇,等.基于第一视角的车辆盲区视野重现[J].软件导刊,2020,19(9):5.
[4] Liao W S,Hsieh T J,Chang Y L.GPU Parallel Computing of Spherical Panorama Video Stitching[C]//IEEE International Conference on Parallel & Distributed Systems.IEEE,2012.
[5] Sanida T,Sideris A,Dasygenis M.A Heterogeneous Implementation of the Sobel Edge Detection Filter Using OpenCL[C]//2020 9th International Conference on Modern Circuits and Systems Technologies (MOCAST),2020.
[6] Huang K,Hu B,Chen L,et al.ADAS on COTS with OpenCL: A Case Study with Lane Detection[J].IEEE Transactions on Computers,2017,67(4):559-565.
[7] Lu X,Yao J,Li K,et al.CannyLines: A parameter-free line segment detector[C]//IEEE International Conference on Image Processing.IEEE,2015.
[8] Gioi R G V,Jérémie Jakubowicz,Morel J M,et al.LSD:A Fast Line Segment Detector with a False Detection Control[J].Pattern Analysis and Machine Intelligence,IEEE Transactions on,2010,32(4):722-732.
[9] Harris C G,Stephens M J.A combined corner and edge detector[C]//Alvey vision conference,1988.
[10] Shi J,Tomasi C.Good Features to Track[C]//IEEE Conference on Computer Vision & Pattern Recognition.IEEE,2000.
PDF(2460 KB)
