摘要
以一款车辆识别软件开发为应用背景,重点介绍基于深度图像处理在车辆识别中的应用实例。针对目前普通相机所拍摄的车辆图像,在识别时产生计算量大,噪声干扰、鲁棒性差的问题,软件通过深度相机Kinect采集图像的深度数据及彩色数据,并完成对两种图像的滤波、感兴趣区域匹配融合、目标特征提取、特征匹配等功能,最终满足对获取的图形图像中的车辆快速识别的功能。
With a vehicle identification software development as the application background, the paper focuses on the application of depth-image processing in vehicle identification. In the identification of the ordinary camera images of vehicles, there is large amount of calculation, noise and poor robustness. The software captures the depth image and color image through the Kinect camera and completes filtering, matching and fusion of region of interest, target feature extraction, feature matching and other function, and ultimately meets the function of rapid identification of vehicles in image obtained.
引文
[1]余涛.Kinect应用开发实战[M].北京:机械工业出版社,2012.
[2]冯伟兴,唐墨,贺波,等.Visual C++数字图像模式识别技术相机[M].北京:机械工业出版社,2010.
[3]刘瑞祯,于仕琪.Open CV教程——基础篇[M].北京:航空航天大学出版社,2007.
[4]吴国斌,李斌,阎骥洲.Kinect人体交互开发实践[M].北京:人民邮电出版社,2013.
[5]余亚玲,张华,刘桂华,等.基于不确定度评价的Kinect深度图预处理[J].计算机应用,2016.
[6]J.R.Parker.Algorithms for Image Processing and Computer Vesion[M].2nd ed,Wiley Publishing,Inc,2012.
[7]Christian Nagel,Bill Evjen,Jay Glynn,et al.Professional C#5.0 and NET 4.5.1[M].9th ed,Wiley Publishing,Inc,2014.
[8]Luigi Gallo,Ernesto Damiani,Giuseppe De Pietro,et al.Comparative evaluation of methods for filtering Kinect depth data[J].Multimedia Tools&Applications,2014.