背景:
最近,在Nvidia的GPU嵌入式开发板Jetson TX1(简称TX1)上移植深度学习目标检测算法YOLO。在TX1上安装了官方提供的opencv版本——OpenCV4Tegra(OpenCV-2.4.12),但是使用该版本opencv VideoCapture在读取摄像头数据时出错,显示错误:
Unable to stop the stream.: Device or resource busy
Unable to stop the stream.: Bad file descriptor
使用Google试图找到解决方法,最后发现官方给出答案,只能重新安装Opencv3.0以上的 版本。我使用Nvidia论坛jkjung提供的方法安装OpenCV3.1.0。安装完成后,我使用的TX1 Ubuntu14.04上安装了两个版本的OpenCV, Opencv-2.4.12和Opencv-3.1.0。因此,我面临着如何在程序中指定使用期望版本的OpenCV,即Ubuntu下多个版本Opencv管理的问题。
解决方案:
现在讨论Ubuntu14.04下分别使用CMake和Makefile编译运行带Opencv程序的两种解决方案。
1. CMake下指定Opencv版本
关键文件:OpenCVConfig.cmake。在opencv编译好后,所在目录中一般会有一个叫OpenCVConfig.cmake的文件,这个文件中指定了CMake要去哪里找OpenCV,其.h文件在哪里等,比如其中一行:
# Provide the include directories to the caller set(OpenCV_INCLUDE_DIRS "/home/ubuntu/src/opencv-3.1.0/build" "/home/ubuntu/src/opencv-3.1.0/include" "/home/ubuntu/src/opencv-3.1.0/include/opencv")
只要让CMake找到这个文件,这个文件就指定了Opencv的所有路径,因此设置OpenCV_DIR为包含OpenCVConfig.cmake的目录,如在我的C++工程CMakeLists.txt中添加
set(OpenCV_DIR "/home/ubuntu/src/opencv-3.1.0/build")
我的OpenCVConfig.cmake在
/home/ubuntu/src/opencv-3.1.0/build
注意,将其添加在project(MyProjectName)之前。
如CMakeLists.txt内容如下:
cmake_minimum_required(VERSION 2.8) set(OpenCV_DIR "/home/ubuntu/src/opencv-3.1.0/build") project( camera ) find_package( OpenCV REQUIRED ) add_executable( camera camera.cpp ) target_link_libraries( camera ${OpenCV_LIBS} )
因此,我们期望使用哪个版本的Opencv,只要找到对应的OpenCVConfig.cmake文件,并且将其路径添加到工程的CMakeLists.txt中即可了。
2. Makefile下指定Opencv版本
关键文件:opencv.pc 。在Makefile下,应该是可以在其中详细设定Opencv路径,要使用到哪些库等等(我没有仔细学习过Makefile的使用)。在此,我想介绍一种简单的在Makefile下指定期望使用的Opencv版本的方法,该方法借助linux下pkg-config命令。
我们常常使用pkg-config --modversion 来查看指定库的版本,比如查看opencv版本pkg-config --modversion opencv。其实pkg-config显示的信息来自于这个库对应的.pc文件,比如安装了opencv后,我们可以在/usr/lib/pkg-config/ 文件夹下找到opencv.pc,内容如下:
# Package Information for pkg-config prefix=/usr exec_prefix=${prefix} libdir=${exec_prefix}/lib includedir_old=${prefix}/include/opencv includedir_new=${prefix}/include Name: OpenCV Description: Open Source Computer Vision Library Version: 2.4.12.2 Libs: -L${exec_prefix}/lib -lopencv_calib3d -lopencv_contrib -lopencv_core -lopencv_features2d -lopencv_flann -lopencv_gpu -lopencv_highgui -lopencv_imgproc -lopencv_legacy -lopencv_ml -lopencv_objdetect -lopencv_photo -lopencv_stitching -lopencv_superres -lopencv_ts -lopencv_video -lopencv_videostab -lopencv_detection_based_tracker -lopencv_esm_panorama -lopencv_facedetect -lopencv_imuvstab -lopencv_tegra -lopencv_vstab -L/usr/local/cuda-7.0/targets/armv7-linux-gnueabihf/lib -lcufft -lnpps -lnppi -lnppc -lcudart -ltbb -lrt -lpthread -lm -ldl Cflags: -I${includedir_old} -I${includedir_new}
当然,我们也可以使用
pkg-config --cflags opencv
pkg-config --libs opencv
分别查看opencv库的包含目录(include/, .h)和库名称(.so)。在Makefile中,可以通过以上命令设定opencv库相关目录。如下:
COMMON+= -DOPENCV CFLAGS+= -DOPENCV LDFLAGS+= `pkg-config --libs opencv` COMMON+= `pkg-config --cflags opencv`
此时Makefile对应的工程使用的opencv库即为opencv.pc文件中所指的版本。本例中为2.4.12.2。
那么可以容易想到,如果我在/usr/lib/pkgconfig文件夹目录下添加opencv-3.1.0.pc文件,是否就可以将版本指向opencv-3.1.0l了呢?答案是肯定的。
我在编译opencv-3.1.0的编译目录下/home/ubuntu/src/opencv-3.1.0/build/目录下找到了unix-install/opencv.pc文件,其内容如下:
# Package Information for pkg-config prefix=/home/ubuntu/opencv-3.1.0 exec_prefix=${prefix} libdir=${exec_prefix}/lib includedir_old=${prefix}/include/opencv includedir_new=${prefix}/include Name: OpenCV Description: Open Source Computer Vision Library Version: 3.1.0 Libs: -L${exec_prefix}/lib -lopencv_cudabgsegm -lopencv_cudaobjdetect -lopencv_cudastereo -lopencv_shape -lopencv_stitching -lopencv_cudafeatures2d -lopencv_superres -lopencv_cudacodec -lopencv_videostab -lopencv_cudaoptflow -lopencv_cudalegacy -lopencv_calib3d -lopencv_features2d -lopencv_objdetect -lopencv_highgui -lopencv_videoio -lopencv_photo -lopencv_imgcodecs -lopencv_cudawarping -lopencv_cudaimgproc -lopencv_cudafilters -lopencv_video -lopencv_ml -lopencv_imgproc -lopencv_flann -lopencv_cudaarithm -lopencv_core -lopencv_cudev
我发现其中的路径是不太准确的,更正为:
# Package Information for pkg-config prefix=/home/ubuntu/src/opencv-3.1.0 exec_prefix=${prefix}/build libdir=${exec_prefix}/lib includedir_old=${prefix}/include/opencv includedir_new=${prefix}/include Name: OpenCV Description: Open Source Computer Vision Library Version: 3.1.0 Libs: -L${exec_prefix}/lib -lopencv_cudabgsegm -lopencv_cudaobjdetect -lopencv_cudastereo -lopencv_shape -lopencv_stitching -lopencv_cudafeatures2d -lopencv_superres -lopencv_cudacodec -lopencv_videostab -lopencv_cudaoptflow -lopencv_cudalegacy -lopencv_calib3d -lopencv_features2d -lopencv_objdetect -lopencv_highgui -lopencv_videoio -lopencv_photo -lopencv_imgcodecs -lopencv_cudawarping -lopencv_cudaimgproc -lopencv_cudafilters -lopencv_video -lopencv_ml -lopencv_imgproc -lopencv_flann -lopencv_cudaarithm -lopencv_core -lopencv_cudev
并将其拷贝到/usr/lib/pkgconfig目录下,并重命名为opencv-3.1.0.pc,将原来的opencv.pc重命名为opencv-2.4.12.pc。
那么当我期望使用opencv-3.1.0时,Makefile中为:
COMMON+= -DOPENCV CFLAGS+= -DOPENCV LDFLAGS+= `pkg-config --libs opencv-3.1.0` COMMON+= `pkg-config --cflags opencv-3.1.0`
当我期望使用opencv-2.4.12时,Makefile中为:
COMMON+= -DOPENCV CFLAGS+= -DOPENCV LDFLAGS+= `pkg-config --libs opencv-2.4.12` COMMON+= `pkg-config --cflags opencv-2.4.12`
从而实现了多个版本OpenCV的使用。
备注:
运行make之后,运行可执行程序时,可能会提示找不到库的错误(not find libopencv-core.so.3.1之类),那时因为动态库目录不在程序搜索库目录的路径下,系统默认的动态库搜索路径有/usr/local/lib等,OpenCV4Tegra(opencv-2.4.12)生成的库都在/usr/local/lib目录,但opencv-3.1.0的库在/home/ubuntu/src/opencv-3.1.0/build/lib中,不在搜索路径中,因此,只需要添加其为搜索路径即可。添加方法为/etc/ld.so.conf.d/libc.conf中添加/home/ubuntu/src/opencv-3.1.0/build/lib。libc.conf文件内容如下:
# libc default configuration /usr/local/lib /home/ubuntu/src/opencv-3.1.0/build/lib