DM8168 OpenCV嘗試與評估(編譯ARM版OpenCV)
交叉編譯opencv2.3.1,並在DM8168 cortex A8中執行圖像處理。
開發環境:
PC:ubuntu12.04LTS。Intel Core 2 Duo CPU [email protected];
目標平臺:SEED DVS DM8168 DVRRDK 02.00.XXXX
第一篇:OpenCV交叉編譯
參考了http://blog.csdn.net/embeddedman/article/details/7416934
1.安裝交叉編譯鏈
使用DVRRDK中的ti_tools/cgt_a8/arm-2009q1/bin/arm-none-linux-xxx
需將其增加環境變量:能夠加到用戶文件夾下的.bashrc或/etc/profile
格式例如以下:
export PATH=/...../bin:$PATH
2.獲得OpenCV源代碼,本文基於2.3.1
3.安裝cmake cmake-gui
4.建立編譯文件夾如/opencv231/build_arm
5.進入文件夾運行cmake-gui
選擇源碼文件夾:/opt/OpenCV-2.31
選擇Build文件夾:/opt/opencv2.3.1_forArm,大家依據自己的喜好設置編譯文件夾吧。
點擊Configure,保持generator為Unix Makefiles。選擇Specify options for cross-compiling,點擊Next
Operating System填寫arm-inux
C Compilers選擇DVRRDK_02.00.00.23/ti_tools/cgt_a8/arm-2009q1/bin/arm-none-linux-gnueabi-gcc
C++ Compilers填寫.......-g++
程序庫的Target Root填寫OpenCV安裝文件夾,然後點擊Finish
上面標註的表示opencv將要安裝的文件夾,默覺得/usr/local,為了便於管理,大家能夠安裝在自己設定的文件夾下,能夠改動,同一時候。網上資料說:另外,我沒有安裝tiff圖像的支持,因此去掉WITH_TIFF(出自:http://blog.csdn.net/eagelangel/article/details/7232364)。在此點擊Configure,然後點擊Gennerate就會生成Makefile。
進入/opt/opencv2.3.1_forArm文件夾。運行make命令,就會編譯opencv,編譯過程中遇到例如以下錯誤:
In file included from /opt/OpenCV-2.3.1/modules/flann/src/precomp.hpp:9,
from /opt/opencv2.3.1_forArm/modules/flann/opencv_flann_pch_dephelp.cxx:1:
/opt/OpenCV-2.3.1/modules/flann/include/opencv2/flann/dist.h: In function ‘T cvflann::abs(T) [with T = long double]‘:
/opt/OpenCV-2.3.1/modules/flann/include/opencv2/flann/dist.h:63: error: ‘fabsl‘ was not declared in this scope
make[2]: *** [modules/flann/CMakeFiles/opencv_flann_pch_dephelp.dir/opencv_flann_pch_dephelp.obj] Error 1
make[1]: *** [modules/flann/CMakeFiles/opencv_flann_pch_dephelp.dir/all] Error 2
make: *** [all] Error 2
解決方法:改動OpenCV-2.3.1/modules/flann/include/opencv2/flann/dist.h文件第63行的源代碼:將absl()改動為abs()
假設編譯過程遇到例如以下錯誤:
Linking CXX executable ../../bin/opencv_test_calib3d
../../lib/libopencv_core.so: undefined reference to `pthread_key_create‘
../../lib/libopencv_core.so: undefined reference to `pthread_getspecific‘
../../lib/libopencv_ts.so: undefined reference to `pthread_key_delete‘
../../lib/libopencv_core.so: undefined reference to `pthread_once‘
../../lib/libopencv_core.so: undefined reference to `clock_gettime‘
../../lib/libopencv_core.so: undefined reference to `pthread_setspecific‘
collect2: ld returned 1 exit status
make[2]: *** [bin/opencv_test_calib3d] Error 1
make[1]: *** [modules/calib3d/CMakeFiles/opencv_test_calib3d.dir/all] Error 2
make: *** [all] Error 2
解決方式:改動/opt/opencv2.3.1文件夾下的CMakeCache.txt。CMAKE_EXE_LINKER_FLAGS原來為空。加上-lpthread -lrt,假設沒有自己加上。
註意:每次錯誤產生。經改動後。僅僅要再次運行make命令就接著編譯,編譯成功後,運行make intall命令就會安裝opencv。
因為本人的安裝文件夾為:/opt/opencv2.3.1_arm,大家能夠看到生成了bin、lib、include等文件夾,例如以下:
第二篇
執行
1.將以上文件夾lib include是必須的拷貝的板子上。文件夾能夠隨意,也能夠放到系統動態庫文件夾。關於這兩點能夠參考:
http://blog.chinaunix.net/uid-23215128-id-2521319.html
一個演示樣例代碼:
#include "cv.h" #include "highgui.h" int main( int argc, char** argv ) { IplImage* pImg; //聲明IplImage指針 //加載圖像。強制轉化為Gray if( argc == 3 && (pImg = cvLoadImage( argv[1], 0)) != 0 ) { IplImage* pImg2 = cvCreateImage(cvGetSize(pImg), pImg->depth, pImg->nChannels); cvCopy(pImg, pImg2, NULL); cvSaveImage(argv[2], pImg2);//把圖像寫入文件 cvReleaseImage( &pImg ); //釋放圖像 cvReleaseImage( &pImg2 ); //釋放圖像 return 0; } return -1; }
採用下面命令方式在pc上進行編譯,當然編譯時所連接的庫是arm版的:
arm-none-linux-gnueabi-g++ test.cpp -o main1 -I/usr/local/arm/opencv/include/opencv -I/usr/local/arm/opencv/include -L/usr/local/arm/opencv/lib -lpthread -ldl -lrt -lopencv_core -lopencv_highgui
(文件夾可能有所不同)
假設PC中 arm的庫的路徑與lib在板子上的路徑不一致的話,運行程序會提示找不到庫,這時能夠通過板卡設置export LD_LIBRARY_PATH= 使程序能夠搜索到庫。
另外最好將板子上opencvlib的文件夾見一個軟連接到/usr/lib.
2.運行
./
結論:
終於執行結果:cvSplit cvMin cvCreateImage等等的操作,大概是臺式機PC的1/40的速度。
計劃:與DSP聯合運算。
DM8168 OpenCV嘗試與評估(編譯ARM版OpenCV)