码迷，mamicode.com
首页 > 移动开发 > 详细

我的Opencv4Android添加V4L2支持的移植记录

时间：2016-07-04 11:59:04      阅读：602      评论：0      收藏：0      [点我收藏+]

标签：

博主QQ:1356438802

引言

前面几篇文章我已经在Windows / Ubuntu / Android 三个平台成功安装了opencv的使用/编程环境，但是我需要在android平台使用usb video  camera（即UVC）。当然在Ubuntu上，默认编译出来的opencv库本身就支持V4L2（UVC的操作API函数集），而Windows是将UVC纳入了DirectShow框架，默认也是支持的；唯独对于Android平台，有且仅有对Android Native Camera的支持，即通常说的前摄像头、后摄像头。所以我接下来的工作就是将V4L2加入Opencv4Andorid！

PS1: 前面那篇ubuntu安装opencv的文章，是在Ubuntu虚拟机里面做的。这篇文章是在Ubuntu物理机上做的，所以工具和环境我又装了一遍。

PS2: 参考文章《OpenCV4Android编译http://blog.sina.com.cn/s/blog_602f87700102vdnw.html》《透析Android以JNI调OpenCV的三种方式(让OpenCVManager永不困扰)http://blog.csdn.net/yanzi1225627/article/details/27863615》
PS3: 本文章的最后编译的opencv4android sdk和几个例程app——http://download.csdn.net/detail/luoyouren/9566690

一、工具和环境变量安装

1. sudo apt-get install cmake

技术分享
技术分享技术分享

技术分享

2. sudo apt-get install ant

技术分享技术分享

技术分享
技术分享
1.8版本可能会出错！
再到http://ant.apache.org官网下载1.9.7版本安装到/opt目录，
修改/usr/bin/ant的软链接，指向/opt/apache-ant-1.9.7/bin/ant
技术分享
技术分享

3. 环境变量：sudo vim /etc/profile

技术分享
技术分享
source /etc/profile

二、直接编译

1. sh scripts/cmake_android_arm.sh

#!/bin/sh
cd `dirname $0`/..
 
mkdir -p build_android_arm
cd build_android_arm
 
cmake -DCMAKE_BUILD_WITH_INSTALL_RPATH=ON -DCMAKE_TOOLCHAIN_FILE=../android/android.toolchain.cmake $@ ../..
log_build_for_android_1.txt，和log_build_for_ubuntu.txt对比
技术分享技术分享

技术分享
技术分享
可以看出对于opencv for android，只提供AndroidNativeCamera的Video I/O，OpenCV modules多了androidcamera / java。

2. 编译

cd build_android_arm
make
make install
技术分享
技术分享
编译完成

技术分享
技术分享

生成最终文件
技术分享
技术分享
（注意我每修改编译一次都会将build_android_arm文件夹复制保存一份，以备之后要使用）SDK: build_android_arm_1

三、修改后编译

1. -DBUILD_EXAMPLES=1

--   Tests and samples:
--     Tests:                       YES
--     Performance tests:           YES
--     C/C++ Examples:              NO

把所有Example都编译出来
技术分享
技术分享

2. 打开对V4L2的支持

a. 查看CMakeLists.txt文件908行处
技术分享
技术分享
if(DEFINED WITH_V4L)
  if(HAVE_CAMV4L)
    set(HAVE_CAMV4L_STR "YES")
  else()
    set(HAVE_CAMV4L_STR "NO")
  endif()
  if(HAVE_CAMV4L2)
    set(HAVE_CAMV4L2_STR "YES")
  elseif(HAVE_VIDEOIO)
    set(HAVE_CAMV4L2_STR "YES(videoio)")
  else()
    set(HAVE_CAMV4L2_STR "NO")
  endif()
  status("    V4L/V4L2:"       HAVE_LIBV4L
             THEN "Using libv4l1 (ver ${ALIASOF_libv4l1_VERSION}) / libv4l2 (ver ${ALIASOF_libv4l2_VERSION})"
             ELSE "${HAVE_CAMV4L_STR}/${HAVE_CAMV4L2_STR}")
endif(DEFINED WITH_V4L)
如果要支持V4L2，必须有WITH_V4L和HAVE_CAMV4L2

b. 但是在CMakeLists.txt文件159行处
技术分享
技术分享
OCV_OPTION(WITH_V4L            "Include Video 4 Linux support"               ON   IF (UNIX AND NOT ANDROID) )
OCV_OPTION(WITH_LIBV4L         "Use libv4l for Video 4 Linux support"        ON   IF (UNIX AND NOT ANDROID) )
UNIX AND NOT ANDROID，意思是，非Android的unix平台才设置为ON

c. 在CMakeLists.txt文件473行处
技术分享
技术分享
cmake/OpenCVFindLibsVideo.cmake
技术分享
技术分享
在这里，已经把HAVE_LIBV4L / HAVE_CAMV4L / HAVE_CAMV4L2 / HAVE_VIDEOIO的设置全部clear

d. 注意必须写在同一行！
WITH_LIBV4L这个已经没有用了，不再设置它。
技术分享
技术分享

e. sh scripts/cmake_android_arm.sh  > log_build_for_android_2.txt
发现没什么变化
技术分享
技术分享

f. 继续修改
OCV_OPTION(WITH_V4L            "Include Video 4 Linux support"               ON   IF (UNIX AND NOT ANDROID) )
CMakeList.txt改成
OCV_OPTION(WITH_V4L            "Include Video 4 Linux support"               ON   IF (UNIX))

并且在cmake/OpenCVFindLibsVideo.cmake中增加打印信息
# --- V4L ---
#luoyouren-20160629
status("  before seting ...")
status("  WITH_V4L:" ${WITH_V4L})
status("  HAVE_LIBV4L:" ${HAVE_LIBV4L})
status("  HAVE_CAMV4L:" ${HAVE_CAMV4L})
status("  HAVE_CAMV4L2:" ${HAVE_CAMV4L2})
status("  HAVE_VIDEOIO:" ${HAVE_VIDEOIO})
ocv_clear_vars(HAVE_LIBV4L HAVE_CAMV4L HAVE_CAMV4L2 HAVE_VIDEOIO)
if(WITH_V4L)
  if(WITH_LIBV4L)
    CHECK_MODULE(libv4l1 HAVE_LIBV4L1)
    CHECK_MODULE(libv4l2 HAVE_LIBV4L2)
    if(HAVE_LIBV4L1 AND HAVE_LIBV4L2)
      set(HAVE_LIBV4L YES)
    else()
      set(HAVE_LIBV4L NO)
    endif()
  endif()
  CHECK_INCLUDE_FILE(linux/videodev.h HAVE_CAMV4L)
  CHECK_INCLUDE_FILE(linux/videodev2.h HAVE_CAMV4L2)
  CHECK_INCLUDE_FILE(sys/videoio.h HAVE_VIDEOIO)
status("  seting ...")
status("  WITH_V4L:" ${WITH_V4L})
status("  HAVE_LIBV4L:" ${HAVE_LIBV4L})
status("  HAVE_CAMV4L:" ${HAVE_CAMV4L})
status("  HAVE_CAMV4L2:" ${HAVE_CAMV4L2})
status("  HAVE_VIDEOIO:" ${HAVE_VIDEOIO})
endif(WITH_V4L)
#luoyouren-20160629
#set(WITH_V4L ON)
#set(HAVE_CAMV4L2 ON)
status("  after seting ...")
status("  WITH_V4L:" ${WITH_V4L})
status("  HAVE_LIBV4L:" ${HAVE_LIBV4L})
status("  HAVE_CAMV4L:" ${HAVE_CAMV4L})
status("  HAVE_CAMV4L2:" ${HAVE_CAMV4L2})
status("  HAVE_VIDEOIO:" ${HAVE_VIDEOIO})

结果显示
V4L/V4L2都没有打开技术分享
技术分享
技术分享
技术分享

before setting时HAVE_CAMV4L2=ON，但是setting后HAVE_CAMV4L2= 
下面三条语句，头文件存在与否决定了HAVE_CAMV4L、HAVE_CAMV4L2、HAVE_VIDEOIO的值
  CHECK_INCLUDE_FILE(linux/videodev.h HAVE_CAMV4L)
  CHECK_INCLUDE_FILE(linux/videodev2.h HAVE_CAMV4L2)
  CHECK_INCLUDE_FILE(sys/videoio.h HAVE_VIDEOIO)
linux/videodev2.h不存在
这个文件在NDK目录下
/opt/ndk/android-ndk-r9/platforms/android-8/arch-arm/usr/include/linux/videodev2.h

先不管这个问题，直接强制设置：
# --- V4L ---
#luoyouren-20160629
status("  before seting ...")
status("  WITH_V4L:" ${WITH_V4L})
status("  HAVE_LIBV4L:" ${HAVE_LIBV4L})
status("  HAVE_CAMV4L:" ${HAVE_CAMV4L})
status("  HAVE_CAMV4L2:" ${HAVE_CAMV4L2})
status("  HAVE_VIDEOIO:" ${HAVE_VIDEOIO})
ocv_clear_vars(HAVE_LIBV4L HAVE_CAMV4L HAVE_CAMV4L2 HAVE_VIDEOIO)
if(WITH_V4L)
  if(WITH_LIBV4L)
    CHECK_MODULE(libv4l1 HAVE_LIBV4L1)
    CHECK_MODULE(libv4l2 HAVE_LIBV4L2)
    if(HAVE_LIBV4L1 AND HAVE_LIBV4L2)
      set(HAVE_LIBV4L YES)
    else()
      set(HAVE_LIBV4L NO)
    endif()
  endif()
  CHECK_INCLUDE_FILE(linux/videodev.h HAVE_CAMV4L)
  CHECK_INCLUDE_FILE(linux/videodev2.h HAVE_CAMV4L2)
  CHECK_INCLUDE_FILE(sys/videoio.h HAVE_VIDEOIO)
status("  seting ...")
status("  WITH_V4L:" ${WITH_V4L})
status("  HAVE_LIBV4L:" ${HAVE_LIBV4L})
status("  HAVE_CAMV4L:" ${HAVE_CAMV4L})
status("  HAVE_CAMV4L2:" ${HAVE_CAMV4L2})
status("  HAVE_VIDEOIO:" ${HAVE_VIDEOIO})
endif(WITH_V4L)
#luoyouren-20160629
set(HAVE_CAMV4L2 ON)
status("  after seting ...")
status("  WITH_V4L:" ${WITH_V4L})
status("  HAVE_LIBV4L:" ${HAVE_LIBV4L})
status("  HAVE_CAMV4L:" ${HAVE_CAMV4L})
status("  HAVE_CAMV4L2:" ${HAVE_CAMV4L2})
status("  HAVE_VIDEOIO:" ${HAVE_VIDEOIO})
结果
技术分享
技术分享
技术分享
技术分享

g. 再编译
cd build_android_arm
make
make install
SDK: build_android_arm_2

四、对V4L2部分代码添加android_log

1. cap_v4l.cpp增加android_log

#include <android/log.h>
#undef LOG_TAG
#undef LOGD
#undef LOGE
#undef LOGI
#define LOG_TAG "OpenCV::android_v4l2"
#define LOGD(...) ((void)__android_log_print(ANDROID_LOG_DEBUG, LOG_TAG, __VA_ARGS__))
#define LOGI(...) ((void)__android_log_print(ANDROID_LOG_INFO, LOG_TAG, __VA_ARGS__))
#define LOGE(...) ((void)__android_log_print(ANDROID_LOG_ERROR, LOG_TAG, __VA_ARGS__))
然后在关键的一些函数增加log打印信息
重新编译，结果SDK: build_android_arm_3

2. 导入app测试新编译的opencv lib是否可用

a. 首先要去除APP对openCV Manager的依赖
详见《yanzi_OpenCV4Android_3》
另外注意lib库的编译方式
LOCAL_PATH := $(call my-dir)  
include $(CLEAR_VARS)  
 
#博客案例：OpenCV4Android释疑:，透析Android以JNI调OpenCV的三种方式(让OpenCVManager永不困扰)
LIB_TYPE = STATIC
ifeq ($(LIB_TYPE), STATIC) 
#STATIC链接时，可以不额外携带其依赖库
#当JNI里面调用了cvCaptureFromCAM去打开AndroidNativeCamera，就需要OPENCV_CAMERA_MODULES:=on
#来增加libnative_camera_r4.4.0.so
	OPENCV_LIB_TYPE:=STATIC
	OPENCV_INSTALL_MODULES:=on
	OPENCV_CAMERA_MODULES:=on
else 
#SHARED链接时，必须携带其依赖库，否则运行报错
	OPENCV_LIB_TYPE:=SHARED
	OPENCV_INSTALL_MODULES:=on
	OPENCV_CAMERA_MODULES:=on
endif 
 
 
#原始openCV4Android SDK
#include ../../openCV_2410_sdk/native/jni/OpenCV.mk
 
#添加V4L2，重新编译后的openCV4Android SDK
include ../../openCV_2410_sdk_v4l2/native/jni/OpenCV.mk
 
$(warning "****************************************")
$(warning $(LOCAL_C_INCLUDES))
 
LOCAL_LDLIBS += -llog 
LOCAL_SHARED_LIBRARIES += \
						libandroid_runtime\
						liblog \
						libcutils \
						libnativehelper \
						libcore/include	
 
LOCAL_C_INCLUDES += $(LOCAL_PATH)
						
LOCAL_SRC_FILES  := ImageProc.cpp  
LOCAL_MODULE     := image_proc  
include $(BUILD_SHARED_LIBRARY)

b. 同时在系统上删除opencv manager相关
Firmware: 706K_uvc_V003_[eng]_A706K-emmc_800X480_20160630_103200_[0630_1045]

c. 安装运行
yanzi_OpenCV4Android_3.apk
01-01 02:15:19.843: I/luo.uvc.jni(6649): open uvc3 ......
01-01 02:15:19.844: I/OPENCV_LUO(6649): usb camera Java_luo_uvc_jni_ImageProc_connectCamera start ....
01-01 02:15:19.844: I/OPENCV_LUO(6649): usb camera (-1) Java_luo_uvc_jni_ImageProc_connectCamera
01-01 02:15:19.844: I/OpenCV::android_v4l2(6649): CvCaptureCAM_V4L_CPP::CvCaptureCAM_V4L_CPP(-1)
01-01 02:15:19.844: I/OpenCV::android_v4l2(6649): icvCaptureFromCAM_V4L::use /dev/video0
01-01 02:15:19.844: E/OPENCV_LUO(6649): usb camera 0 open success!
01-01 02:15:19.844: I/luo.uvc.jni(6649): open uvc success!!!
从log信息可以看出，JNI已经成功打开了/dev/video0设备
而且实验得知index = -1 或者0，都会打开/dev/video0设备

3. 测试客户的CameraTouch.apk

a. 直接运行奔溃

02:20:51.968: D/dalvikvm(6834): Trying to load lib /data/app-lib/CameraTouch/libNDKBase.so 0x41ce1eb0
01-01 02:20:51.986: E/ION_DEBUG(6834): [add_system_map_entry]so libNDKBase.so address 0x612b7000 size 3563520 has been load into memory ion_debug_lock 0x404ad01c
01-01 02:20:52.033: D/dalvikvm(6834): Added shared lib /data/app-lib/CameraTouch/libNDKBase.so 0x41ce1eb0
01-01 02:20:52.034: D/dalvikvm(6834): No JNI_OnLoad found in /data/app-lib/CameraTouch/libNDKBase.so 0x41ce1eb0, skipping init
01-01 02:20:52.035: D/NDKBASE(6834): CMPBase::CMPBase()
01-01 02:20:52.036: D/NDKBASE(6834): CHelper::Open END
01-01 02:20:52.036: D/NDKBASE(6834): Java_com_indesign_cameratouch_NDKBase_OpenTouchNDK
01-01 02:20:52.037: D/NDKBASE(6834): CImageBase::captureThread start
01-01 02:20:52.037: D/NDKBASE(6834): CImageBase::captureThread start111
01-01 02:20:52.038: D/NDKBASE(6834): CImageBase::captureThread start00
01-01 02:20:52.038: D/NDKBASE(6834): CImageBase::captureThread start11
01-01 02:20:52.040: D/NDKBASE(6834): CCamInfo::checkDevice start
01-01 02:20:52.040: D/NDKBASE(6834): CCamInfo::checkDevice 1111
01-01 02:20:52.041: D/NDKBASE(6834): usb_device_added, Name: /dev/bus/usb/001/004
01-01 02:20:52.042: D/NDKBASE(6834): usb_device_added, Name: /dev/bus/usb/001/001
01-01 02:20:52.042: D/NDKBASE(6834): CCamInfo::checkDevice END0
01-01 02:20:52.042: D/NDKBASE(6834): CImageBase::captureThread start22

b. 把新编译出来的opencv4android lib库预置入系统再运行
Firmware: 706K_uvc_V003_[eng]_A706K-emmc_800X480_20160630_103200_[0630_1711]

c. 增加cvCreateCameraCapture()中的log再编译lib再植入系统
SDK: build_android_arm_4
Firmware: 706K_uvc_V003_[eng]_A706K-emmc_800X480_20160630_103200_[0630_1905]

07-01 02:41:07.644: D/NDKBASE(6805): CMPBase::CMPBase()
07-01 02:41:07.645: D/NDKBASE(6805): CHelper::Open END
07-01 02:41:07.645: D/NDKBASE(6805): Java_com_indesign_cameratouch_NDKBase_OpenTouchNDK
07-01 02:41:07.645: D/NDKBASE(6805): CImageBase::captureThread start
07-01 02:41:07.645: D/NDKBASE(6805): CImageBase::captureThread start111
07-01 02:41:07.645: D/NDKBASE(6805): CImageBase::captureThread start00
07-01 02:41:07.645: D/NDKBASE(6805): CImageBase::captureThread start11
07-01 02:41:07.645: D/NDKBASE(6805): CCamInfo::checkDevice start
07-01 02:41:07.645: D/NDKBASE(6805): CCamInfo::checkDevice 1111
07-01 02:41:07.646: D/NDKBASE(6805): usb_device_added, Name: /dev/bus/usb/001/002
07-01 02:41:07.646: D/NDKBASE(6805): usb_device_added, Name: /dev/bus/usb/001/001
07-01 02:41:07.646: D/NDKBASE(6805): CCamInfo::checkDevice END1
07-01 02:41:07.646: D/NDKBASE(6805): CImageBase::captureThread start333
07-01 02:41:07.646: D/NDKBASE(6805): CameraId: 0, checking.
07-01 02:41:07.646: D/NDKBASE(6805): CameraId:cvCaptureFromCAM star
07-01 02:41:07.647: D/OpenCV::camera(6805): CvCapture_Android::CvCapture_Android(0)
07-01 02:41:07.647: D/OpenCV::camera(6805): Library name: libNDKBase.so
07-01 02:41:07.647: D/OpenCV::camera(6805): Library base address: 0x61191000
07-01 02:41:07.691: D/OpenCV::camera(6805): Libraries folder found: /data/app-lib/CameraTouch/
07-01 02:41:07.691: D/OpenCV::camera(6805): CameraWrapperConnector::connectToLib: folderPath=/data/app-lib/CameraTouch/
07-01 02:41:07.706: V/ActivityManager(684): Broadcast: Intent { act=android.intent.action.DROPBOX_ENTRY_ADDED flg=0x10 (has extras) } ordered=false userid=0 callerApp=ProcessRecord{41d6b1a0 684:system/1000}
07-01 02:41:07.706: D/dalvikvm(684): threadid=83: exiting
07-01 02:41:07.714: D/dalvikvm(684): threadid=83: bye!
07-01 02:41:07.841: E/OpenCV::camera(6805): ||libnative_camera_r4.1.1.so
07-01 02:41:07.841: E/OpenCV::camera(6805): ||libnative_camera_r4.0.3.so
07-01 02:41:07.841: E/OpenCV::camera(6805): ||libnative_camera_r2.3.3.so
07-01 02:41:07.841: E/OpenCV::camera(6805): ||libnative_camera_r4.2.0.so
07-01 02:41:07.841: E/OpenCV::camera(6805): ||libnative_camera_r2.2.0.so
07-01 02:41:07.841: E/OpenCV::camera(6805): ||libnative_camera_r3.0.1.so
07-01 02:41:07.841: E/OpenCV::camera(6805): ||libnative_camera_r4.3.0.so
07-01 02:41:07.841: E/OpenCV::camera(6805): ||libnative_camera_r4.0.0.so
07-01 02:41:07.841: E/OpenCV::camera(6805): ||libnative_camera_r4.4.0.so
07-01 02:41:07.841: D/OpenCV::camera(6805): try to load library ‘libnative_camera_r4.4.0.so‘
07-01 02:41:07.852: E/ION_DEBUG(6805): [add_system_map_entry]so libnative_camera_r4.4.0.so address 0x614f9000 size 282624 has been load into memory ion_debug_lock 0x4048601c
07-01 02:41:07.853: D/OpenCV::camera(6805): Loaded library ‘/data/app-lib/CameraTouch/libnative_camera_r4.4.0.so‘
07-01 02:41:07.854: D/OpenCV_NativeCamera(6805): CameraHandler::initCameraConnect(0x6122beed, 0, 0x6055fde8, 0x0)
07-01 02:41:07.855: D/OpenCV_NativeCamera(6805): Current process name for camera init: com.indesign.cameratouch
07-01 02:41:08.036: D/CameraService(132): CameraService::connect E (pid 6805 "com.indesign.cameratouch", id 0)
07-01 02:41:08.036: E/CameraService(132): CameraService::connect X (pid 6805) rejected (invalid cameraId 0).
07-01 02:41:08.037: W/CameraBase(6805): An error occurred while connecting to camera: 0
07-01 02:41:08.038: E/OpenCV_NativeCamera(6805): initCameraConnect: Unable to connect to CameraService
07-01 02:41:08.038: E/OpenCV::camera(6805): CameraWrapperConnector::connectWrapper ERROR: the initializing function returned false
07-01 02:41:08.038: E/OpenCV::camera(6805): Native_camera returned opening error: 6
07-01 02:41:08.038: D/NDKBASE(6805): CameraId: 0, fail.
07-01 02:41:08.038: D/NDKBASE(6805): CImageBase::captureThread start333
07-01 02:41:08.038: D/NDKBASE(6805): CameraId: 1, checking.
07-01 02:41:08.038: D/NDKBASE(6805): CameraId:cvCaptureFromCAM star
07-01 02:41:08.038: D/OpenCV::camera(6805): CvCapture_Android::CvCapture_Android(1)
07-01 02:41:08.038: D/OpenCV_NativeCamera(6805): CameraHandler::initCameraConnect(0x6122beed, 1, 0x6055fde8, 0x0)
07-01 02:41:08.038: D/OpenCV_NativeCamera(6805): Current process name for camera init: com.indesign.cameratouch
07-01 02:41:08.040: D/CameraService(132): CameraService::connect E (pid 6805 "com.indesign.cameratouch", id 1)
07-01 02:41:08.040: E/CameraService(132): CameraService::connect X (pid 6805) rejected (invalid cameraId 1).
07-01 02:41:08.041: W/CameraBase(6805): An error occurred while connecting to camera: 1
07-01 02:41:08.041: E/OpenCV_NativeCamera(6805): initCameraConnect: Unable to connect to CameraService
07-01 02:41:08.041: E/OpenCV::camera(6805): CameraWrapperConnector::connectWrapper ERROR: the initializing function returned false
07-01 02:41:08.041: E/OpenCV::camera(6805): Native_camera returned opening error: 6
07-01 02:41:08.041: D/NDKBASE(6805): CameraId: 1, fail.

经验证CameraTouch.apk始终调用的是CvCapture_Android去打开Native Camera，并没有去调用CvCaptureCAM_V4L_CPP去打开video0设备，猜测CameraTouch.apk的opencv依赖库没有加入V4L2.

4. 更改yanzi_OpenCV4Android.apk实现获取video0的图像

a. 去除Android系统所有opencv相关文件
OpenCV_2.4.10_Manager_2.19_armv7a-neon.apk
Opencv for android lib.so 
CameraTouch.apk

Firmware: 706K_uvc_V004_[eng]_A706K-emmc_800X480_20160701_185957_[0702_0906]

b. 获取video0的图像

frame = cvQueryFrame(capture); //取图片帧
cvQueryFrame调用出问题，继续修改opencv源码增加log
SDK: build_android_arm_5

yanzi_OpenCV4Android_4.apk测试结果：崩溃
技术分享
技术分享
分析：select timeout说明没有获取到frame，但是icvRetrieveFrameCAM_V4L函数进行了frame.imageData拷贝工作，估计操作了空指针，于是报SIGSEGV段错误！

c. 修改bug

case PALETTE_YUYV:
		LOGI("icvRetrieveFrameCAM_V4L: capture->palette = PALETTE_YUYV");
        yuyv_to_rgb24(capture->form.fmt.pix.width,
                  capture->form.fmt.pix.height,
                  (unsigned char*)(capture->buffers[capture->bufferIndex].start),
                  (unsigned char*)capture->frame.imageData);
        break;
其中的源数据地址(unsigned char*)(capture->buffers[capture->bufferIndex].start可能是无效的，
所以预先设置一个默认值， 使得源数据地址是有效地址，防止段错误：
capture->bufferIndex = MAX_V4L_BUFFERS;

SDK: build_android_arm_6

d. yanzi_OpenCV4Android_4.apk运行正常
可以获取的图像帧数据，前面的bug实际上是，初始化摄像头时，set format 为YUYV格式而且return success，但是我的MTK Android平台不支持YUYV格式，只支持MJPEG格式，所以出现select timeout，然后就出现段错误。


至此，opencv4android sdk对于V4L2的支持，已经添加完成，已经可以正常获取图像。接下来需要测试其稳定性，另外新的sdk对于Android Native Camera的支持理论上来说，没有被破坏，具体需要实验验证！









我的Opencv4Android添加V4L2支持的移植记录

标签：

原文地址：http://blog.csdn.net/luoyouren/article/details/51818523
踩
(0)
赞
(1)
   
举报
评论 一句话评论（0）
登录后才能评论！
分享档案
更多>
2021年07月29日 (22)
2021年07月28日 (40)
2021年07月27日 (32)
2021年07月26日 (79)
2021年07月23日 (29)
2021年07月22日 (30)
2021年07月21日 (42)
2021年07月20日 (16)
2021年07月19日 (90)
2021年07月16日 (35)
周排行
mamicode.com排行更多图片
更多
友情链接
兰亭集智   国之画   百度统计   站长统计   阿里云   chrome插件   新版天听网
关于我们 - 联系我们 - 留言反馈
© 2014 mamicode.com 版权所有  联系我们:gaon5@hotmail.com
迷上了代码！