最近需要将Python人证对比模型移植安卓端.安卓端需要使用openCV简单的人像提取处理.在配置openCV前首先需要配置NDK环境.

NDK的介绍()

JNI与NDK详细介绍: 

一 NDK在Ubuntu的配置

参考: 

二 NDK在IntelliJ IDEA中的配置

参考: 

三 NDK开发实例

参考: 

在main/java下创建创建一个NDKUtils.Java类:

public class NDKUtils {    public native String invokeCmethod();    static {        System.loadLibrary("NDKUtils");    }}

右键选择 External Tools ---> javah

会在app目录下生成jni目录和目录内一个com_**_demo_utils_NDKUtils.h文件

复制.h 文件修改后缀名为 .c (该文件就是C 语言的实现文件)

修改com_**_demo_utils_NDKUtils.c内容:

JNIEXPORT jstring JNICALL Java_com_**_demo_utils_NDKUtils_jniInit  (JNIEnv *env, jobject obj){                       return (*env)->NewStringUTF(env,"NDK 获取Native 字符串");                       }

 

将jni目录拷贝至main/java目录下.

在jni目录下新建Android.mk和Application.mk:

Android.mk:

LOCAL_PATH:=$(call my-dir)include $(CLEAR_VARS)LOCAL_MODULE := NDKUtilsLOCAL_SRC_FILES := com_**_demo_utils_NDKUtils.cinclude $(BUILD_SHARED_LIBRARY)

Application.mk:

APP_PLATFORM := android-23APP_ABI := armeabi

在IntelliJ IDEA的命令窗口(Terminal)执行:

cd app/src/main/

继续输入:

ndk-build

该命令是编译C文件，然后生成 .so 文件，生成在 app\src\main\libs 目录下。

完成后会在main目录下生成libs和obj两个目录.

配置build.gradle 文件,在android {}中添加:

sourceSets{        main{            jni.srcDirs = [] // disable automatic ndk-build call, which ignore our Android.mk            jniLibs.srcDir 'src/main/libs'        }    }

在MainActivity中使用:

package com.***.opencvdemo;import android.support.v7.app.AppCompatActivity;import android.os.Bundle;import android.widget.TextView;public class MainOpencvActivity extends AppCompatActivity {    @Override    protected void onCreate(Bundle savedInstanceState) {        super.onCreate(savedInstanceState);        setContentView(R.layout.activity_main_opencv);        NDKUtils ndk = new NDKUtils();        TextView text = (TextView) findViewById(R.id.tv);        text.setText(ndk.invokeCmethod());    }}

运行程序会输出字符串:   NDK 获取Native 字符串 

以上测试成功.

 

四 Android+openCV

首先下载openCV包,地址:

最新的3.3,我下载了3.2

opencv配置参考: 

NDKUtils.Java代码:

package com.***.opencvdemo;public class NDKUtils {    //public native String invokeCmethod();    public static native int[] gray(int[] buf, int w, int h);    static {        //System.loadLibrary("NDKUtils");        System.loadLibrary("OpenCV");//导入生成的链接库文件    }}

右键选择 External Tools ---> javah  同上,把生成目录jni拷贝到 src/main.

复制.h文件 修改后缀名为 .cpp   (.c文件编译会报错) 参考: 

编辑 com_***_opencvdemo_NDKUtils.cpp:

#include 
     
      #include 
      
       #include 
       
        #include 
        
         #include 
         
          #include 
          
           using namespace cv;extern "C" JNIEXPORT jintArray JNICALL Java_com_***_opencvdemo_NDKUtils_gray( JNIEnv *env, jclass obj, jintArray buf, int w, int h) { jint *cbuf; cbuf = env->GetIntArrayElements(buf, JNI_FALSE ); if (cbuf == NULL) { return 0; } Mat imgData(h, w, CV_8UC4, (unsigned char *) cbuf); uchar* ptr = imgData.ptr(0); for(int i = 0; i < w*h; i ++){ //计算公式：Y(亮度) = 0.299*R + 0.587*G + 0.114*B //对于一个int四字节，其彩色值存储方式为：BGRA int grayScale = (int)(ptr[4*i+2]*0.299 + ptr[4*i+1]*0.587 + ptr[4*i+0]*0.114); ptr[4*i+1] = grayScale; ptr[4*i+2] = grayScale; ptr[4*i+0] = grayScale; } int size = w * h; jintArray result = env->NewIntArray(size); env->SetIntArrayRegion(result, 0, size, cbuf); env->ReleaseIntArrayElements(buf, cbuf, 0); return result; }
          
         
        
       
      
     

 

把 sdk/native 目录 拷贝到 项目的 jni 目录下.

 

修改:

Android.mk:

LOCAL_PATH:=$(call my-dir)include $(CLEAR_VARS)OpenCV_INSTALL_MODULES := onOpenCV_CAMERA_MODULES := offOPENCV_LIB_TYPE :=STATICifeq ("$(wildcard $(OPENCV_MK_PATH))","")include $(LOCAL_PATH)/native/jni/OpenCV.mkelseinclude $(OPENCV_MK_PATH)endifLOCAL_MODULE := OpenCVLOCAL_SRC_FILES := com_***_opencvdemo_NDKUtils.cppLOCAL_LDLIBS +=  -lm -lloginclude $(BUILD_SHARED_LIBRARY)

Application.mk:

APP_STL := gnustl_staticAPP_CPPFLAGS := -frtti -fexceptionsAPP_PLATFORM := android-9

 

ndk-build:生成.so文件

MainOpencvActivity:

package com.***.opencvdemo;import android.graphics.Bitmap;import android.graphics.drawable.BitmapDrawable;import android.support.v7.app.AppCompatActivity;import android.os.Bundle;import android.widget.ImageView;public class MainOpencvActivity extends AppCompatActivity {    @Override    protected void onCreate(Bundle savedInstanceState) {        super.onCreate(savedInstanceState);        setContentView(R.layout.activity_main_opencv);        Bitmap bitmap = ((BitmapDrawable) getResources().getDrawable(                R.mipmap.aaa)).getBitmap();        int w = bitmap.getWidth(), h = bitmap.getHeight();        int[] pix = new int[w * h];        bitmap.getPixels(pix, 0, w, 0, 0, w, h);        int [] resultPixes= NDKUtils.gray(pix,w,h);        Bitmap result = Bitmap.createBitmap(w,h, Bitmap.Config.RGB_565);        result.setPixels(resultPixes, 0, w, 0, 0,w, h);        ImageView img = (ImageView)findViewById(R.id.image2);        img.setImageBitmap(result);    }}

附加 activity_main_opencv.xml:

运行效果: