OpenCV4 Android 搭建入门 demo

短命女 2023-02-14 13:04 50阅读 0赞

# 下载并集成 OpenCV #

## 下载 ##

首先到 OpenCV 官网，下载 Android 版本  
![在这里插入图片描述][20200602211523504.png]  
下载后的目录结构如下  
![在这里插入图片描述][20200602212148259.png]

## 导入项目并做修改 ##

打开 Android Studio，新建项目  
然后 File ——> New ——> Import Module，选择 `D:\Downloads\opencv-4.2.0-android-sdk\OpenCV-android-sdk\sdk` (该目录根据你自己下载的位置而定)，引入的时候会让你输入模块名，我是填 opencv，你们可以随意用自己喜欢的名字

对 opencv 的 build.gradle 做 2 处修改

1.  compileSdkVersion 和 targetSdkVersion 都改为最新版  
    ![在这里插入图片描述][20200602212729471.png]2. 编译的 Java 版本改为最新的  
    ![在这里插入图片描述][20200602212849637.png]

## 国内被墙，gradle 同步失败问题 ##

修改项目根目录的 build.gradle 文件，通用模板如下（包括flutter同步失败，证数错误等问题，都可以使用该模板解决）

// Top-level build file where you can add configuration options common to all sub-projects/modules.
    
    buildscript {
        repositories {
            //使用阿里云国内镜像
            maven{ url 'https://maven.aliyun.com/nexus/content/groups/public/'}
            maven{url 'https://maven.aliyun.com/repository/google'}
            maven{url 'https://maven.aliyun.com/repository/jcenter'}
            mavenCentral()
            mavenLocal()
            google()
            jcenter()
        }
        dependencies {
            classpath 'com.android.tools.build:gradle:3.6.3'
    
            // NOTE: Do not place your application dependencies here; they belong
            // in the individual module build.gradle files
            //https://github.com/JakeWharton/butterknife
            classpath 'com.jakewharton:butterknife-gradle-plugin:10.2.1'
            //https://github.com/greenrobot/greenDAO
            classpath 'org.greenrobot:greendao-gradle-plugin:3.2.2'
            //AOP方案一
    //        classpath 'com.hujiang.aspectjx:gradle-android-plugin-aspectjx:2.0.4'
            //AOP方案二：
    //        classpath 'org.aspectj:aspectjtools:1.8.13'
    //        classpath 'org.aspectj:aspectjweaver:1.8.13'
        }
    }
    
    allprojects {
        repositories {
            maven { url 'https://maven.aliyun.com/nexus/content/groups/public/' }
            maven { url 'https://maven.aliyun.com/nexus/content/repositories/jcenter' }
            maven { url 'https://maven.aliyun.com/nexus/content/repositories/google' }
            maven { url 'https://maven.aliyun.com/nexus/content/repositories/gradle-plugin' }
            mavenCentral()
            // 用来引入 rxpermission, GreenDaoUpgradeHelper 等第三方库
            maven { url 'https://jitpack.io' }
            mavenLocal()
            // bmob后台数据库，用来更新apk
            maven { url "https://raw.github.com/bmob/bmob-android-sdk/master" }
            google()
            jcenter()
        }
    }
    
    task clean(type: Delete) {
        delete rootProject.buildDir
    }

![在这里插入图片描述][20200602213614127.png]

## testOnly 问题 ##

Android Studio 3.0会在debug apk的manifest文件application标签里自动添加 android:testOnly="true"属性，导致IDE中run跑出的apk在大部分手机上只能用 `adb install -t <apk>` 来安装

在项目根目录的 gradle.properties 中添加 `android.injected.testOnly=false`

![在这里插入图片描述][20200602213438580.png]同步一下，至此，编译成功

# 原理讲解 #

修改 MainActivity.java

public class MainActivity extends AppCompatActivity {
        @Override
        protected void onCreate(Bundle savedInstanceState) {
            super.onCreate(savedInstanceState);
            setContentView(R.layout.activity_main);
            //初始化
            if (OpenCVLoader.initDebug()) {
                Toast.makeText(this, "OpenCVLoader初始化成功", Toast.LENGTH_SHORT).show();
            }
    	}
    }

OpenCVLoader 会尝试加载OpenCV 要用到的所有 so 库，如果全部加载成功并返回true，则说明集成正确。

> 在 OpenCV 3.0 之前，要使用 OpenCV 的功能，都必须额外安装一个 OpenCV Manager 的 apk，但 3.0 版本之后就不用了， OpenCV 的所有功能都已经被集成进 so 了，只需要调用OpenCVLoader.initAsync 然后 OpenCV 会在初始化完毕之后回调 LoaderCallbackInterface 接口的 onManagerConnected 来通知我们初始化完毕。

initDebug 代码如下

public class OpenCVLoader
    {
    	public static boolean initDebug()
        {
            return StaticHelper.initOpenCV(false);
        }
    	（略去后面代码）
    }

initOpenCV 代码如下

class StaticHelper {
    
        public static boolean initOpenCV(boolean InitCuda)
        {
            boolean result;
            String libs = "";
    
            if(InitCuda)
            {
                loadLibrary("cudart");
                loadLibrary("nppc");
                loadLibrary("nppi");
                loadLibrary("npps");
                loadLibrary("cufft");
                loadLibrary("cublas");
            }
    
            Log.d(TAG, "Trying to get library list");
    
            try
            {
                System.loadLibrary("opencv_info");
                libs = getLibraryList();
            }
            catch(UnsatisfiedLinkError e)
            {
                Log.e(TAG, "OpenCV error: Cannot load info library for OpenCV");
            }
    
            Log.d(TAG, "Library list: \"" + libs + "\"");
            Log.d(TAG, "First attempt to load libs");
            if (initOpenCVLibs(libs))
            {
                Log.d(TAG, "First attempt to load libs is OK");
                String eol = System.getProperty("line.separator");
                for (String str : Core.getBuildInformation().split(eol))
                    Log.i(TAG, str);
    
                result = true;
            }
            else
            {
                Log.d(TAG, "First attempt to load libs fails");
                result = false;
            }
    
            return result;
        }
    	
    	（略去后面代码） 
    }

加载成功后，日志会打印如下两段内容  
![在这里插入图片描述][20200602214920163.png]  
![在这里插入图片描述][20200602215212127.png]

## 加载和初始化库的两类函数 ##

在 OpenCVLoader.java 中，有两类函数用来加载 OpenCV 库

*  同步函数 initDebug(boolean InitCuda)
 *  异步函数 initAsync(String Version, Context AppContext, LoaderCallbackInterface Callback)

源码如下，从注释可以看出，两类函数的功能都是一样的 —— 从应用程序包中，加载并初始化 OpenCV 库，它差不多类似 system.loadLibrary(“opencv\_java”)

public class OpenCVLoader {
    	/**
         * Current OpenCV Library version
         */
        public static final String OPENCV_VERSION = "4.2.0";
    
    
        /**
         * Loads and initializes OpenCV library from current application package. Roughly, it's an analog of system.loadLibrary("opencv_java").
         * @return Returns true is initialization of OpenCV was successful.
         */
        public static boolean initDebug()
        {
            return StaticHelper.initOpenCV(false);
        }
    
        /**
         * Loads and initializes OpenCV library from current application package. Roughly, it's an analog of system.loadLibrary("opencv_java").
         * @param InitCuda load and initialize CUDA runtime libraries.
         * @return Returns true is initialization of OpenCV was successful.
         */
        public static boolean initDebug(boolean InitCuda)
        {
            return StaticHelper.initOpenCV(InitCuda);
        }
    
        /**
         * Loads and initializes OpenCV library using OpenCV Engine service.
         * @param Version OpenCV library version.
         * @param AppContext application context for connecting to the service.
         * @param Callback object, that implements LoaderCallbackInterface for handling the connection status.
         * @return Returns true if initialization of OpenCV is successful.
         */
        public static boolean initAsync(String Version, Context AppContext,
                LoaderCallbackInterface Callback)
        {
            return AsyncServiceHelper.initOpenCV(Version, AppContext, Callback);
        }
    }

实际使用上，我们一般会先调用其同步方法判断是否已加载成功，如果未成功再调用异步加载函数，比如

if(OpenCVLoader.initDebug()) {
    	//执行你需要的opencv函数
    } else {
    	OpenCVLoader.initAsync(OpenCVLoader.OPENCV_VERSION, this, baseLoaderCallback);
    }

[20200602211523504.png]: https://img-blog.csdnimg.cn/20200602211523504.png
[20200602212148259.png]: https://img-blog.csdnimg.cn/20200602212148259.png
[20200602212729471.png]: https://img-blog.csdnimg.cn/20200602212729471.png
[20200602212849637.png]: https://img-blog.csdnimg.cn/20200602212849637.png
[20200602213614127.png]: https://img-blog.csdnimg.cn/20200602213614127.png
[20200602213438580.png]: https://img-blog.csdnimg.cn/20200602213438580.png
[20200602214920163.png]: https://img-blog.csdnimg.cn/20200602214920163.png
[20200602215212127.png]: https://img-blog.csdnimg.cn/20200602215212127.png