Shulou(Shulou.com)06/01 Report--

Android studio how to use jni to compile opencv, many novices are not very clear about this, in order to help you solve this problem, the following editor will explain in detail for you, people with this need can come to learn, I hope you can get something.

1. Sigh about the process (if you want to see the tutorial directly, please skip this section)

Before I write the details, I'd like to talk about the process of doing this thing, because the tutor said to have an image matching androi APP, which is the similarity between the two images, similar to the facial scan unlocking introduced by Android 5.0.

At that time, to achieve such a thing, there must be no off-the-shelf API available, the first thing that comes to mind is undoubtedly opencv, which has a powerful set of image processing functions in the library, and its development language is mainly Clippers. However, there are also jar packages available for android development, and if you only use the effects already written inside, you will definitely not be able to complete image matching.

In other words, I have to call the functions inside it and then combine my own algorithm to achieve such a function, and then use the ndk environment to achieve jni programming. I will introduce and use the C++ code written by myself on the basis of generating .so dynamic library.

At first, the train of thought was very clear, and then we started the opencv jni programming tutorial of Baidu android studio (hereinafter referred to as as). It is a great pity that what we can find is that the examples of as, opencv and jni are almost zero, and many of them are eclipse. I have no choice but to do it myself.

When I first started, I quickly got all the equipment work done, and the so dynamic library file (described below) was also compiled, but at this point, I came across a bugboat that made me give up in the first phase!

The bug is: (I will make it clear below, its real cause and solution)

Fatal error: opencv2/opencv.hpp: No such file or directory, which means that the header file opencv2/opencv.hpp in the cpp file I want to compile cannot be found. At that time, whether it is to consult others, Baidu, google or search books, can not be solved, it takes a full week!

Keep the project information one by one and give up.

Until 2 days ago, decided to try again, and formally resolved today, this article is now published.

2, operating environment

Win 7, system

Android studio version 0.8.0 beta, using build:gradle:0.12.+,tools version: 21.1.2 Die API 21

Opencv for android package, the version I use is OpenCV-3.0.0-android-sdk,2.4.9, which can be downloaded from the official website of opencv. I provide a link here.

Http://downloads.sourceforge.net/project/opencvlibrary/opencv-android/3.0.0/OpenCV-3.0.0-android-sdk-1.zip?r=http%3A%2F%2Fopencv.org%2F&ts=1436167636&use_mirror=nchc

Compile. so dynamic library uses cygwin and installs all packages. Here's a hint: you don't have to use it, you can compile it directly with cmd.

Ndk for android-ndk-r10d (it is strongly recommended to use R9 or R10 series, because these two can be compiled in cmd. So), R10d can support up to 21 android api, if yours is 22, please modify, otherwise there will not be compiled jni.h header files, or other header files, you will find that other people's source code can not be compiled in you.

3. Preparation work

1the download, installation and configuration of mai Mui muri NDK, not to mention this part, there are many online tutorials, many are feasible.

According to http://blog.csdn.net/asmcvc/article/details/9311573, I said above, you don't have to use it. The cmd that comes with win can also be compiled. If you use cygwin, be prepared to download and install it for a very long time, with a total file volume of more than 20 gigabytes! It took me more than nine hours.

3When the sdk download of the for android is completed. Open this folder, sdk/native/libs, there are many platform folders, which can appear in it, which proves that you can set the .so file to generate the corresponding architecture in the following Application.mk. Let me give you an example. Mine is:

There is a sentence in the Application.mk file introduced below, which is used to set the .so file to generate the corresponding schema. I am armeabi-7a here. If you want to generate all, write: = all, note, this is likely to make an error. The error message is that a certain architecture cannot be found. So, I want you to see clearly what schemas are in the above folder. These pits cannot be found online. If you want to generate two kinds of pits, It can be compiled in turn, and the second compilation will not be covered by different architectures. Now open sdk/native/jni, and if there is no accident, there must be a file called OpenCV.mk, which is the file we want to refer to when we introduce opencv C++ library into the android.mk script file. Please open it with notepad or Notepad++.

4Murray-learn the basic content information of Android.mk and Application.mk files: below I use the default Android.mk to illustrate, and the Application.mk of my example to illustrate.

They are all script files.

Android.mk

Application.mk

4, compile .so

Create a new project using your as, and then create a jni folder under your project's main directory, thus creating:

After it is created, it goes like this:

First compile the header file .h of the project. Generally, after compilation, its name structure is: package name _ class name.h

The compilation command is as follows, please enter it in Terminal under your as:

SourcePath: d:\ work\ androidstudio\ VisualRecognition\ app\ src\ main\ java (absolute path)

TargetPath: d:\ work\ androidstudio\ VisualRecognition\ visual\ src\ main\ jni (absolute path)

TargetClassName: com.yf.visualrecognition.UnityPlayerActivity (your package name + class name)

Format: javah-d ${SourceFile}-classpath ${TargetPath} ${TargetClassName}

Console instruction: javah-d D:\ work\ androidstudio\ VisualRecognition\ visual\ src\ main\ jni-classpath D:\ work\ androidstudio\ VisualRecognition\ app\ src\ main\ java io.github.froger.jni.MyActivity

Then create Android.mk, Application.mk and the .cpp or .c file you want to compile under your jni folder. The contents of the first two can be similar to what I described above, and .cpp I provide one here.

Android.mk, Application.mk, and ImgFuncpp are as follows. Util.c is an empty file. It is created to avoid another bug, not to mention:

The Android.mk file is as follows

LOCAL_PATH: = $(call my-dir) include $(CLEAR_VARS) OPENCV_LIB_TYPE:=STATICifeq ("$(wildcard $(OPENCV_MK_PATH))", "") include E:\ OpenCV-3.0.0-android-sdk-1\ OpenCV-android-sdk\ sdk\ native\ jni\ OpenCV.mkelseinclude $(OPENCV_MK_PATH) endifLOCAL_MODULE: = ImgFunLOCAL_SRC_FILES: = ImgFun.cppLOCAL_LDLIBS + =-lm-lloginclude $(BUILD_SHARED_LIBRARY)

The Application.mk file is as follows

APP_STL: = gnustl_staticAPP_CPPFLAGS: =-frtti-fexceptionsAPP_ABI: = armeabi-v7a # this sentence sets the generated cpu instruction type. Hint: at present, most Android phones support too many types under armeabi,libs, and compiling the apk package will be too large APP_PLATFORM: = android-8 # this sentence is to set the minimum Android platform, which can be avoided.

The ImgFun.cpp file is as follows

1 # include 2 # include 3 # include 4 # include 5 using namespace cv; 6 IplImage * change4channelTo3InIplImage (IplImage * src); 7 8 extern "C" {9 JNIEXPORT jintArray JNICALL Java_io_github_froger_jni_MyActivity_ImgFun (10 JNIEnv* env, jobject obj, jintArray buf, int w, int h); 11 JNIEXPORT jintArray JNICALL Java_io_github_froger_jni_MyActivity_ImgFun (12 JNIEnv* env, jobject obj, jintArray buf, int w, int h) {13 14 jint * cbuf;15 cbuf = env- > GetIntArrayElements (buf, false) 16 if (cbuf = = NULL) {17 return 0 myimg 18} 19 20 Mat myimg (h, w, CV_8UC4, (unsigned char*) cbuf); 21 IplImage image=IplImage (myimg); 22 IplImage* image3channel = change4channelTo3InIplImage (& image); 23 24 IplImage* pCannyImage=cvCreateImage (cvGetSize (image3channel), IPL_DEPTH_8U,1); 25 26 cvCanny (image3channel,pCannyImage,50,150,3); 27 28 int* outImage=new int [wigh]; 29 for (int iimageData) 32} 33 34 int size = w * hscape 35 jintArray result = env- > NewIntArray (size); 36 env- > SetIntArrayRegion (result, 0, size, outImage); 37 env- > ReleaseIntArrayElements (buf, cbuf, 0); 38 return result;39} 40} 41 42 IplImage * change4channelTo3InIplImage (IplImage * src) {43 if (src- > nChannels! = 4) {44 return NULL;45} 46 47 IplImage * destImg = cvCreateImage (cvGetSize (src), IPL_DEPTH_8U, 3); 48 for (int row = 0; row)

Height; row++) {49 for (int col = 0; col)

Width; col++) {50 CvScalar s = cvGet2D (src, row, col); 51 cvSet2D (destImg, row, col, s); 52} 53} 54 55 return destImg;56}

There are a few words to explain in the .cpp file above. Note that the .c file is different from the .cpp file:

1, please use extern "C" {} to wrap the definition of your C++ function body and the variables inside. The function declaration can be on the outside.

2JNIEXPort jintArray JNICALL Java_io_github_froger_jni_MyActivity_ImgFun (JNIEnv* env, jobject obj, jintArray buf, int w, int h)

3JintArray is the return value of the function you defined. What I have here is the int array, which has a j in front of the type. If it is a string, it is jstring, the array plus Array.

4JNICALL Java this sentence remains the same, all are the same, notice that the j of java is uppercase

5There is the name of your package

6. MyActivity your class name. Refer to the class name of C++ function here.

7GramImgFun is the name of the function you want to call in java, and those that do not need to be called directly do not need to be written

8JNIEnv * env, jobject obj, this is fixed, the first one means virtual machine reference, the second is the project

9 the parameters of the functions of JintArray buf, int w, int h.

Well, the above introduction is done, and then there is the formal operation of compiling .so (I use cmd as an example here, because it is easier to operate, and cygwin can also be used).

You can compile in the cmd of as or in the cmd box of the system, first use the command to go to the directory of the current jni folder, for example, mine is

D:asproject/JniDemo/app/main/jni, then use the command ndk-build, (using the ndk-build command requires that you have configured the ndk environment, please refer to the tutorials above on Baidu), then enter. If there is no accident, the following file will be generated:

The .so file is what we need. Now open the build.gradle file under your project app and add it to android {}:

SourceSets {main () {jniLibs.srcDirs = ['src/main/libs']}}

This is for the use of .so files, above we are just production!

OK, basically done here, but it was after this step that when I was running the program, a simple and deadly bug appeared, which led me to look for it for nearly 2 weeks. It's stupid to think of it now.

5. The key problems encountered and their solutions

Running the program, the following error occurs, declared here, not only opencv2/opencv.hpp, but also other hpp.

The reason for this is:

It turns out that after we finished compiling the .so file, android studio, we set up the opencv function library introduced in the Android.mk file, which has been compiled into the .so dynamic library, and the cpp file we need for compilation is in the jni folder, so naturally there is no opencv library to rely on.

Solution:

After you have compiled the. so file, you can comment or delete the contents of the cpp or c file, otherwise you will throw out errors that can not be found in the header file when you run the program. Hey, it is really bitter tears, such a bug has taken me so much time, but fortunately, it has been solved.

6, realize the effect.

1 package io.github.froger.jni; 2 3 import android.app.Activity; 4 import android.graphics.Bitmap; 5 import android.graphics.drawable.BitmapDrawable; 6 import android.os.Bundle; 7 import android.view.View; 8 import android.widget.Button; 9 import android.widget.ImageView;10 11 public class MyActivity extends Activity {12 / * * Called when the activity is first created. * / 13 ImageView imgView;14 Button btnNDK, btnRestore;15 public static native int [] ImgFun (int [] buf, int w, int h); 16 static {17 System.loadLibrary ("ImgFun"); 18} 19 @ Override20 public void onCreate (Bundle savedInstanceState) {21 super.onCreate (savedInstanceState); 22 setContentView (R.layout.activity_my); 23 24 this.setTitle ("convert grayscale images using NDK") 25 btnRestore = (Button) this.findViewById (R.id.btnRestore); 26 / / btnRestore.setText (ImgFun ()); 27 btnRestore.setOnClickListener (new ClickEvent ()); 28 btnNDK = (Button) this.findViewById (R.id.btnNDK); 29 btnNDK.setOnClickListener (new ClickEvent ()); 30 imgView = (ImageView) this.findViewById (R.id.ImageView01) 31 Bitmap img = (BitmapDrawable) getResources (). GetDrawable (32 R.drawable.ic_launcher). GetBitmap (); 33 imgView.setImageBitmap (img); 34} 35 36 class ClickEvent implements View.OnClickListener {37 public void onClick (View v) {38 / / btnRestore.setText (ImgFun ()) 39 if (v = = btnNDK) {40 long current = System.currentTimeMillis (); 41 Bitmap img1 = ((BitmapDrawable) getResources (). GetDrawable (42 R.drawable.ic_launcher)) .getBitmap (); 43 int w = img1.getWidth (), h = img1.getHeight (); 44 int [] pix = new int [w * h] 45 img1.getPixels (pix, 0, w, 0, 0, w, h); 46 int [] resultInt = ImgFun (pix, w, h); 47 Bitmap resultImg = Bitmap.createBitmap (w, h, Bitmap.Config.RGB_565); 48 resultImg.setPixels (resultInt, 0, w, 0, w, h); 49 long performance = System.currentTimeMillis ()-current 50 imgView.setImageBitmap (resultImg); 51} else if (v = = btnRestore) {52 Bitmap img2 = ((BitmapDrawable) getResources (). GetDrawable (53 R.drawable.ic_launcher)). GetBitmap (); 54 imgView.setImageBitmap (img2); 55} 56} 57} 58 59 60}

Is it helpful for you to read the above content? If you want to know more about the relevant knowledge or read more related articles, please follow the industry information channel, thank you for your support.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.