Android开发—基于OpenCV实现相机实时图像识别跟踪

图像识别

什么是图像识别

图像识别，是指利用计算机对图像进行处理、分析和理解，以识别各种不同模式的目标和对像的技术。根据观测到的图像，对其中的物体分辨其类别，做出有意义的判断。利用现代信息处理与计算技术来模拟和完成人类的认识、理解过程。一般而言，一个图像识别系统主要由三个部分组成，分别是：图像分割、图像特征提取以及分类器的识别分类。

其中，图像分割将图像划分为多个有意义的区域，然后将每个区域的图像进行特征提取，最后分类器根据提取的图像特征对图像进行相对应的分类。实际上，图像识别和图像分割并不存在严格的界限。从某种意义上，图像分割的过程就是图像识别的过程。图像分割着重于对象和背景的关系，研究的是对象在特定背景下所表现出来的整体属性，而图像识别则着重于对象本身的属性。

图像识别的研究现状

图像识别的发展经历了三个阶段：文字识别、数字图像处理与识别、物体识别。

图像识别作为计算视觉技术体系中的重要一环，一直备受重视。微软在两年前就公布了一项里程碑式的成果：它的图像系统识别图片的错误率比人类还要低。如今，图像识别技术又发展到一个新高度。这有赖于更多数据的开放、更多基础工具的开源、产业链的更新迭代，以及高性能的AI计算芯片、深度摄像头和优秀的深度学习算法等的进步，这些都为图像识别技术向更深处发展提供了源源不断的动力。

其实对于图像识别技术，大家已经不陌生，人脸识别、虹膜识别、指纹识别等都属于这个范畴，但是图像识别远不只如此，它涵盖了生物识别、物体与场景识别、视频识别三大类。发展至今，尽管与理想还相距甚远，但日渐成熟的图像识别技术已开始探索在各类行业的应用。

Android图像识别相关技术

1. OpenCV
   基于BSD许可（开源）发行的跨平台计算机视觉库，可以运行在Linux、Windows、Android和Mac OS操作系统上。
   轻量级而且高效——由一系列 C 函数和少量 C++ 类构成，同时提供了Python、Ruby、MATLAB等语言的接口，实现了图像处理和计算机视觉方面的很多通用算法
2. TensorFlow
   TensorFlow是一个深度学习框架，支持Linux平台，Windows平台，Mac平台，甚至手机移动设备等各种平台。
   TensorFlow提供了非常丰富的深度学习相关的API，可以说目前所有深度学习框架里，提供的API最全的，包括基本的向量矩阵计算、各种优化算法、各种卷积神经网络和循环神经网络基本单元的实现、以及可视化的辅助工具、等等。
3. YOLO
   YOLO (You Only Look Once)是一种快速和准确的实时对象检测算法。
   YOLOv3 在 TensorFlow 中实现的完整数据管道。它可用在数据集上来训练和评估自己的目标检测模型。
4. ……

基于OpenCV实现

介绍使用OpenCV来实现指定图像识别的DEMO：

实现思路

①打开应用的同时开启摄像头
②对实时摄像头拍摄的图像封装成MAT对象进行逐帧比对：

1. 获取目标特征并针对各特征集获取描述符
2. 获取两个描述符集合间的匹配项
3. 获取参考图像和空间匹配图像间的单应性
4. 当图像矩阵符合单应性时，绘制跟踪图像的轮廓线

代码部分

权限设置

AndroidMainifest.xml

	<uses-permission android:name="android.permission.CAMERA" />
    <uses-permission android:name="android.permission.WRITE_EXTERNAL_STORAGE" />
    <uses-permission android:name="android.permission.READ_EXTERNAL_STORAGE" />

    <uses-feature android:name="android.hardware.camera" />
    <uses-feature
        android:name="android.hardware.camera.autofocus"
        android:required="false" />
    <uses-feature
        android:name="android.hardware.camera.flash"
        android:required="false" />

权限提示方法

private void requestPermissions() {
        final int REQUEST_CODE = 1;
        if (ContextCompat.checkSelfPermission(this, Manifest.permission.CAMERA) != PackageManager.PERMISSION_GRANTED) {
            ActivityCompat.requestPermissions(this, new String[]{
                            Manifest.permission.CAMERA, Manifest.permission.WRITE_EXTERNAL_STORAGE},
                    REQUEST_CODE);
        }
    }

界面设计

activity_img_recognition.xml

<RelativeLayout xmlns:android="http://schemas.android.com/apk/res/android"
    xmlns:opencv="http://schemas.android.com/apk/res-auto"
    xmlns:tools="http://schemas.android.com/tools"
    android:id="@+id/activity_img_recognition"
    android:layout_width="match_parent"
    android:layout_height="match_parent"
    tools:context="com.sueed.imagerecognition.CameraActivity">

    <org.opencv.android.JavaCameraView
        android:id="@+id/jcv"
        android:layout_width="match_parent"
        android:layout_height="match_parent"
        android:visibility="gone"
        opencv:camera_id="any"
        opencv:show_fps="true" />
RelativeLayout>

主要逻辑代码

CameraActivity.java 【相机启动获取图像和包装MAT相关】

因为OpenCV中JavaCameraView继承自SurfaceView，若有需要可以自定义编写extends SurfaceView implements SurfaceHolder.Callback的xxxSurfaceView替换使用。

package com.sueed.imagerecognition;

import android.Manifest;
import android.content.Intent;
import android.content.pm.PackageManager;
import android.os.Bundle;
import android.util.Log;
import android.view.Menu;
import android.view.MenuItem;
import android.view.SurfaceView;
import android.view.View;
import android.view.WindowManager;
import android.widget.ImageView;
import android.widget.RelativeLayout;
import android.widget.Toast;

import androidx.appcompat.app.AppCompatActivity;
import androidx.core.app.ActivityCompat;
import androidx.core.content.ContextCompat;

import com.sueed.imagerecognition.filters.Filter;
import com.sueed.imagerecognition.filters.NoneFilter;
import com.sueed.imagerecognition.filters.ar.ImageDetectionFilter;
import com.sueed.imagerecognition.imagerecognition.R;

import org.opencv.android.CameraBridgeViewBase;
import org.opencv.android.CameraBridgeViewBase.CvCameraViewFrame;
import org.opencv.android.CameraBridgeViewBase.CvCameraViewListener2;
import org.opencv.android.JavaCameraView;
import org.opencv.android.OpenCVLoader;
import org.opencv.core.Mat;

import java.io.IOException;

// Use the deprecated Camera class.
@SuppressWarnings("deprecation")
public final class CameraActivity extends AppCompatActivity implements CvCameraViewListener2 {

    // A tag for log output.
    private static final String TAG = CameraActivity.class.getSimpleName();

    // The filters.
    private Filter[] mImageDetectionFilters;

    // The indices of the active filters.
    private int mImageDetectionFilterIndex;

    // The camera view.
    private CameraBridgeViewBase mCameraView;

    @Override
    protected void onCreate(final Bundle savedInstanceState) {
        super.onCreate(savedInstanceState);
        getWindow().addFlags(WindowManager.LayoutParams.FLAG_KEEP_SCREEN_ON);
        //init CameraView
        mCameraView = new JavaCameraView(this, 0);
        mCameraView.setMaxFrameSize(size.MaxWidth, size.MaxHeight);
        mCameraView.setCvCameraViewListener(this);
        setContentView(mCameraView);
        requestPermissions();
        mCameraView.enableView();
    }

    @Override
    public void onPause() {
        if (mCameraView != null) {
            mCameraView.disableView();
        }
        super.onPause();
    }

    @Override
    public void onResume() {
        super.onResume();
        OpenCVLoader.initDebug();
    }

    @Override
    public void onDestroy() {
        if (mCameraView != null) {
            mCameraView.disableView();
        }
        super.onDestroy();
    }

    @Override
    public boolean onCreateOptionsMenu(final Menu menu) {
        getMenuInflater().inflate(R.menu.activity_camera, menu);
        return true;
    }

    @Override
    public boolean onOptionsItemSelected(final MenuItem item) {
        switch (item.getItemId()) {
            case R.id.menu_next_image_detection_filter:
                mImageDetectionFilterIndex++;
                if (mImageDetectionFilters != null && mImageDetectionFilterIndex == mImageDetectionFilters.length) {
                    mImageDetectionFilterIndex = 0;
                }
                return true;

            default:
                return super.onOptionsItemSelected(item);
        }
    }

    @Override
    public void onCameraViewStarted(final int width, final int height) {
        Filter Enkidu = null;
        try {
            Enkidu = new ImageDetectionFilter(CameraActivity.this, R.drawable.enkidu);
        } catch (IOException e) {
            e.printStackTrace();
        }


        Filter akbarHunting = null;
        try {
            akbarHunting = new ImageDetectionFilter(CameraActivity.this, R.drawable.akbar_hunting_with_cheetahs);
        } catch (IOException e) {
            Log.e(TAG, "Failed to load drawable: " + "akbar_hunting_with_cheetahs");
            e.printStackTrace();
        }

        mImageDetectionFilters = new Filter[]{
                new NoneFilter(),
                Enkidu,
                akbarHunting
        };
    }

    @Override
    public void onCameraViewStopped() {
    }

    @Override
    public Mat onCameraFrame(final CvCameraViewFrame inputFrame) {
        final Mat rgba = inputFrame.rgba();

        if (mImageDetectionFilters != null) {
            mImageDetectionFilters[mImageDetectionFilterIndex].apply(rgba, rgba);
        }

        return rgba;
    }

}

ImageRecognitionFilter.java【图像特征过滤比对及绘制追踪绿框】

package com.nummist.secondsight.filters.ar;

import java.io.IOException;
import java.util.ArrayList;
import java.util.List;

import org.opencv.android.Utils;
import org.opencv.calib3d.Calib3d;
import org.opencv.core.Core;
import org.opencv.core.CvType;
import org.opencv.core.DMatch;
import org.opencv.core.KeyPoint;
import org.opencv.core.Mat;
import org.opencv.core.MatOfDMatch;
import org.opencv.core.MatOfKeyPoint;
import org.opencv.core.MatOfPoint;
import org.opencv.core.MatOfPoint2f;
import org.opencv.core.Point;
import org.opencv.core.Scalar;
import org.opencv.features2d.DescriptorExtractor;
import org.opencv.features2d.DescriptorMatcher;
import org.opencv.features2d.FeatureDetector;
import org.opencv.imgcodecs.Imgcodecs;
import org.opencv.imgproc.Imgproc;

import android.content.Context;

import com.nummist.secondsight.filters.Filter;

public final class ImageDetectionFilter implements Filter {

    // The reference image (this detector's target).
    private final Mat mReferenceImage;
    // Features of the reference image.
    private final MatOfKeyPoint mReferenceKeypoints = new MatOfKeyPoint();
    // Descriptors of the reference image's features.
    private final Mat mReferenceDescriptors = new Mat();
    // The corner coordinates of the reference image, in pixels.
    // CvType defines the color depth, number of channels, and
    // channel layout in the image. Here, each point is represented
    // by two 32-bit floats.
    private final Mat mReferenceCorners = new Mat(4, 1, CvType.CV_32FC2);

    // Features of the scene (the current frame).
    private final MatOfKeyPoint mSceneKeypoints = new MatOfKeyPoint();
    // Descriptors of the scene's features.
    private final Mat mSceneDescriptors = new Mat();
    // Tentative corner coordinates detected in the scene, in
    // pixels.
    private final Mat mCandidateSceneCorners = new Mat(4, 1, CvType.CV_32FC2);
    // Good corner coordinates detected in the scene, in pixels.
    private final Mat mSceneCorners = new Mat(0, 0, CvType.CV_32FC2);
    // The good detected corner coordinates, in pixels, as integers.
    private final MatOfPoint mIntSceneCorners = new MatOfPoint();

    // A grayscale version of the scene.
    private final Mat mGraySrc = new Mat();
    // Tentative matches of scene features and reference features.
    private final MatOfDMatch mMatches = new MatOfDMatch();

    // A feature detector, which finds features in images.
    private final FeatureDetector mFeatureDetector = FeatureDetector.create(FeatureDetector.ORB);
    // A descriptor extractor, which creates descriptors of
    // features.
    private final DescriptorExtractor mDescriptorExtractor = DescriptorExtractor.create(DescriptorExtractor.ORB);
    // A descriptor matcher, which matches features based on their
    // descriptors.
    private final DescriptorMatcher mDescriptorMatcher = DescriptorMatcher.create(DescriptorMatcher.BRUTEFORCE_HAMMINGLUT);

    // The color of the outline drawn around the detected image.
    private final Scalar mLineColor = new Scalar(0, 255, 0);

    public ImageDetectionFilter(final Context context, final int referenceImageResourceID) throws IOException {

        // Load the reference image from the app's resources.
        // It is loaded in BGR (blue, green, red) format.
        mReferenceImage = Utils.loadResource(context, referenceImageResourceID, Imgcodecs.CV_LOAD_IMAGE_COLOR);

        // Create grayscale and RGBA versions of the reference image.
        final Mat referenceImageGray = new Mat();
        Imgproc.cvtColor(mReferenceImage, referenceImageGray, Imgproc.COLOR_BGR2GRAY);
        Imgproc.cvtColor(mReferenceImage, mReferenceImage, Imgproc.COLOR_BGR2RGBA);

        // Store the reference image's corner coordinates, in pixels.
        mReferenceCorners.put(0, 0, new double[]{0.0, 0.0});
        mReferenceCorners.put(1, 0, new double[]{referenceImageGray.cols(), 0.0});
        mReferenceCorners.put(2, 0, new double[]{referenceImageGray.cols(), referenceImageGray.rows()});
        mReferenceCorners.put(3, 0, new double[]{0.0, referenceImageGray.rows()});

        // Detect the reference features and compute their
        // descriptors.
        mFeatureDetector.detect(referenceImageGray, mReferenceKeypoints);
        mDescriptorExtractor.compute(referenceImageGray, mReferenceKeypoints, mReferenceDescriptors);
    }

    @Override
    public void apply(final Mat src, final Mat dst) {

        // Convert the scene to grayscale.
        Imgproc.cvtColor(src, mGraySrc, Imgproc.COLOR_RGBA2GRAY);

        // Detect the scene features, compute their descriptors,
        // and match the scene descriptors to reference descriptors.
        mFeatureDetector.detect(mGraySrc, mSceneKeypoints);
        mDescriptorExtractor.compute(mGraySrc, mSceneKeypoints, mSceneDescriptors);
        mDescriptorMatcher.match(mSceneDescriptors, mReferenceDescriptors, mMatches);

        // Attempt to find the target image's corners in the scene.
        findSceneCorners();

        // If the corners have been found, draw an outline around the
        // target image.
        // Else, draw a thumbnail of the target image.
        draw(src, dst);
    }

    private void findSceneCorners() {

        final List<DMatch> matchesList = mMatches.toList();
        if (matchesList.size() < 4) {
            // There are too few matches to find the homography.
            return;
        }

        final List<KeyPoint> referenceKeypointsList = mReferenceKeypoints.toList();
        final List<KeyPoint> sceneKeypointsList = mSceneKeypoints.toList();

        // Calculate the max and min distances between keypoints.
        double maxDist = 0.0;
        double minDist = Double.MAX_VALUE;
        for (final DMatch match : matchesList) {
            final double dist = match.distance;
            if (dist < minDist) {
                minDist = dist;
            }
            if (dist > maxDist) {
                maxDist = dist;
            }
        }

        // The thresholds for minDist are chosen subjectively
        // based on testing. The unit is not related to pixel
        // distances; it is related to the number of failed tests
        // for similarity between the matched descriptors.
        if (minDist > 50.0) {
            // The target is completely lost.
            // Discard any previously found corners.
            mSceneCorners.create(0, 0, mSceneCorners.type());
            return;
        } else if (minDist > 25.0) {
            // The target is lost but maybe it is still close.
            // Keep any previously found corners.
            return;
        }

        // Identify "good" keypoints based on match distance.
        final ArrayList<Point> goodReferencePointsList = new ArrayList<Point>();
        final ArrayList<Point> goodScenePointsList = new ArrayList<Point>();
        final double maxGoodMatchDist = 1.75 * minDist;
        for (final DMatch match : matchesList) {
            if (match.distance < maxGoodMatchDist) {
                goodReferencePointsList.add(referenceKeypointsList.get(match.trainIdx).pt);
                goodScenePointsList.add(sceneKeypointsList.get(match.queryIdx).pt);
            }
        }

        if (goodReferencePointsList.size() < 4 || goodScenePointsList.size() < 4) {
            // There are too few good points to find the homography.
            return;
        }

        // There are enough good points to find the homography.
        // (Otherwise, the method would have already returned.)

        // Convert the matched points to MatOfPoint2f format, as
        // required by the Calib3d.findHomography function.
        final MatOfPoint2f goodReferencePoints = new MatOfPoint2f();
        goodReferencePoints.fromList(goodReferencePointsList);
        final MatOfPoint2f goodScenePoints = new MatOfPoint2f();
        goodScenePoints.fromList(goodScenePointsList);

        // Find the homography.
        final Mat homography = Calib3d.findHomography(goodReferencePoints, goodScenePoints);

        // Use the homography to project the reference corner
        // coordinates into scene coordinates.
        Core.perspectiveTransform(mReferenceCorners, mCandidateSceneCorners, homography);

        // Convert the scene corners to integer format, as required
        // by the Imgproc.isContourConvex function.
        mCandidateSceneCorners.convertTo(mIntSceneCorners, CvType.CV_32S);

        // Check whether the corners form a convex polygon. If not,
        // (that is, if the corners form a concave polygon), the
        // detection result is invalid because no real perspective can
        // make the corners of a rectangular image look like a concave
        // polygon!
        if (Imgproc.isContourConvex(mIntSceneCorners)) {
            // The corners form a convex polygon, so record them as
            // valid scene corners.
            mCandidateSceneCorners.copyTo(mSceneCorners);
        }
    }

    protected void draw(final Mat src, final Mat dst) {

        if (dst != src) {
            src.copyTo(dst);
        }

        if (mSceneCorners.height() < 4) {
            // The target has not been found.

            // Draw a thumbnail of the target in the upper-left
            // corner so that the user knows what it is.

            // Compute the thumbnail's larger dimension as half the
            // video frame's smaller dimension.
            int height = mReferenceImage.height();
            int width = mReferenceImage.width();
            final int maxDimension = Math.min(dst.width(), dst.height()) / 2;
            final double aspectRatio = width / (double) height;
            if (height > width) {
                height = maxDimension;
                width = (int) (height * aspectRatio);
            } else {
                width = maxDimension;
                height = (int) (width / aspectRatio);
            }

            // Select the region of interest (ROI) where the thumbnail
            // will be drawn.
            final Mat dstROI = dst.submat(0, height, 0, width);

            // Copy a resized reference image into the ROI.
            Imgproc.resize(mReferenceImage, dstROI, dstROI.size(), 0.0, 0.0, Imgproc.INTER_AREA);

            return;
        }

        // Outline the found target in green.
        Imgproc.line(dst, new Point(mSceneCorners.get(0, 0)), new Point(mSceneCorners.get(1, 0)), mLineColor, 4);
        Imgproc.line(dst, new Point(mSceneCorners.get(1, 0)), new Point(mSceneCorners.get(2, 0)), mLineColor, 4);
        Imgproc.line(dst, new Point(mSceneCorners.get(2, 0)), new Point(mSceneCorners.get(3, 0)), mLineColor, 4);
        Imgproc.line(dst, new Point(mSceneCorners.get(3, 0)), new Point(mSceneCorners.get(0, 0)), mLineColor, 4);
    }
}

实现效果图

确认允许权限：
[外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传(img-Ii2Kg1dH-1609501715043)(https://i.ibb.co/8db22TB/bo1.jpg =400x800)] [外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传(img-7Oy1osjx-1609501715044)(https://i.ibb.co/Kzfr6gG/bo2.jpg =400x800)]
实时追踪指定图像
[外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传(img-apk42FOU-1609501715045)(https://i.ibb.co/fXDKgtk/bo3.jpg =400x800)][外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传(img-3S9pNQm0-1609501715048)(https://i.ibb.co/BCk9gXs/bo4.jpg =400x800)]

结语

本文只实现了需要提供完整原图进行比对才能实现图像识别，还有许多更加智能方便的识别技术和方法，比如：HOG、SIFT、SURF 等方法经由正负样本库进行训练后可以从图像中提取一些特征，并通过特征确定物体类别。OpenCV库中也仍有很大一部分的功能在本文中未能进行实践，亟待今后继续探索和研究。

作者：陈舒婷

原文链接：https://blog.csdn.net/qq_42871911/article/details/111460677