Android中的人脸检测(静态跟动态)

（1）背景。

?????? Google 于2006年8月收购Neven Vision 公司 （该公司拥有10多项应用于移动设备领域的图像识别的专利），以此获得了图像识别的技术，并加入到android中。Android 中的人脸识别技术，用到的底层库：android/external/neven/，framework 层：frameworks/base/media/java/android/media/FaceDetector.java。

?????? Java 层接口的限制：A，只能接受Bitmap 格式的数据；B，只能识别双眼距离大于20 像素的人脸像（当然，这个可在framework层中修改）；C，只能检测出人脸的位置（双眼的中心点及距离），不能对人脸进行匹配（查找指定的脸谱）。

??????? 人脸识别技术的应用：A，为Camera 添加人脸识别的功能，使得Camera 的取景器上能标识出人脸范围；如果硬件支持，可以对人脸进行对焦。B，为相册程序添加按人脸索引相册的功能，按人脸索引相册，按人脸分组，搜索相册。

（2）Neven库给上层提供的主要方法：
??????? A，android.media.FaceDetector .FaceDetector(int width, int height, int maxFaces)：Creates a FaceDetector, configured with the size of the images to be analysed and the maximum number of faces that can be detected. These parameters cannot be changed once the object is constructed.

??????? B，int android.media.FaceDetector .findFaces(Bitmap bitmap, Face [] faces)：Finds all the faces found in a given Bitmap . The supplied array is populated with FaceDetector.Face s for each face found. The bitmap must be in 565 format (for now).

（3）?静态图片处理代码实例：

? ? ? ?通过对位图的处理，捕获位图中的人脸，并以绿框显示，有多个人脸就提示多个绿框。首先新建一个activity，由于位图资源会用代码显示出来，所以不需在layout中使用widget。

[java]?view plaincopy

?

1. package?com.example.mydetect2;??
2. ??
3. import?android.os.Bundle;??
4. import?android.app.Activity;??
5. import?android.util.Log;??
6. import?android.view.Menu;??
7. import?android.content.Context;???
8. import?android.graphics.Bitmap;???
9. import?android.graphics.BitmapFactory;???
10. import?android.graphics.Canvas;???
11. import?android.graphics.Color;???
12. import?android.graphics.Paint;???
13. import?android.graphics.PointF;???
14. import?android.media.FaceDetector;??//人脸识别的关键类??
15. import?android.media.FaceDetector.Face;???
16. import?android.view.View;???
17. ??
18. public?class?MainActivity2?extends?Activity?{??
19. ??
20. ????@Override??
21. ????protected?void?onCreate(Bundle?savedInstanceState)?{??
22. ????????super.onCreate(savedInstanceState);??
23. ????????//setContentView(R.layout.activity_main_activity2);??
24. ????????setContentView(new?myView(this));???//使用自建的view来显示??
25. ????????Log.i("zhangcheng","MainActivity2?run?here");??
26. ????}??
27. ??
28. ????private?class?myView?extends?View{??
29. ????????private?int?imageWidth,?imageHeight;??
30. ????????private?int?numberOfFace?=?5;???????//最大检测的人脸数??
31. ????????private?FaceDetector?myFaceDetect;??//人脸识别类的实例??
32. ????????private?FaceDetector.Face[]?myFace;?//存储多张人脸的数组变量??
33. ????????float?myEyesDistance;???????????//两眼之间的距离??
34. ????????int?numberOfFaceDetected;???????//实际检测到的人脸数??
35. ????????Bitmap?myBitmap;??
36. ??
37. ????????public?myView(Context?context){?????//view类的构造函数，必须有??
38. ????????????super(context);???
39. ????????????BitmapFactory.Options?BitmapFactoryOptionsbfo?=?new?BitmapFactory.Options();???
40. ????????????BitmapFactoryOptionsbfo.inPreferredConfig?=?Bitmap.Config.RGB_565;??//构造位图生成的参数，必须为565。类名+enum??
41. ????????????myBitmap?=?BitmapFactory.decodeResource(getResources(),?R.drawable.baby,?BitmapFactoryOptionsbfo);?????
42. ????????????imageWidth?=?myBitmap.getWidth();???
43. ????????????imageHeight?=?myBitmap.getHeight();???
44. ????????????myFace?=?new?FaceDetector.Face[numberOfFace];???????//分配人脸数组空间??
45. ????????????myFaceDetect?=?new?FaceDetector(imageWidth,?imageHeight,?numberOfFace);???
46. ????????????numberOfFaceDetected?=?myFaceDetect.findFaces(myBitmap,?myFace);????//FaceDetector?构造实例并解析人脸??
47. ????????????Log.i("zhangcheng","numberOfFaceDetected?is?"?+?numberOfFaceDetected);??
48. ????????}??
49. ??????????
50. ????????protected?void?onDraw(Canvas?canvas){???????????//override函数，必有??
51. ????????????canvas.drawBitmap(myBitmap,?0,?0,?null);????//画出位图???
52. ????????????Paint?myPaint?=?new?Paint();???
53. ????????????myPaint.setColor(Color.GREEN);???
54. ????????????myPaint.setStyle(Paint.Style.STROKE);???
55. ????????????myPaint.setStrokeWidth(3);??????????//设置位图上paint操作的参数??
56. ??
57. ????????????for(int?i=0;?i?<?numberOfFaceDetected;?i++){??
58. ????????????????Face?face?=?myFace[i];??
59. ????????????????PointF?myMidPoint?=?new?PointF();???
60. ????????????????face.getMidPoint(myMidPoint);???
61. ????????????????myEyesDistance?=?face.eyesDistance();???//得到人脸中心点和眼间距离参数，并对每个人脸进行画框??
62. ????????????????canvas.drawRect(????????????//矩形框的位置参数??
63. ????????????????????????(int)(myMidPoint.x?-?myEyesDistance),???
64. ????????????????????????(int)(myMidPoint.y?-?myEyesDistance),???
65. ????????????????????????(int)(myMidPoint.x?+?myEyesDistance),???
66. ????????????????????????(int)(myMidPoint.y?+?myEyesDistance),???
67. ????????????????????????myPaint);??
68. ????????????}??
69. ????????}??
70. ????}??
71. }??

????? 以上为activity，工程的xml文件没有什么特殊地方。最后得到的结果如下，图片资源是png的也可以。

?

?（4） 动态预览识别人脸代码实例

?????? 该过程用于后台工作，没有界面也没有预览。所以没有采用上面那种处理位图资源的方式。Import的类就不列出了，核心的代码和流程如下：

A，打开摄像头，获得初步摄像头回调数据，用到是setpreviewcallback

protected Camera mCameraDevice = null;// 摄像头对象实例

private long mScanBeginTime = 0;?? // 扫描开始时间
private long mScanEndTime = 0;?? // 扫描结束时间
private long mSpecPreviewTime = 0;?? // 扫描持续时间

private int orientionOfCamera ;?? //前置摄像头layout角度

int numberOfFaceDetected;??? //最终识别人脸数目

[java]?view plaincopy

?

1. public?void?startFaceDetection()?{??
2. ????????try?{??
3. ????????????????mCameraDevice?=?Camera.open(1);?????//打开前置??
4. ????????????????if?(mCameraDevice?!=?null)??
5. ????????????????????Log.i(TAG,?"open?cameradevice?success!?");??
6. ????????????}?catch?(Exception?e)?{?????????????//Exception代替很多具体的异常??
7. ????????????????mCameraDevice?=?null;??
8. ????????????????Log.w(TAG,?"open?cameraFail");??
9. ????????????????mHandler.postDelayed(r,5000);???//如果摄像头被占用，人眼识别每5秒检测看有没有释放前置??
10. ????????????????return;??
11. ????????}???
12. ??????????????
13. ????????Log.i(TAG,?"startFaceDetection");??
14. ????????Camera.Parameters?parameters?=?mCameraDevice.getParameters();??
15. ????????setCameraDisplayOrientation(1,mCameraDevice);??????????????//设置预览方向???
16. ??????????????
17. ????????mCameraDevice.setPreviewCallback(new?PreviewCallback(){??
18. ????????????public?void?onPreviewFrame(byte[]?data,?Camera?camera){??
19. ????????????????mScanEndTime?=?System.currentTimeMillis();???//记录摄像头返回数据的时间??
20. ????????????????mSpecPreviewTime?=?mScanEndTime?-?mScanBeginTime;??//从onPreviewFrame获取摄像头数据的时间??
21. ????????????????Log.i(TAG,?"onPreviewFrame?and?mSpecPreviewTime?=?"?+?String.valueOf(mSpecPreviewTime));??
22. ????????????????Camera.Size?localSize?=?camera.getParameters().getPreviewSize();??//获得预览分辨率??
23. ????????????????YuvImage?localYuvImage?=?new?YuvImage(data,?17,?localSize.width,?localSize.height,?null);??
24. ????????????????ByteArrayOutputStream?localByteArrayOutputStream?=?new?ByteArrayOutputStream();??
25. ????????????????localYuvImage.compressToJpeg(new?Rect(0,?0,?localSize.width,?localSize.height),?80,?localByteArrayOutputStream);????//把摄像头回调数据转成YUV，再按图像尺寸压缩成JPEG，从输出流中转成数组??
26. ????????????????byte[]?arrayOfByte?=?localByteArrayOutputStream.toByteArray();??
27. ????????????????CameraRelease();???//及早释放camera资源，避免影响camera设备的正常调用??
28. ????????????????StoreByteImage(arrayOfByte);??
29. ????????????}??
30. ????????});??
31. ??
32. ????????mCameraDevice.startPreview();?????????//该语句可放在回调后面，当执行到这里，调用前面的setPreviewCallback??
33. ????????mScanBeginTime?=?System.currentTimeMillis();//?记录下系统开始扫描的时间??
34. ????}??

B，设置预览方向的函数说明，该函数比较重要，因为方向直接影响bitmap构造时的矩阵旋转角度，影响最终人脸识别的成功与否

[java]?view plaincopy

?

1. public?void?setCameraDisplayOrientation(int?paramInt,?Camera?paramCamera){??
2. ????????CameraInfo?info?=?new?CameraInfo();??
3. ????????Camera.getCameraInfo(paramInt,?info);??
4. ????????int?rotation?=?((WindowManager)getSystemService("window")).getDefaultDisplay().getRotation();??//获得显示器件角度??
5. ????????int?degrees?=?0;??
6. ????????Log.i(TAG,"getRotation's?rotation?is?"?+?String.valueOf(rotation));??
7. ????????switch?(rotation)?{??
8. ????????????case?Surface.ROTATION_0:?degrees?=?0;?break;??
9. ????????????case?Surface.ROTATION_90:?degrees?=?90;?break;??
10. ????????????case?Surface.ROTATION_180:?degrees?=?180;?break;??
11. ????????????case?Surface.ROTATION_270:?degrees?=?270;?break;??
12. ????????}??
13. ??
14. ????????orientionOfCamera?=?info.orientation;??????//获得摄像头的安装旋转角度??
15. ????????int?result;??
16. ????????if?(info.facing?==?Camera.CameraInfo.CAMERA_FACING_FRONT)?{??
17. ????????????result?=?(info.orientation?+?degrees)?%?360;??
18. ????????????result?=?(360?-?result)?%?360;??//?compensate?the?mirror??
19. ????????}?else?{??//?back-facing??
20. ????????????result?=?(info.orientation?-?degrees?+?360)?%?360;??
21. ????????}??
22. ????????paramCamera.setDisplayOrientation(result);??//注意前后置的处理，前置是映象画面，该段是SDK文档的标准DEMO??
23. ????}??

C，对摄像头回调数据进行转换并最终解成BITMAP后再人脸识别的过程

[java]?view plaincopy

?

1. public?void?StoreByteImage(byte[]?paramArrayOfByte){??
2. ????????mSpecStopTime?=?System.currentTimeMillis();??
3. ????????mSpecCameraTime?=?mSpecStopTime?-?mScanBeginTime;??
4. ???????????????????????
5. ????????Log.i(TAG,?"StoreByteImage?and?mSpecCameraTime?is?"?+?String.valueOf(mSpecCameraTime));??
6. ??
7. ????????BitmapFactory.Options?localOptions?=?new?BitmapFactory.Options();??
8. ????????????Bitmap?localBitmap1?=?BitmapFactory.decodeByteArray(paramArrayOfByte,?0,?paramArrayOfByte.length,?localOptions);??
9. ????????????int?i?=?localBitmap1.getWidth();??
10. ????????????int?j?=?localBitmap1.getHeight();???//从上步解出的JPEG数组中接出BMP，即RAW->JPEG->BMP??
11. ????????????Matrix?localMatrix?=?new?Matrix();??
12. ????????????//int?k?=?cameraResOr;??
13. ????????????Bitmap?localBitmap2?=?null;??
14. ????????????FaceDetector?localFaceDetector?=?null;??
15. ??
16. ????????switch(orientionOfCamera){???//根据前置安装旋转的角度来重新构造BMP??
17. ????????????case?0:??
18. ????????????????localFaceDetector?=?new?FaceDetector(i,?j,?1);??
19. ????????????????????????localMatrix.postRotate(0.0F,?i?/?2,?j?/?2);??
20. ????????????????????????localBitmap2?=?Bitmap.createBitmap(i,?j,?Bitmap.Config.RGB_565);??
21. ????????????????break;??
22. ????????????case?90:??
23. ????????????????localFaceDetector?=?new?FaceDetector(j,?i,?1);???//长宽互换??
24. ????????????????????????localMatrix.postRotate(-270.0F,?j?/?2,?i?/?2);??//正90度的话就反方向转270度，一样效果??
25. ????????????????????????localBitmap2?=?Bitmap.createBitmap(i,?j,?Bitmap.Config.RGB_565);??
26. ????????????????break;????????????????????????
27. ????????????case?180:??
28. ????????????????localFaceDetector?=?new?FaceDetector(i,?j,?1);??
29. ????????????????????????localMatrix.postRotate(-180.0F,?i?/?2,?j?/?2);??
30. ????????????????????????localBitmap2?=?Bitmap.createBitmap(i,?j,?Bitmap.Config.RGB_565);??
31. ????????????????break;??
32. ????????????case?270:??
33. ????????????????localFaceDetector?=?new?FaceDetector(j,?i,?1);??
34. ????????????????????????localMatrix.postRotate(-90.0F,?j?/?2,?i?/?2);??
35. ????????????????????????localBitmap2?=?Bitmap.createBitmap(j,?i,?Bitmap.Config.RGB_565);??//localBitmap2应是没有数据的??
36. ????????????????break;??
37. ????????}??
38. ??
39. ????????FaceDetector.Face[]?arrayOfFace?=?new?FaceDetector.Face[1];??
40. ????????????Paint?localPaint1?=?new?Paint();??
41. ????????????Paint?localPaint2?=?new?Paint();??
42. ????????localPaint1.setDither(true);??
43. ????????????localPaint2.setColor(-65536);??
44. ????????????localPaint2.setStyle(Paint.Style.STROKE);??
45. ????????????localPaint2.setStrokeWidth(2.0F);??
46. ????????????Canvas?localCanvas?=?new?Canvas();??
47. ????????????localCanvas.setBitmap(localBitmap2);??
48. ????????????localCanvas.setMatrix(localMatrix);??
49. ????????????localCanvas.drawBitmap(localBitmap1,?0.0F,?0.0F,?localPaint1);?//该处将localBitmap1和localBitmap2关联（可不要？）??
50. ??
51. ????????numberOfFaceDetected?=?localFaceDetector.findFaces(localBitmap2,?arrayOfFace);?//返回识脸的结果??
52. ????????????localBitmap2.recycle();??
53. ????????????localBitmap1.recycle();???//释放位图资源??
54. ??
55. ????????FaceDetectDeal(numberOfFaceDetected);??
56. ????}??

?

?

参考原文：http://blog.renren.com/share/234025064/6227088558

参考原文：http://www.csdn.net/article/a/2012-04-27/2805150

参考原文：http://www.cnblogs.com/feisky/archive/2010/09/12/1824320.html

?

转自：http://blog.csdn.net/zhandoushi1982/article/details/8613916