iOS7—Day by day—Day18:Detecting Facial Features With CoreImage

这篇文章是天天品尝iOS7甜点系列的一部分，你可以查看完整的系列目录：天天品尝iOS7甜点

Introduction – 介绍

运用AVFoundation和CoreImage,从iOS5开始就介绍了有关脸部识别的功能。在iOS7中，运用CoreImage进行脸部识别已经改进了许多新的识别特性(包括微笑和眨眼).这个API是十分简单易用的，所以我们创建一个应用程序，它运用AVFoundation进行查找脸部，然后在使用CoreImage来对寻找照片中的微笑和闭眼。

本章的实例程序能够在github上面进行访问，访问地址:github.com/ShinobiControls/iOS7-day-by-day

Face detection with AVFoundation – 运用AVFoundation进行面部识别

在Day 16中，我们使用AVFoundation中的AVCatpureMetadataOutput类来寻找和解析二维码。面部识别也是使用同样的方式，和二维码一样都是metadata对象。我们将会创建一个AVCaptureMetadataOutput对象，但是它具有不懂的metadata类型:

 
   AVCaptureMetadataObject *output = [[AVCaptureMetadataOutput alloc] init];
// Have to add the output before settig metadata types
[_session addOutput:output];
// We're only interested in faces
[output setMetadataObjectTypes:@[AVMetadataObjectTypeFace]];
// This VC is the delegate, Please call use on the main queue
[output setMetadataObjectsDelegate:self queue:dispatch_get_main_queue()];
 
 
 

我们和之前一样来实现代理方法:

 
   - (void)captureOutput:(AVCpatureOutput *)captureOutput didOutputMetadataObjects:(NSArray *)metadataObjects fromConnection:(AVCaptureConnection *)connection {
  for (AVMetadataObject *metadataObject in metadataObjects) {
      if ([metadataObject.type isEqualToString:AVMetadataObjectTypeFace]) {
          // Take an image of the face and pass to CoreImage for detection
          AVCaptureConnection *stillConnection = [_stillImageOutput connectionWithMediaType:AVMediaTypeVideo];
          [_stillImageOutput captureStillImageSaynchronouslyFromConnection:stillConnection completionHandler:^(CMSampleBufferRef imageDataSampleBuffer, NSError *error) {
              if (error) {
                  NSLog(@"There was a problem");
                  return;
              }
              
              NSData *jpegData = [AVCaptureStillImageOutput jpegStillImageNSDataRepresentation:imageDataSampleBuffer];
              
              UIImage *smileImage = [UIImage imageWithData:jpegData];
              _previewLayer.hidden = YES;
              [_session stopRunning];
              self.imageView.hidden = NO;
              self.imageView.image = smileyImage;
              self.activityView.hidden = NO;
              self.statusLabel.text = @"Processing";
              self.statusLabel.hidden = NO;
              
              CIImage *image = [CIImage imageWithData:jpegData];
              [self imageContainsSmiles:image callback:^(BOOL happyFace) {
                  if (happyFace) {
                      self.statusLabel.text = @"Happy Face Found!";
                  }else {
                      self.statusLabel.text = @"Not a good photo...";
                  }
                  self.activityView.hidden = YES;
                  self.retakeButton.hidden = NO;
              }];
          }];
      }
  }
}
 
 
 

和二维码一样都是十分相似的，只是现在我们在会话中添加了一个新的输出类型(AVCaptureStillImageOutput)。这就允许我们把拍的照片作为输入源,这就是captureStillImageAsynchronouslyFromConnection:completionHandler:所做的。我们当我们通知AVFoundation来进行脸部识别的时候，我们需要在当前的输入中保持一个image,然后停止会话.

我们创建一个JPEG代表捕获到的图像:

 
   NSData *jpegData = [AVCaptureStillImageOutput jpegStillImageNSDataRepresentation:imageDataSampleBuffer];

现在我们把它放置到一个UIImageView中，然后创建一个CIImage版本，为CoreImage来进行面部特征识别做准备。我们将会来看看你imageContainsSmiles:callback:的方法。

Feature finding with CoreImage – 运用CoreImage来寻找面部特性

CoreImage需要一个CIContext和一个CIDetector:

 
   if (!_ciContext) {
  _ciContext = [CIContext contextWithOptions:nil];
}

if (!_faceDetector) {
  _faceDetector = [CIDetector detectorOfType:CIDetectorTypeFace context:_ciContext options:nil];
}

我们得到查询器来执行它的检索，我们需要调用featuresInImage:options:方法:

 
   NSArray *features = [_faceDetector featuresInImage:image options:@{CIDetectorEyeBlink: @YES, CIDetectorSmile: @YES, CIDetectorImageOrientation: @5];

为了能够让查询器来执行微笑和眨眼的检索，我们需要指定查询器的参数(CIDetectorEyeBlink和CIDetectorSmile)。CoreImage面部查询器是需要指定方向的，因此，我们需要设置查询器的方向来匹配程序中的设计.

现在我们需要循环特性数组(它包含CIFaceFeature对象)，然后我们需要辨别寻找出哪个包含了微笑或者眨眼的动作:

 
   BOOL happyPicture = NO;
if ([features count] > 0) {
  happyPicture = YES;
}
for (CIFeature *feature in features) {
  if ([feature isKindOfClass:[CIFaceFeature class]]) {
      CIFaceFeature *faceFeature = (CIFaceFeature *)feature;
      if (!faceFeature.hasSmile) {
          happyPicture = NO;
      }
      if (faceFeature.leftEyeClosed || faceFeature.rightEyeClosed) {
          happyPicture = NO;
      }
  }
}
 
 
 

最后我们在主线程上面执行一个回调函数:

 
   dispatch_async(dispatch_get_main_queue(), ^{
  callback(happyPicture);
});

我们的回调函数更新标签来描述是不是一个好的图片:

 
   [self imageContainsSmiles:image callback:^(BOOL happyFace) {
  if (happyFace) {
      self.statusLabel.text = @"Happy Face Found!";
  }else {
      self.statusLabel.text = @"Not a good photo...";
  }
  self.activityView.hidden = YES;
  self.retakeButton.hidden = NO;
}];
 
 
 