iOS 基于AVFoundation框架搭建人脸识别，并获取识别到的人脸图片

这段时间抽空做了下人脸识别功能，人脸识别有很多第三方的SDK，例如：Face++，腾讯、讯飞、opencv等，但其实iOS 原生已经支持人脸识别，网上也有很多人脸识别的demo，但基本检测到人脸后就没后续操作了，因此这里分享下识别到人脸后获取识别到的人脸图片。

识别的原理是：客户端检测到人脸，然后将识别到的人脸照片请求后台接口，让后台做人脸校验，成功后返回相关信息！

下面给出基于AVFoundation框架搭建人脸检测功能代码:

一、导入 框架，并设置相关代理和属性

#import 
#define kWidth [UIScreen mainScreen].bounds.size.width
#define kHeight [UIScreen mainScreen].bounds.size.height
#define WS(weakSelf) __weak __typeof(&*self) weakSelf = self

@interface FaceViewController ()
@property (nonatomic,strong) AVCaptureSession *session;
@property (nonatomic,strong) AVCaptureVideoPreviewLayer *previewLayer;
@property (nonatomic,strong) AVCaptureDeviceInput*input;
@property (nonatomic,strong) AVCaptureVideoDataOutput *videoOutput;
@property(nonatomic,strong) UIImageView *faceImgView;
@property(nonatomic,assign)BOOL isFirst;

@end

二、界面初始化

- (void)viewDidLoad {
    [super viewDidLoad];
    self.title = @"人脸识别";
    
    _isFirst = YES;
    [self deviceInit];
    [self initUI];
}


-(void)initUI{
    _faceImgView = [[UIImageView alloc] initWithFrame:CGRectMake(kWidth - 120 , 64, 120, 120)];
    _faceImgView.backgroundColor = [UIColor blueColor];
    [self.view addSubview:_faceImgView];
    
    
    UILabel *titleLab = [[UILabel alloc] initWithFrame:CGRectMake(52, 100, kWidth - 108, 18)];
    titleLab.text = @"请对准脸部拍摄  提高认证成功率";
    titleLab.textAlignment = NSTextAlignmentCenter;
    titleLab.textColor = [UIColor redColor];
    titleLab.font = [UIFont systemFontOfSize:17];
    [self.view addSubview:titleLab];
}

三、相机设备初始化

-(void)deviceInit{
    // 获取输入设备（摄像头）
    NSArray *devices = [AVCaptureDeviceDiscoverySession discoverySessionWithDeviceTypes:@[AVCaptureDeviceTypeBuiltInWideAngleCamera] mediaType:AVMediaTypeVideo position:AVCaptureDevicePositionBack].devices;
    AVCaptureDevice *deviceF = devices[0];
    
    // 根据输入设备创建输入对象
    self.input = [[AVCaptureDeviceInput alloc] initWithDevice:deviceF error:nil];
    
    // 设置代理监听输出对象输出的数据
    self.videoOutput = [[AVCaptureVideoDataOutput alloc] init];
    
    // 对实时视频帧进行相关的渲染操作,指定代理
    [_videoOutput setSampleBufferDelegate:self queue:dispatch_get_main_queue()];
    
    self.session = [[AVCaptureSession alloc] init];
    
    // 设置输出质量(高像素输出)
    if ([self.session canSetSessionPreset:AVCaptureSessionPreset640x480]) {
        [self.session setSessionPreset:AVCaptureSessionPreset640x480];
    }
    // 添加输入和输出到会话
    [self.session beginConfiguration];
    
    if ([self.session canAddInput:_input]) {
        [self.session addInput:_input];
    }
    
    if ([self.session canAddOutput:_videoOutput]) {
        [self.session addOutput:_videoOutput];
    }
    
    [self.session commitConfiguration];
    
    AVCaptureSession *session = (AVCaptureSession *)self.session;
    
    //8.创建预览图层
    _previewLayer = [[AVCaptureVideoPreviewLayer alloc] initWithSession:session];
    _previewLayer.videoGravity = AVLayerVideoGravityResizeAspectFill;
    _previewLayer.frame = self.view.bounds;
    [self.view.layer insertSublayer:_previewLayer atIndex:0];
    

    //10. 开始扫描
    [self.session startRunning];

}

四、将CMSampleBufferRef转为NSImage

- (void )imageFromSampleBuffer:(CMSampleBufferRef) sampleBuffer
{
    //CIImage -> CGImageRef -> UIImage
    CVImageBufferRef imageBuffer = CMSampleBufferGetImageBuffer(sampleBuffer);  //拿到缓冲区帧数据
    CIImage *ciImage = [CIImage imageWithCVPixelBuffer:imageBuffer];            //创建CIImage对象
    CIContext *temporaryContext = [CIContext contextWithOptions:nil];           //创建上下文
    
    //识别脸部
    CIDetector *detector=[CIDetector detectorOfType:CIDetectorTypeFace context:temporaryContext options:@{CIDetectorAccuracy: CIDetectorAccuracyLow}]; //CIDetectorAccuracyLow：识别精度低，但识别速度快、性能高
                                                              //CIDetectorAccuracyHigh：识别精度高、但识别速度比较慢
    NSArray *faceArray = [detector featuresInImage:ciImage
                                           options:nil];
    
    //得到人脸图片的尺寸
    if (faceArray.count) {
        NSLog(@"faceArray == %@",faceArray);
        WS(weakSelf);
        for (CIFaceFeature * faceFeature in faceArray) {
           if (faceFeature.hasLeftEyePosition && faceFeature.hasRightEyePosition  && faceFeature.hasMouthPosition) {
                NSLog(@"_isFirst == %d",_isFirst);
               //这个布尔值用于判断检测到人脸后，获取到人脸照片，不用再进行持续检测
                if (_isFirst) {
                    //因为刚开始扫描到的人脸是模糊照片，所以延迟几秒获取
                    dispatch_after(dispatch_time(DISPATCH_TIME_NOW, (int64_t)(2.0 * NSEC_PER_SEC)), dispatch_get_main_queue(), ^{
                        CGImageRef cgImageRef = [temporaryContext createCGImage:ciImage fromRect:faceFeature.bounds];
                        
                        //resultImg即为获得的人脸图片
                        UIImage   *resultImg = [[UIImage alloc] initWithCGImage:cgImageRef scale:0.1 orientation:UIImageOrientationLeftMirrored];
                        
                        //显示人脸图片，这里可以将图片转为NSdata类型后，请求后台接口
                        [self uploadFaceImg:resultImg];
                        //置为NO
                        weakSelf.isFirst = NO;
                    });
                   
                }
           }
        }
    }
}

六、 显示捕捉到的人脸图片

//显示图片，这里可以请求后台接口
-(void)uploadFaceImg:(UIImage *)image{
    _faceImgView.image = image;

     WS(weakSelf);
    //这里设置为2秒后可以进行继续检测
    dispatch_after(dispatch_time(DISPATCH_TIME_NOW, (int64_t)(2.0 * NSEC_PER_SEC)), dispatch_get_main_queue(), ^{
        weakSelf.isFirst = YES;
    });
}

七、实现 AVCaptureVideoDataOutput获取实时图像的代理

#pragma mark - AVCaptureVideoDataOutputSampleBufferDelegate
//AVCaptureVideoDataOutput获取实时图像，这个代理方法的回调频率很快，几乎与手机屏幕的刷新频率一样快
- (void)captureOutput:(AVCaptureOutput*)output didOutputSampleBuffer:(CMSampleBufferRef)sampleBuffer fromConnection:(AVCaptureConnection*)connection{
    [self imageFromSampleBuffer:sampleBuffer];
}

结语：

以上就是人脸识别功能代码, 如有问题请下方留言指正！

如有帮助请支持一下

Demo地址: https://github.com/zhwIdea/FaceDetect

下一篇文章：基于GPUImage框架进行人脸识别