FaceDetector

CoreImage之人脸识别OC版

核心代码:

#pragma mark - 识别人脸
- (void)faceDetectWithImage:(UIImage *)image {
    // 图像识别能力：可以在CIDetectorAccuracyHigh(较强的处理能力)与CIDetectorAccuracyLow(较弱的处理能力)中选择
    NSDictionary *opts = [NSDictionary dictionaryWithObject:
                          CIDetectorAccuracyHigh forKey:CIDetectorAccuracy];
    // 将图像转换为CIImage
    CIImage *faceImage = [CIImage imageWithCGImage:image.CGImage];
    CIDetector *faceDetector=[CIDetector detectorOfType:CIDetectorTypeFace context:nil options:opts];
    // 识别出人脸数组
    NSArray *features = [faceDetector featuresInImage:faceImage];
    // 得到图片的尺寸
    CGSize inputImageSize = [faceImage extent].size;
    //将image沿y轴对称
    CGAffineTransform transform = CGAffineTransformScale(CGAffineTransformIdentity, 1, -1);
    //将图片上移
    transform = CGAffineTransformTranslate(transform, 0, -inputImageSize.height);
    
    // 取出所有人脸
    for (CIFaceFeature *faceFeature in features){
        //获取人脸的frame
        CGRect faceViewBounds = CGRectApplyAffineTransform(faceFeature.bounds, transform);
        CGSize viewSize = _imageView.bounds.size;
        CGFloat scale = MIN(viewSize.width / inputImageSize.width,
                            viewSize.height / inputImageSize.height);
        CGFloat offsetX = (viewSize.width - inputImageSize.width * scale) / 2;
        CGFloat offsetY = (viewSize.height - inputImageSize.height * scale) / 2;
        // 缩放
        CGAffineTransform scaleTransform = CGAffineTransformMakeScale(scale, scale);
        // 修正
        faceViewBounds = CGRectApplyAffineTransform(faceViewBounds,scaleTransform);
        faceViewBounds.origin.x += offsetX;
        faceViewBounds.origin.y += offsetY;
    
        //描绘人脸区域
        UIView* faceView = [[UIView alloc] initWithFrame:faceViewBounds];
        faceView.layer.borderWidth = 2;
        faceView.layer.borderColor = [[UIColor redColor] CGColor];
        [_imageView addSubview:faceView];
    
        // 判断是否有左眼位置
        if(faceFeature.hasLeftEyePosition){}
        // 判断是否有右眼位置
        if(faceFeature.hasRightEyePosition){}
        // 判断是否有嘴位置
        if(faceFeature.hasMouthPosition){}
    }
}

原理请参见我的简书 ---iOS之使用CoreImage进行人脸识别, 文章中的代码使用的是Swift编写的，在文章结尾处也给出了下载地址。OC版和Swift版本原理一样，就是代码有些稍微不同，请注意。

VictDog/FaceDetector

FaceDetector