iOS人脸识别、自定义相机、图像扫描、系统自带二维码识别

作者: abb266389fd0 | 来源:发表于2016-08-15 14:48 被阅读12071次

前段时间遇到一个需求，需要实时扫描图像，并且设定摄像头的尺寸为1080x1920，然后从中间截取出512x512的图片传给第三方SDK做进一步业务处理，直到SDK返回正确的处理结果。

一顿Google，发现图像预览、人脸识别、二维码识别这些苹果都帮我们做好了，而且它们都是基于AVFoundation框架实现的。

话不多说，上代码~！

用到的类，主要有这些：

//硬件设备
@property (nonatomic, strong) AVCaptureDevice *device;
//输入流
@property (nonatomic, strong) AVCaptureDeviceInput *input;
//协调输入输出流的数据
@property (nonatomic, strong) AVCaptureSession *session;
//预览层
@property (nonatomic, strong) AVCaptureVideoPreviewLayer *previewLayer;

//输出流
@property (nonatomic, strong) AVCaptureStillImageOutput *stillImageOutput;  //用于捕捉静态图片
@property (nonatomic, strong) AVCaptureVideoDataOutput *videoDataOutput;    //原始视频帧，用于获取实时图像以及视频录制
@property (nonatomic, strong) AVCaptureMetadataOutput *metadataOutput;      //用于二维码识别以及人脸识别

首先我们需要在手机上把图像显示出来

1.1 获取硬件设备

-(AVCaptureDevice *)device{
    if (_device == nil) {
        _device = [AVCaptureDevice defaultDeviceWithMediaType:AVMediaTypeVideo];
        if ([_device lockForConfiguration:nil]) {
            //自动闪光灯
            if ([_device isFlashModeSupported:AVCaptureFlashModeAuto]) {
                [_device setFlashMode:AVCaptureFlashModeAuto];
            }
            //自动白平衡
            if ([_device isWhiteBalanceModeSupported:AVCaptureWhiteBalanceModeContinuousAutoWhiteBalance]) {
                [_device setWhiteBalanceMode:AVCaptureWhiteBalanceModeContinuousAutoWhiteBalance];
            }
            //自动对焦
            if ([_device isFocusModeSupported:AVCaptureFocusModeContinuousAutoFocus]) {
                [_device setFocusMode:AVCaptureFocusModeContinuousAutoFocus];
            }
            //自动曝光
            if ([_device isExposureModeSupported:AVCaptureExposureModeContinuousAutoExposure]) {
                [_device setExposureMode:AVCaptureExposureModeContinuousAutoExposure];
            }
            [_device unlockForConfiguration];
        }
    }
    return _device;
}

device有很多属性可以调整(注意调整device属性的时候需要上锁, 调整完再解锁)：
闪光灯

typedef NS_ENUM(NSInteger, AVCaptureFlashMode) {
 AVCaptureFlashModeOff  = 0,
 AVCaptureFlashModeOn   = 1,
 AVCaptureFlashModeAuto = 2
} NS_AVAILABLE(10_7, 4_0) __TVOS_PROHIBITED;

前后置摄像头

typedef NS_ENUM(NSInteger, AVCaptureDevicePosition) {
 AVCaptureDevicePositionUnspecified         = 0,
 AVCaptureDevicePositionBack                = 1,
 AVCaptureDevicePositionFront               = 2
} NS_AVAILABLE(10_7, 4_0) __TVOS_PROHIBITED;

手电筒

typedef NS_ENUM(NSInteger, AVCaptureTorchMode) {
 AVCaptureTorchModeOff  = 0,
 AVCaptureTorchModeOn   = 1,
 AVCaptureTorchModeAuto = 2,
} NS_AVAILABLE(10_7, 4_0) __TVOS_PROHIBITED;

对焦

typedef NS_ENUM(NSInteger, AVCaptureFocusMode) {
 AVCaptureFocusModeLocked              = 0,
 AVCaptureFocusModeAutoFocus           = 1,
 AVCaptureFocusModeContinuousAutoFocus = 2,
} NS_AVAILABLE(10_7, 4_0) __TVOS_PROHIBITED;

曝光

typedef NS_ENUM(NSInteger, AVCaptureExposureMode) {
 AVCaptureExposureModeLocked                            = 0,
 AVCaptureExposureModeAutoExpose                        = 1,
 AVCaptureExposureModeContinuousAutoExposure            = 2,
 AVCaptureExposureModeCustom NS_ENUM_AVAILABLE_IOS(8_0) = 3,
} NS_AVAILABLE(10_7, 4_0) __TVOS_PROHIBITED;

白平衡

typedef NS_ENUM(NSInteger, AVCaptureWhiteBalanceMode) {
 AVCaptureWhiteBalanceModeLocked                      = 0,
 AVCaptureWhiteBalanceModeAutoWhiteBalance            = 1,
 AVCaptureWhiteBalanceModeContinuousAutoWhiteBalance  = 2,
} NS_AVAILABLE(10_7, 4_0) __TVOS_PROHIBITED;

1.2 获取硬件的输入流

-(AVCaptureDeviceInput *)input{
    if (_input == nil) {
        _input = [[AVCaptureDeviceInput alloc] initWithDevice:self.device error:nil];
    }
    return _input;
}

创建输入流的时候，会弹出alert向用户获取相机权限
1.3 需要一个用来协调输入和输出数据的会话，然后把input添加到会话中

-(AVCaptureSession *)session{
    if (_session == nil) {
        _session = [[AVCaptureSession alloc] init];
        if ([_session canAddInput:self.input]) {
            [_session addInput:self.input];
        }
    }
    return _session;
}

1.4 然后我们需要一个预览图像的层

-(AVCaptureVideoPreviewLayer *)previewLayer{
    if (_previewLayer == nil) {
        _previewLayer = [[AVCaptureVideoPreviewLayer alloc] initWithSession:self.session];
        _previewLayer.frame = self.view.layer.bounds;
    }
    return _previewLayer;
}

1.5 最后把previewLayer添加到self.view.layer上

[self.view.layer addSublayer:self.previewLayer];

1.6 找个合适的位置，让session运行起来，比如viewWillAppear

-(void)viewWillAppear:(BOOL)animated{
    [super viewWillAppear:animated];
    [self.session startRunning];
}

搞一个按钮用来控制手电筒

#pragma mark - 手电筒
-(void)openTorch:(UIButton*)button{
    button.selected = !button.selected;
    Class captureDeviceClass = NSClassFromString(@"AVCaptureDevice");
    if (captureDeviceClass != nil) {
        if ([self.device hasTorch] && [self.device hasFlash]){
            [self.device lockForConfiguration:nil];
            if (button.selected) {
                [self.device setTorchMode:AVCaptureTorchModeOn];
            } else {
                [self.device setTorchMode:AVCaptureTorchModeOff];
            }
            [self.device unlockForConfiguration];
        }
    }
}

再搞一个按钮来切换前后置摄像头

#pragma mark - 切换前后摄像头
-(void)switchCamera{
    NSUInteger cameraCount = [[AVCaptureDevice devicesWithMediaType:AVMediaTypeVideo] count];
    if (cameraCount > 1) {
        AVCaptureDevice *newCamera = nil;
        AVCaptureDeviceInput *newInput = nil;
        AVCaptureDevicePosition position = [[self.input device] position];
        if (position == AVCaptureDevicePositionFront){
            newCamera = [self cameraWithPosition:AVCaptureDevicePositionBack];
        }else {
            newCamera = [self cameraWithPosition:AVCaptureDevicePositionFront];
        }
        newInput = [AVCaptureDeviceInput deviceInputWithDevice:newCamera error:nil];
        if (newInput != nil) {
            [self.session beginConfiguration];
            [self.session removeInput:self.input];
            if ([self.session canAddInput:newInput]) {
                [self.session addInput:newInput];
                self.input = newInput;
            }else {
                [self.session addInput:self.input];
            }
            [self.session commitConfiguration];
        }
    }
}

-(AVCaptureDevice *)cameraWithPosition:(AVCaptureDevicePosition)position{
    NSArray *devices = [AVCaptureDevice devicesWithMediaType:AVMediaTypeVideo];
    for ( AVCaptureDevice *device in devices )
        if ( device.position == position ) return device;
    return nil;
}

使用AVCaptureStillImageOutput捕获静态图片

4.1 创建一个AVCaptureStillImageOutput对象

-(AVCaptureStillImageOutput *)stillImageOutput{
    if (_stillImageOutput == nil) {
        _stillImageOutput = [[AVCaptureStillImageOutput alloc] init];
    }
    return _stillImageOutput;
}

4.2 将stillImageOutput添加到session中

if ([_session canAddOutput:self.stillImageOutput]) {
       [_session addOutput:self.stillImageOutput];
 }

4.3 搞个拍照按钮，截取静态图片

//AVCaptureStillImageOutput截取静态图片，会有快门声
-(void)screenshot{
    AVCaptureConnection * videoConnection = [self.stillImageOutput connectionWithMediaType:AVMediaTypeVideo];
    if (!videoConnection) {
        NSLog(@"take photo failed!");
        return;
    }
    
    [self.stillImageOutput captureStillImageAsynchronouslyFromConnection:videoConnection completionHandler:^(CMSampleBufferRef imageDataSampleBuffer, NSError *error) {
        if (imageDataSampleBuffer == NULL) {
            return;
        }
        NSData * imageData = [AVCaptureStillImageOutput jpegStillImageNSDataRepresentation:imageDataSampleBuffer];
        UIImage *image = [UIImage imageWithData:imageData];
        [self saveImageToPhotoAlbum:image];
    }];
}

使用AVCaptureVideoOutput实时获取预览图像，这也是楼主的项目需求所在

5.1 创建AVCaptureVideoOutput对象

-(AVCaptureVideoDataOutput *)videoDataOutput{
    if (_videoDataOutput == nil) {
        _videoDataOutput = [[AVCaptureVideoDataOutput alloc] init];
        [_videoDataOutput setSampleBufferDelegate:self queue:dispatch_get_main_queue()];
    }
    return _videoDataOutput;
}

5.2 将videoDataOutput添加session中

if ([_session canAddOutput:self.videoDataOutput]) {
       [_session addOutput:self.videoDataOutput];
}

5.3 遵守AVCaptureVideoDataOutputSampleBufferDelegate协议，并实现它的代理方法

#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 *)captureOutput didOutputSampleBuffer:(CMSampleBufferRef)sampleBuffer fromConnection:(AVCaptureConnection *)connection{
        largeImage = [self imageFromSampleBuffer:sampleBuffer];
}

5.4 实现imageFromSampleBuffer:方法，将CMSampleBufferRef转为NSImage

//CMSampleBufferRef转NSImage
-(UIImage *)imageFromSampleBuffer:(CMSampleBufferRef)sampleBuffer{
    // 为媒体数据设置一个CMSampleBuffer的Core Video图像缓存对象
    CVImageBufferRef imageBuffer = CMSampleBufferGetImageBuffer(sampleBuffer);
    // 锁定pixel buffer的基地址
    CVPixelBufferLockBaseAddress(imageBuffer, 0);
    // 得到pixel buffer的基地址
    void *baseAddress = CVPixelBufferGetBaseAddress(imageBuffer);
    // 得到pixel buffer的行字节数
    size_t bytesPerRow = CVPixelBufferGetBytesPerRow(imageBuffer);
    // 得到pixel buffer的宽和高
    size_t width = CVPixelBufferGetWidth(imageBuffer);
    size_t height = CVPixelBufferGetHeight(imageBuffer);
    // 创建一个依赖于设备的RGB颜色空间
    CGColorSpaceRef colorSpace = CGColorSpaceCreateDeviceRGB();
    // 用抽样缓存的数据创建一个位图格式的图形上下文（graphics context）对象
    CGContextRef context = CGBitmapContextCreate(baseAddress, width, height, 8, bytesPerRow, colorSpace, kCGBitmapByteOrder32Little | kCGImageAlphaPremultipliedFirst);
    // 根据这个位图context中的像素数据创建一个Quartz image对象
    CGImageRef quartzImage = CGBitmapContextCreateImage(context);
    // 解锁pixel buffer
    CVPixelBufferUnlockBaseAddress(imageBuffer,0);
    // 释放context和颜色空间
    CGContextRelease(context); CGColorSpaceRelease(colorSpace);
    // 用Quartz image创建一个UIImage对象image
    UIImage *image = [UIImage imageWithCGImage:quartzImage];
    // 释放Quartz image对象
    CGImageRelease(quartzImage);
    return (image);
}

眼看大功告成，结果一运行，创建core graphic上下文的时候报错：

CGBitmapContextCreate: invalid data bytes/row: should be at least 7680 for 8 integer bits/component, 3 components, kCGImageAlphaPremultipliedFirst.**
CGBitmapContextCreateImage: invalid context 0x0. If you want to see the backtrace, please set CG_CONTEXT_SHOW_BACKTRACE environmental variable.**

又是一通Google，发现stackoverflow上很多这种问答，楼主英语不好，又是一通翻译，发现大家都是在说像素组件位数什么的，摸索半天找到解决办法，设置videoDataOutput的像素格式：

[_videoDataOutput setVideoSettings:[NSDictionary dictionaryWithObject:[NSNumber numberWithInt:kCVPixelFormatType_32BGRA] forKey:(id)kCVPixelBufferPixelFormatTypeKey]];

再运行，还有问题，获取到的图片是颠倒的，尼玛，真是多灾多难，不过这个简单，很快找到解决方法，设置一下视频的方向：

#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 *)captureOutput didOutputSampleBuffer:(CMSampleBufferRef)sampleBuffer fromConnection:(AVCaptureConnection *)connection{
    [connection setVideoOrientation:AVCaptureVideoOrientationPortrait];
    largeImage = [self imageFromSampleBuffer:sampleBuffer];
}

5.5 还记得楼主一开始的需求吗，设定摄像头的尺寸为1080x1920，然后从中间截取出512x512的图片传给第三方SDK做进一步业务处理：

[_session setSessionPreset:AVCaptureSessionPreset1920x1080];

smallImage = [largeImage imageCompressTargetSize:CGSizeMake(512.0f, 512.0f)];

到这里为止，楼主的需求就大功告成啦

使用AVCaptureMetadataOutput识别二维码

6.1 创建AVCaptureMetadataOutput对象

-(AVCaptureMetadataOutput *)metadataOutput{
    if (_metadataOutput == nil) {
        _metadataOutput = [[AVCaptureMetadataOutput alloc]init];
        [_metadataOutput setMetadataObjectsDelegate:self queue:dispatch_get_main_queue()];
        //设置扫描区域
        _metadataOutput.rectOfInterest = self.view.bounds;
    }
    return _metadataOutput;
}

6.2 将metadataOutput添加到session中，并且设置扫描类型

if ([_session canAddOutput:self.metadataOutput]) {
       [_session addOutput:self.metadataOutput];
       //设置扫码格式
       self.metadataOutput.metadataObjectTypes = @[
                                                   AVMetadataObjectTypeQRCode,
                                                   AVMetadataObjectTypeEAN13Code,
                                                   AVMetadataObjectTypeEAN8Code,
                                                   AVMetadataObjectTypeCode128Code
                                                   ];
}

6.3 遵守AVCaptureMetadataOutputObjectsDelegate协议，并实现其代理方法

#pragma mark - AVCaptureMetadataOutputObjectsDelegate
-(void)captureOutput:(AVCaptureOutput *)captureOutput didOutputMetadataObjects:(NSArray *)metadataObjects fromConnection:(AVCaptureConnection *)connection{
    if (metadataObjects.count>0) {
        [self.session stopRunning];
        AVMetadataMachineReadableCodeObject *metadataObject = [metadataObjects objectAtIndex :0];
        NSLog(@"二维码内容 ： %@",metadataObject.stringValue);
    }
}

关于人脸识别

人脸识别也是基于AVCaptureMetadataOutput实现的，跟二维码识别的区别在于，扫描类型：

self.metadataOutput.metadataObjectTypes = @[AVMetadataObjectTypeFace];

#pragma mark - AVCaptureMetadataOutputObjectsDelegate
-(void)captureOutput:(AVCaptureOutput *)captureOutput didOutputMetadataObjects:(NSArray *)metadataObjects fromConnection:(AVCaptureConnection *)connection{
    if (metadataObjects.count>0) {
        [self.session stopRunning];
        AVMetadataMachineReadableCodeObject *metadataObject = [metadataObjects objectAtIndex :0];
        if (metadataObject.type == AVMetadataObjectTypeFace) {
            AVMetadataObject *objec = [self.previewLayer transformedMetadataObjectForMetadataObject:metadataObject];
            NSLog(@"%@",objec);
        }
    }
}

至于怎么利用它来实现具体的功能需求，楼主也很方哈，这里有个链接可以参考一下：基于 OpenCV 的人脸识别

好啦，就这么多了，代码在这里，水平有限，有不对的地方还请多多指教
参考资料：iOS 上的相机捕捉

网友评论

Torin76:求demo
5e9e739529af:你好这个在ipad上面会有留白不知道怎么解决，能帮下看看不
littleDad:你最后分享的objc 上的opencv人脸识别那个代码好早了。不能用了。你有其它相关的人脸视频资料嘛。我用的是opencv3.3版本。
kingkong1221:二维码自动放大和灯光根自动开启怎么设置?
翻滚的炒勺2013:写的太棒了!!!
岁月煮酒你与狗:博主你的demo 运行崩，求解释运行版本10.3.1
abb266389fd0:@僵小白 plist 里面需要添加相机和相册的权限
专业bug篡写员:楼主您好，为什么我的5.3的代理方法有时候不执行，有时候执行两次就不执行了呢
042a0e1be73f:谢谢，学习了！
白水灬煮一切:不错楼主，继续努力！太棒了！
夜色下的港湾_主程序:不错哦赏一块
abb266389fd0:@夜色下的港湾_主程序谢赏
cjgioscjg:你
酷酷的七哥哥:还不错
351e3f5fd77e:不错，正需要用
陈晨XX:android的你同事怎么做的啊？分享下？
abb266389fd0:@陈晨XX 不是公司项目
black__jack:非常详细，非常感谢！
lovedrose:mark
酸三角:学习了
清無:图呢。。。
saman0:点赞
垚子:1
萌萌的白天:看了半天都没看到重点。。。就是想看算法。。。
abb266389fd0:@萌萌的白天标题吹牛了哈哈，没什么算法，都是直接用苹果封装好的东西，就是一个简单的学习笔记😄
iOS旭大大:作为小白，我刚刚看到上面的协议，是最近学的，慢慢我会读懂的