CoreImage 下CIDetector.h自带了四种识别功能
```swift / * 人脸识别 / CORE_IMAGE_EXPORT NSString const CIDetectorTypeFace NS_AVAILABLE(10_7, 5_0);
/ * 矩形边缘识别 / CORE_IMAGE_EXPORT NSString const CIDetectorTypeRectangle NS_AVAILABLE(10_10, 8_0);
/* 二维码识别 / CORE_IMAGE_EXPORT NSString const CIDetectorTypeQRCode NS_AVAILABLE(10_10, 8_0);
/* 文本识别 / #if OBJC2 CORE_IMAGE_EXPORT NSString const CIDetectorTypeText NS_AVAILABLE(10_11, 9_0);
接下来用`CIDetectorTypeRectangle `对图片的矩形状边缘进行识别，效果图如下 （__Demo链接文章底部已给出__）
### 部分代码：
* 初始化一个高精度的识别器
```objc
// 高精度边缘识别器
- (CIDetector *)highAccuracyRectangleDetector
{
static CIDetector *detector = nil;
static dispatch_once_t onceToken;
dispatch_once(&onceToken, ^
{
detector = [CIDetector detectorOfType:CIDetectorTypeRectangle context:nil options:@{CIDetectorAccuracy : CIDetectorAccuracyHigh}];
});
return detector;
}
- 调用照相机捕获摄像头图像
NSArray *possibleDevices = [AVCaptureDevice devicesWithMediaType:AVMediaTypeVideo];
AVCaptureDevice *device = [possibleDevices firstObject];
if (!device) return;
_imageDedectionConfidence = 0.0;
AVCaptureSession *session = [[AVCaptureSession alloc] init];
self.captureSession = session;
[session beginConfiguration];
self.captureDevice = device;
NSError *error = nil;
AVCaptureDeviceInput* input = [AVCaptureDeviceInput deviceInputWithDevice:device error:&error];
session.sessionPreset = AVCaptureSessionPresetPhoto;
[session addInput:input];
AVCaptureVideoDataOutput *dataOutput = [[AVCaptureVideoDataOutput alloc] init];
[dataOutput setAlwaysDiscardsLateVideoFrames:YES];
[dataOutput setVideoSettings:@{(id)kCVPixelBufferPixelFormatTypeKey:@(kCVPixelFormatType_32BGRA)}];
[dataOutput setSampleBufferDelegate:self queue:dispatch_get_main_queue()];
[session addOutput:dataOutput];
self.stillImageOutput = [[AVCaptureStillImageOutput alloc] init];
[session addOutput:self.stillImageOutput];
AVCaptureConnection *connection = [dataOutput.connections firstObject];
[connection setVideoOrientation:AVCaptureVideoOrientationPortrait];
- 还需要有个显示已捕获图像的容器
self.context = [[EAGLContext alloc] initWithAPI:kEAGLRenderingAPIOpenGLES2];
GLKView *view = [[GLKView alloc] initWithFrame:self.bounds];
view.autoresizingMask = UIViewAutoresizingFlexibleWidth | UIViewAutoresizingFlexibleHeight;
view.translatesAutoresizingMaskIntoConstraints = YES;
view.context = self.context;
view.contentScaleFactor = 1.0f;
view.drawableDepthFormat = GLKViewDrawableDepthFormat24;
[self insertSubview:view atIndex:0];
_glkView = view;
glGenRenderbuffers(1, &_renderBuffer);
glBindRenderbuffer(GL_RENDERBUFFER, _renderBuffer);
//图像将绘制进_coreImageContext内
_coreImageContext = [CIContext contextWithEAGLContext:self.context];
[EAGLContext setCurrentContext:self.context];
- 遵循
AVCaptureVideoDataOutputSampleBufferDelegate代理，捕获到图像之后，会调用以下方法
-(void)captureOutput:(AVCaptureOutput *)captureOutput didOutputSampleBuffer:(CMSampleBufferRef)sampleBuffer fromConnection:(AVCaptureConnection *)connection
调用CIDetector 进行识别，并且获取最大不规则四边形
// 从缓冲区中获取CIImage
CVPixelBufferRef pixelBuffer = (CVPixelBufferRef)CMSampleBufferGetImageBuffer(sampleBuffer);
CIImage *image = [CIImage imageWithCVPixelBuffer:pixelBuffer];
// 用高精度边缘识别器 识别特征
NSArray <CIFeature *>*features = [[self highAccuracyRectangleDetector] featuresInImage:image];
// 选取特征列表中最大的不规则四边形
_borderDetectLastRectangleFeature = [self biggestRectangleInRectangles:features];
- 识别到边缘之后使用
CAShapeLayer将边缘绘制并显示
// 绘制边缘检测图层
- (void)drawBorderDetectRectWithImageRect:(CGRect)imageRect topLeft:(CGPoint)topLeft topRight:(CGPoint)topRight bottomLeft:(CGPoint)bottomLeft bottomRight:(CGPoint)bottomRight
{
if (!_rectOverlay) {
_rectOverlay = [CAShapeLayer layer];
_rectOverlay.fillRule = kCAFillRuleEvenOdd;
_rectOverlay.fillColor = [UIColor colorWithRed:73/255.0 green:130/255.0 blue:180/255.0 alpha:0.4].CGColor;
_rectOverlay.strokeColor = [UIColor whiteColor].CGColor;
_rectOverlay.lineWidth = 5.0f;
}
if (!_rectOverlay.superlayer) {
self.layer.masksToBounds = YES;
[self.layer addSublayer:_rectOverlay];
}
// 将图像空间的坐标系转换成uikit坐标系
TransformCIFeatureRect featureRect = [self transfromRealRectWithImageRect:imageRect topLeft:topLeft topRight:topRight bottomLeft:bottomLeft bottomRight:bottomRight];
// 边缘识别路径
UIBezierPath *path = [UIBezierPath new];
[path moveToPoint:featureRect.topLeft];
[path addLineToPoint:featureRect.topRight];
[path addLineToPoint:featureRect.bottomRight];
[path addLineToPoint:featureRect.bottomLeft];
[path closePath];
// 背景遮罩路径
UIBezierPath *rectPath = [UIBezierPath bezierPathWithRect:CGRectMake(-5,
-5,
self.frame.size.width + 10,
self.frame.size.height + 10)];
[rectPath setUsesEvenOddFillRule:YES];
[rectPath appendPath:path];
_rectOverlay.path = rectPath.CGPath;
}
即可显示出实时识别的效果了
- 最后拍照之后的裁剪，使用该滤镜将识别出的不规则四边形转换成矩形，即可转换成正正方方的矩形了
/// 将任意四边形转换成长方形
- (CIImage *)correctPerspectiveForImage:(CIImage *)image withFeatures:(CIRectangleFeature *)rectangleFeature
{
NSMutableDictionary *rectangleCoordinates = [NSMutableDictionary new];
rectangleCoordinates[@"inputTopLeft"] = [CIVector vectorWithCGPoint:rectangleFeature.topLeft];
rectangleCoordinates[@"inputTopRight"] = [CIVector vectorWithCGPoint:rectangleFeature.topRight];
rectangleCoordinates[@"inputBottomLeft"] = [CIVector vectorWithCGPoint:rectangleFeature.bottomLeft];
rectangleCoordinates[@"inputBottomRight"] = [CIVector vectorWithCGPoint:rectangleFeature.bottomRight];
return [image imageByApplyingFilter:@"CIPerspectiveCorrection" withInputParameters:rectangleCoordinates];
}
// TODO : 识别出边缘之后，可以手动设置边缘范围
Demo地址： github传送门：https://github.com/madaoCN/MADRectDetect 好心人点下Star呗