status-im
/
react-native-camera
mirror of https://github.com/status-im/react-native-camera.git
2
0
Fork 0
Code Issues Projects Releases Wiki Activity

feat(Android): add TextBlock-Detection via new prop objectsToDetect

This commit is contained in:
Malte Peters 2018-03-21 16:28:43 +01:00
parent 7acd1f23db
commit 5dfa6d2d43
4 changed files with 242 additions and 72 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
android/src/main/java/org/reactnative/camera/CameraViewManager.java
View File

@ -119,6 +119,11 @@ public class CameraViewManager extends ViewGroupManager<RNCameraView> {
view.setFaceDetectionExpectedOrientation(faceDetectionExpectedOrientation);
}
@ReactProp(name = "objectsToDetect")
public void updateObjectsToDetect(RNCameraView view, int objectsToDetect) {
view.updateObjectsToDetect(objectsToDetect);
}
@ReactProp(name = "faceDetectorEnabled")
public void setFaceDetecting(RNCameraView view, boolean faceDetectorEnabled) {
view.setShouldDetectFaces(faceDetectorEnabled);

8
android/src/main/java/org/reactnative/camera/RNCameraView.java
View File

@ -62,6 +62,7 @@ public class RNCameraView extends CameraView implements LifecycleEventListener,
private boolean mShouldDetectFaces = false;
private boolean mShouldScanBarCodes = false;
private int mFaceDetectionExpectedOrientation = -1;
private int mObjectsToDetect = 0;
private int mFaceDetectorMode = RNFaceDetector.FAST_MODE;
private int mFaceDetectionLandmarks = RNFaceDetector.NO_LANDMARKS;
private int mFaceDetectionClassifications = RNFaceDetector.NO_CLASSIFICATIONS;
@ -251,6 +252,13 @@ public class RNCameraView extends CameraView implements LifecycleEventListener,
}
}
public void updateObjectsToDetect(int objectsToDetect){
mObjectsToDetect = objectsToDetect;
if(openCVProcessor != null){
openCVProcessor.updateObjectsToDetect(objectsToDetect);
}
}
public void setFaceDetectionLandmarks(int landmarks) {
mFaceDetectionLandmarks = landmarks;
// if (mFaceDetector != null) {

205
android/src/main/java/org/reactnative/opencv/OpenCVProcessor.java
View File

@ -11,24 +11,37 @@ import java.io.File;
import java.io.FileOutputStream;
import java.io.IOException;
import java.io.InputStream;
import java.util.ArrayList;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
import org.opencv.core.Core;
import org.opencv.core.CvType;
import org.opencv.core.Mat;
import org.opencv.core.MatOfInt4;
import org.opencv.core.MatOfPoint;
import org.opencv.core.MatOfPoint2f;
import org.opencv.core.MatOfRect;
import org.opencv.core.Point;
import org.opencv.core.Rect;
import org.opencv.core.RotatedRect;
import org.opencv.core.Scalar;
import org.opencv.core.Size;
import org.opencv.imgcodecs.Imgcodecs;
import org.opencv.imgproc.Imgproc;
import org.opencv.objdetect.CascadeClassifier;
import org.reactnative.camera.R;
import static org.opencv.core.CvType.CV_32F;
import static org.opencv.imgproc.Imgproc.morphologyEx;
public class OpenCVProcessor {
private CascadeClassifier faceDetector;
private int frame = 0;
private Context reactContext;
private int faceDetectionExpectedOrientation = -1;
private int objectsToDetect = 0;
private boolean saveDemoFrame = false;
public OpenCVProcessor(Context context) {
this.reactContext = context;
@ -64,15 +77,7 @@ public class OpenCVProcessor {
Imgcodecs.imwrite("/sdcard/nect/" + String.valueOf(System.currentTimeMillis()) + ".jpg", mat);
}
public SparseArray<Map<String, Float>> detect(byte[] imageData, int width, int height, int rotation) {
SparseArray<Map<String, Float>> faces = new SparseArray();
if (this.frame % 15 == 0) {
Mat mat = new Mat((height / 2) + height, width, CvType.CV_8UC1);
mat.put(0, 0, imageData);
Mat grayMat = new Mat();
Imgproc.cvtColor(mat, grayMat, Imgproc.COLOR_YUV2GRAY_420);
private int rotateImage(Mat image, int rotation){
int imageRotation = 1;
switch(rotation) {
case 90:
@ -112,31 +117,69 @@ public class OpenCVProcessor {
switch(rotationToBeApplied){
case 2:
Core.transpose(grayMat, grayMat);
Core.flip(grayMat, grayMat,1);
Core.transpose(image, image);
Core.flip(image, image,1);
break;
case 3:
Core.flip(grayMat, grayMat,-1);
Core.flip(image, image,-1);
break;
case 0:
Core.transpose(grayMat, grayMat);
Core.flip(grayMat, grayMat,0);
Core.transpose(image, image);
Core.flip(image, image,0);
break;
}
return expectedFaceOrientation;
}
private float resizeImage(Mat image, float width){
float scale = width / image.cols();
Imgproc.resize(image, image, new Size(), scale, scale, 2);
return scale;
}
public SparseArray<Map<String, Float>> detect(byte[] imageData, int width, int height, int rotation) {
SparseArray<Map<String, Float>> objects = new SparseArray();
if (this.frame % 15 == 0) {
Mat mat = new Mat((height / 2) + height, width, CvType.CV_8UC1);
mat.put(0, 0, imageData);
Mat grayMat = new Mat();
Imgproc.cvtColor(mat, grayMat, Imgproc.COLOR_YUV2GRAY_420);
switch(objectsToDetect){
case 0:
objects = detectFaces(grayMat, rotation);
break;
case 1:
objects = detectTextBlocks(grayMat, rotation);
break;
}
}
this.frame++;
return objects;
}
private SparseArray<Map<String, Float>> detectFaces(Mat image, int rotation) {
SparseArray<Map<String, Float>> faces = new SparseArray();
int expectedFaceOrientation = rotateImage(image, rotation);
float imageWidth = 480f;
float scale = imageWidth / grayMat.cols();
float imageHeight = grayMat.rows() * scale;
float scale = resizeImage(image, imageWidth);
Imgproc.resize(grayMat, grayMat, new Size(), scale, scale, 2);
float imageHeight = image.rows();
// if (this.frame == 30) {
// Log.d(ReactConstants.TAG, "---SAVE IMAGE!!--- ");
// saveMatToDisk(grayMat);
// }
// Save Demo Frame
if (saveDemoFrame && this.frame == 30) {
Log.d(ReactConstants.TAG, "---SAVE IMAGE!!--- ");
saveMatToDisk(image);
}
MatOfRect rec = new MatOfRect();
this.faceDetector.detectMultiScale(grayMat, rec, 1.3, 3, 0, new Size(50, 50), new Size());
this.faceDetector.detectMultiScale(image, rec, 1.3, 3, 0, new Size(50, 50), new Size());
Rect[] detectedObjects = rec.toArray();
if (detectedObjects.length > 0) {
@ -152,12 +195,126 @@ public class OpenCVProcessor {
faces.append(i, face);
}
}
}
this.frame++;
return faces;
}
private SparseArray<Map<String, Float>> detectTextBlocks(Mat image, int rotation) {
SparseArray<Map<String, Float>> objects = new SparseArray();
int orientation = rotateImage(image, rotation);
float algorithmWidth = 1080f;
float imageWidth = 480f;
float algorithmScale = imageWidth / algorithmWidth;
float scale = resizeImage(image, imageWidth);
float imageHeight = image.rows();
float rectKernX = 17f * algorithmScale;
float rectKernY = 6f * algorithmScale;
float sqKernXY = 40f * algorithmScale;
float minSize = 3000f * algorithmScale;
float maxSize = 100000f * algorithmScale;
Mat processedImage = image.clone();
// initialize a rectangular and square structuring kernel
//float factor = (float)min(image.rows, image.cols) / 600.;
Mat rectKernel = Imgproc.getStructuringElement(Imgproc.MORPH_RECT, new Size(rectKernX, rectKernY));
Mat rectKernel2 = Imgproc.getStructuringElement(Imgproc.MORPH_RECT, new Size(sqKernXY, (int)(0.666666*sqKernXY)));
// Smooth the image using a 3x3 Gaussian, then apply the blackhat morphological
// operator to find dark regions on a light background
Imgproc.GaussianBlur(processedImage, processedImage, new Size(3, 3), 0);
morphologyEx(processedImage, processedImage, Imgproc.MORPH_BLACKHAT, rectKernel);
// Compute the Scharr gradient of the blackhat image
Mat imageGrad = new Mat();
//Sobel(processedImage, imageGrad, CV_32F, 1, 0, CV_SCHARR);
Imgproc.Sobel(processedImage, imageGrad, CV_32F, 1, 0);
// Core.convertScaleAbs(imageGrad/8, processedImage);
Core.convertScaleAbs(imageGrad, processedImage);
// Apply a closing operation using the rectangular kernel to close gaps in between
// letters, then apply Otsu's thresholding method
morphologyEx(processedImage, processedImage, Imgproc.MORPH_CLOSE, rectKernel);
Imgproc.threshold(processedImage, processedImage, 0, 255, Imgproc.THRESH_BINARY | Imgproc.THRESH_OTSU);
// , 1, 1
Imgproc.erode(processedImage, processedImage, new Mat(), new Point(-1, -1), 2);
// Perform another closing operation, this time using the square kernel to close gaps
// between lines of TextBlocks
morphologyEx(processedImage, processedImage, Imgproc.MORPH_CLOSE, rectKernel2);
// Find contours in the thresholded image and sort them by size
float minContourArea = minSize;
float maxContourArea = maxSize;
// https://github.com/codertimo/Vision/issues/7
// http://answers.opencv.org/question/6206/opencv4android-conversion-from-matofkeypoint-to-matofpoint2f/
List<MatOfPoint> contours = new ArrayList<>();
MatOfInt4 hierarchy = new MatOfInt4();
Imgproc.findContours(processedImage, contours, hierarchy, Imgproc.RETR_EXTERNAL, Imgproc.CHAIN_APPROX_SIMPLE);
// Create a result vector
List<RotatedRect> minRects = new ArrayList<>();
for (int i = 0, I = contours.size(); i < I; ++i) {
// Filter by provided area limits
if (Imgproc.contourArea(contours.get(i)) > minContourArea && Imgproc.contourArea(contours.get(i)) < maxContourArea)
minRects.add(Imgproc.minAreaRect(new MatOfPoint2f(contours.get(i).toArray())));
}
if(saveDemoFrame){
Mat debugDrawing = image.clone();
for (int i = 0, I = minRects.size(); i < I; ++i) {
Point[] rect_points = new Point[4];
minRects.get(i).points( rect_points );
for( int j = 0; j < 4; ++j )
Imgproc.line( debugDrawing, rect_points[j], rect_points[(j+1)%4], new Scalar(255,0,0), 1, 8, 0 );
}
saveMatToDisk(debugDrawing);
}
if (minRects.size() > 0) {
for (int i = 0; i < minRects.size(); i++) {
Point[] rect_points = new Point[4];
minRects.get(i).points( rect_points );
float xRel = (float) rect_points[1].x / imageWidth;
float yRel = (float) rect_points[1].y / imageHeight;
float widthRel = (float) Math.abs(rect_points[3].x - rect_points[1].x) / imageWidth;
float heightRel = (float) Math.abs(rect_points[3].y - rect_points[1].y) / imageHeight;
float sizeRel = Math.abs(widthRel * heightRel);
float ratio = (float) Math.abs(rect_points[3].x - rect_points[1].x) / (float) Math.abs(rect_points[3].y - rect_points[1].y);
// if object large enough
if(sizeRel >= 0.025 & ratio >= 5.5 & ratio <= 8.5){
Map<String, Float> object = new HashMap();
object.put("x", xRel);
object.put("y", yRel);
object.put("width", widthRel);
object.put("height", heightRel);
object.put("orientation", (float) orientation);
objects.append(i, object);
}
}
}
return objects;
}
public void setFaceDetectionExpectedOrientation(int expectedOrientation){
faceDetectionExpectedOrientation = expectedOrientation;
}
public void updateObjectsToDetect(int objToDetect){
objectsToDetect = objToDetect;
}
}

2
ios/RN/OpenCVProcessor.mm
View File

@ -267,7 +267,7 @@
float ratio = fabsf(rect_points[3].x - rect_points[1].x) / fabsf(rect_points[3].y - rect_points[1].y);
// if object large enough
if(sizeRel >= 0.03 & ratio >= 6. & ratio <= 8.){
if(sizeRel >= 0.025 & ratio >= 5.5 & ratio <= 8.5){
NSDictionary *objectDescriptor = @{
@"x" : [NSNumber numberWithFloat:xRel],
@"y" : [NSNumber numberWithFloat:yRel],