People Detection in Processing with OpenCV

This is the original OpenCV people detection example ported to Processing with the Java library of OpenCV 2.4.8. It can achieve more than 20 frames per second. Here is a sample test snapshot.
 

 

import processing.video.*;
 
import java.util.*;
import java.nio.*;
 
import org.opencv.core.Core;
import org.opencv.core.Mat;
import org.opencv.core.CvType;
import org.opencv.core.Scalar;
import org.opencv.objdetect.HOGDescriptor;
import org.opencv.core.MatOfRect;
import org.opencv.core.MatOfDouble;
import org.opencv.core.Rect;
import org.opencv.core.Size;
import org.opencv.imgproc.Imgproc;
 
Capture cap;
PImage small;
HOGDescriptor hog;
 
byte [] bArray;
int [] iArray;
int pixCnt1, pixCnt2;
int w, h;
float ratio;
 
void setup() {
  size(640, 480);
  ratio = 0.5;
  w = int(width*ratio);
  h = int(height*ratio);
 
  background(0);
  // Define and initialise the default capture device.
  cap = new Capture(this, width, height);
  cap.start();
 
  // Load the OpenCV native library.
  System.loadLibrary(Core.NATIVE_LIBRARY_NAME);
  println(Core.VERSION);
 
  // pixCnt1 is the number of bytes in the pixel buffer.
  // pixCnt2 is the number of integers in the PImage pixels buffer.
  pixCnt1 = w*h*4;
  pixCnt2 = w*h;
 
  // bArray is the temporary byte array buffer for OpenCV cv::Mat.
  // iArray is the temporary integer array buffer for PImage pixels.
  bArray = new byte[pixCnt1];
  iArray = new int[pixCnt2];
 
  small = createImage(w, h, ARGB);
  hog = new HOGDescriptor();
  hog.setSVMDetector(HOGDescriptor.getDefaultPeopleDetector());
  noFill();
  stroke(255, 255, 0);
}
 
void draw() {
  if (cap.available()) {
    cap.read();
  } 
  else {
    return;
  }
  image(cap, 0, 0);
  // Resize the video to a smaller PImage.
  small.copy(cap, 0, 0, width, height, 0, 0, w, h);
  // Copy the webcam image to the temporary integer array iArray.
  arrayCopy(small.pixels, iArray);
 
  // Define the temporary Java byte and integer buffers. 
  // They share the same storage.
  ByteBuffer bBuf = ByteBuffer.allocate(pixCnt1);
  IntBuffer iBuf = bBuf.asIntBuffer();
 
  // Copy the webcam image to the byte buffer iBuf.
  iBuf.put(iArray);
 
  // Copy the webcam image to the byte array bArray.
  bBuf.get(bArray);
 
  // Create the OpenCV cv::Mat.
  Mat m1 = new Mat(h, w, CvType.CV_8UC4);
 
  // Initialise the matrix m1 with content from bArray.
  m1.put(0, 0, bArray);
  // Prepare the grayscale matrix.
  Mat m3 = new Mat(h, w, CvType.CV_8UC1);
  Imgproc.cvtColor(m1, m3, Imgproc.COLOR_BGRA2GRAY);
 
  MatOfRect found = new MatOfRect();
  MatOfDouble weight = new MatOfDouble();
 
  hog.detectMultiScale(m3, found, weight, 0, new Size(8, 8), new Size(32, 32), 1.05, 2, false);
 
  Rect [] rects = found.toArray();
  if (rects.length > 0) {
    for (int i=0; i<rects.length; i++) {
      rect(rects[i].x/ratio, rects[i].y/ratio, rects[i].width/ratio, rects[i].height/ratio);
    }
  }
  text("Frame Rate: " + round(frameRate), 500, 50);
}

One Reply to “People Detection in Processing with OpenCV”

Comments are closed.