基于Java的源码-jviolajones(人脸检测算法).zip

package detection; import java.awt.Rectangle; import java.awt.image.BufferedImage; import java.awt.Point; import java.io.*; import org.jdom.*; import org.jdom.input.*; import org.jdom.filter.*; import java.util.ArrayList; import java.util.Arrays; import java.util.LinkedList; import java.util.List; import java.util.Iterator; import java.util.Scanner; import javax.imageio.ImageIO; import javax.swing.JFrame; public class Detector { /** The list of classifiers that the test image should pass to be considered as an image.*/ List<Stage> stages; Point size; /**Factory method. Builds a detector from an XML file. * @param filename The XML file (generated by OpenCV) describing the Haar Cascade. * @return The corresponding detector. */ public static Detector create(String filename) { /* Read XML file */ org.jdom.Document document=null; SAXBuilder sxb = new SAXBuilder(); try { document = sxb.build(new File(filename)); } catch(Exception e){e.printStackTrace();} return new Detector(document); } /** Detector constructor. * Builds, from a XML document (i.e. the result of parsing an XML file, the corresponding Haar cascade. * @param document The XML document (parsing of file generated by OpenCV) describing the Haar cascade. */ public Detector(org.jdom.Document document) { /* The detector is constituted by stages, each of them telling whether the considered zone represents the object * with probability a bit greater than 0.5. If a zone passes all stages, it is considered as representing the object.*/ stages=new LinkedList<Stage>(); /* Read the size (in pixels) of the detector. */ Element racine = (Element) document.getRootElement().getChildren().get(0); Scanner scanner = new Scanner(racine.getChild("size").getText()); size = new Point(scanner.nextInt(),scanner.nextInt()); /* Iterate over the stages nodes to read the stages. */ Iterator it=racine.getChild("stages").getChildren("_").iterator(); while(it.hasNext()) { Element stage=(Element)it.next(); /* Read the stage threshold. */ float thres=Float.parseFloat(stage.getChild("stage_threshold").getText()); /*Read all trees of the stage. */ Iterator it2=stage.getChild("trees").getChildren("_").iterator(); Stage st=new Stage(thres); while(it2.hasNext()) { Element tree = ((Element)it2.next()); Tree t = new Tree(); Iterator it4 = tree.getChildren("_").iterator(); while(it4.hasNext()) { Element feature=(Element) it4.next(); float thres2=Float.parseFloat(feature.getChild("threshold").getText()); int left_node=-1; float left_val = 0; boolean has_left_val =false; int right_node=-1; float right_val = 0; boolean has_right_val =false; Element e; if((e=feature.getChild("left_val"))!=null) { left_val = Float.parseFloat(e.getText()); has_left_val=true; } else { left_node = Integer.parseInt(feature.getChild("left_node").getText()); has_left_val=false; } if((e=feature.getChild("right_val"))!=null) { right_val = Float.parseFloat(e.getText()); has_right_val=true; } else { right_node = Integer.parseInt(feature.getChild("right_node").getText()); has_right_val=false; } Feature f = new Feature(thres2,left_val,left_node,has_left_val,right_val,right_node,has_right_val,size); Iterator it3=feature.getChild("feature").getChild("rects").getChildren("_").iterator(); while(it3.hasNext()) { String s = ((Element) it3.next()).getText().trim(); Rect r = Rect.fromString(s); f.add(r); } t.addFeature(f); } st.addTree(t); } stages.add(st); } } /** Returns the list of detected objects in an image applying the Viola-Jones algorithm. * * The algorithm tests, from sliding windows on the image, of variable size, which regions should be considered as searched objects. * Please see Wikipedia for a description of the algorithm. * @param file The image file to scan. * @param baseScale The initial ratio between the window size and the Haar classifier size (default 2). * @param scale_inc The scale increment of the window size, at each step (default 1.25). * @param increment The shift of the window at each sub-step, in terms of percentage of the window size. * @return the list of rectangles containing searched objects, expressed in pixels. */ public List<java.awt.Rectangle> getFaces(String file,float baseScale, float scale_inc,float increment, int min_neighbors,boolean doCannyPruning) { try { BufferedImage image = ImageIO.read(new File(file)); return getFaces(image, baseScale, scale_inc, increment, min_neighbors, doCannyPruning); } catch (IOException e) { e.printStackTrace(); } return null; } public List<java.awt.Rectangle> getFaces(BufferedImage image,float baseScale, float scale_inc,float increment, int min_neighbors,boolean doCannyPruning) { //StopWatch sw = new StopWatch(); //sw.start(); List<Rectangle> ret=new ArrayList<Rectangle>(); int width=image.getWidth(); int height=image.getHeight(); /* Compute the max scale of the detector, i.e. the size of the image divided by the size of the detector. */ float maxScale = (Math.min((width+0.f)/size.x,(height+0.0f)/size.y)); /* Compute the grayscale image, the integral image and the squared integral image.*/ int[][] grayImage=new int[width][height]; int[][] img = new int[width][height]; int[][] squares=new int[width][height]; for(int i=0;i<width;i++) { int col=0; int col2=0; for(int j=0;j<height;j++) { int c = image.getRGB(i,j); int red = (c & 0x00ff0000) >> 16; int green = (c & 0x0000ff00) >> 8; int blue = c & 0x000000ff; int value=(30*red +59*green +11*blue)/100; img[i][j]=value; grayImage[i][j]=(i>0?grayImage[i-1][j]:0)+col+value; squares[i][j]=(i>0?squares[i-1][j]:0)+col2+value*value; col+=value; col2+=value*value; } } /* Eventually compute the gradient of the image, if option is on. */ int[][] canny = null; if(doCannyPruning) canny = CannyPruner.getIntegralCanny(img); /*Heart of the algorithm : detection */ /*For each scale of the detection window */ for(float scale=baseScale;scale<maxScale;scale*=scale_inc) { /*Compute the sliding step of the window*/ int step=(int) (scale*size.x*increment); int size=(int) (scale*this.size.x); /*For each position of the window on the image, check whether the object is detected there.*/ for(int i=0;i<width-size;i+=step) { for(int j=0;j<height-size;j+=step) { /* If Canny pruning is on, compute the edge density of the zone. * If it is too low, the object should not be there so skip the region.*/ if(doCannyPruning) { int edges_density = canny[i+size][j+size]+canny[i][j]-canny[i][j+size]-canny[i+size][j]; int d = edges_density/size/size; if(d<20||d>100) continue; } boolean pass=true; /* Perform each stage of the detector on the window. If one stage fails, the zone is rejected.*/ for(Stage s:stages) { if(!s.pass(grayImage,squares,i,j,scale)) {pass=false; break;} } /* If the window passed all stages, add it to the results. */ if(pass) ret.add(new Rectangle(i,j,size,size)); } } } //sw.print("Single threaded :