#!/usr/bin/python """ This program is demonstration for face and object detection using haar-like feat ures.

The program finds faces in a camera image or video stream and displays a red box around them. Original C implementation by: ? Python implementation by: Roman Stanchak, James Bowman """ import sys import cv2.cv as cv from optparse import OptionParser # # # # # # # Parameters for haar detection From the API: The default parameters (scale_factor=2, min_neighbors=3, flags=0) are tuned for accurate yet slow object detection. For a faster operation on real video images the settings are: scale_factor=1.2, min_neighbors=2, flags=CV_HAAR_DO_CANNY_PRUNING, min_size=<minimum possible face size

min_size = (20, 20) image_scale = 2 haar_scale = 1.2 min_neighbors = 2 haar_flags = 0 # variables to store average hue and saturation global H global S H=S=0 class getHistValues: # not used right def __init__(self,img,roi=None): if roi: cv.SetImageROI( img._hue, cv.SetImageROI( img._val, cv.SetImageROI( img._sat, now roi ); roi ); roi );

cv.CalcHist([img._hue, img._sat, img._val], img._histHSV, 0 ); (_, _, _, maxBin) = cv.GetMinMaxHistValue( img._histHSV ); # raise offset and multiply bins to best HSV values (h, s, v) = [x + 1 for x in maxBin]; (hv, sv, vv) = ( 180/img._histBins, 255/25, 255/25); values = (h*hv, s*sv, v*vv); if roi: cv.ResetImageROI( img._hue ); cv.ResetImageROI( img._sat ); cv.ResetImageROI( img._val ); return values; class getSkinColor: def __init__(self,img):

cascade = cv.Load("haarcascades\haarcascade_frontalface_alt.xml") rect = (0,0,1,1) # allocate temporary images gray = cv.CreateImage((img.width,img.height), 8, 1) small_img = cv.CreateImage((cv.Round(img.width / image_scale), cv.Round (img.height / image_scale)), 8, 1) # convert color input image to grayscale cv.CvtColor(img, gray, cv.CV_BGR2GRAY) # # scale input image for faster processing cv.Resize(gray, small_img, cv.CV_INTER_LINEAR) cv.EqualizeHist(gray, gray) faces = cv.HaarDetectObjects(gray, cascade, cv.CreateMemStorage(0), haar_scale, min_neighbors, haar_flags, min_ size) if faces: for ((x, y, w, h), n) in faces: print "(x, y, w, h) = (%d, %d, %d, %d)" % (x, y, w, h) print "width is %d and height is %d" % (img.width, img.height) faceX faceY faceW faceH = = = = x y w h

# #

reScale = 0.2 horScl = int(faceW * reScale) verScl = int(faceH * reScale) rect = (faceX+horScl,faceY+verScl,faceW-(horScl*2),faceH-(verScl *2)) # the input to cv.HaarDetectObjects was resized, so scale the # bounding box of each face and convert it to two CvPoints pt1 = (int(x), int(y)) pt2 = (int((x + w)), int((y + h))) cv.Rectangle(img, pt1, pt2, cv.RGB(255, 0, 0), 3, 8, 0) cv.ShowImage("result", img) #sets the Region of Interest cv.SetImageROI(img, rect); face = cv.CreateImage(cv.GetSize(img),img.depth,img.nChannels); #copy subimage */ cv.Copy(img, face); #always reset the Region of Interest */ cv.ResetImageROI(img); cv.ShowImage("resultFace", face) faceHSV = cv.CreateImage(cv.GetSize(face), 8, 3) cv.CvtColor(face, faceHSV, cv.CV_BGR2HSV); hueFace = cv.CreateMat(face.height, face.width, cv.CV_8UC1) satFace = cv.CreateMat(face.height, face.width, cv.CV_8UC1)

valFace = cv.CreateMat(face.height, face.width, cv.CV_8UC1) cv.Split(faceHSV, hueFace, satFace, valFace, None) cv.ShowImage("hue", hueFace) #intitialize value to store average Hue value maxi=0 mini=1000 global H global S H = S = 0 for x in range(0, hueFace.height): for y in range(0, hueFace.width): H = H + hueFace[x,y] S = S + satFace[x,y] V = V + valFace[x,y] if hueFace[x,y]>maxi: maxi=hueFace[x,y] if hueFace[x,y]<mini: mini=hueFace[x,y] numPixels = hueFace.height * hueFace.width H = H/numPixels S = S/numPixels V = V/numPixels print "skin color (H, S) = (%f, %f)" % (H,S)

class detect_and_draw: def __init__(self,img): small_img = cv.CreateImage((cv.Round(img.width / image_scale),cv.Round(i mg.height / image_scale)), 8, 3) cv.Resize(img, small_img, cv.CV_INTER_LINEAR) if H==0 and S ==0: getSkinColor(small_img) imgHSV = cv.CreateImage(cv.GetSize(small_img), 8, 3) cv.CvtColor(small_img, imgHSV, cv.CV_BGR2HSV); hueImg = cv.CreateMat(small_img.height, small_img.width, cv.CV_8UC1) satImg = cv.CreateMat(small_img.height, small_img.width, cv.CV_8UC1) valImg = cv.CreateMat(small_img.height, small_img.width, cv.CV_8UC1) cv.Split(imgHSV, hueImg, satImg, valImg, None) cv.ShowImage("hueImg", hueImg) hueTrshld = cv.CreateMat(hueImg.height, hueImg.width, cv.CV_8UC1) hueDiff = 30 satDiff = 80 for x in range(0, hueTrshld.height): for y in range(0, hueTrshld.width): hueTrshld[x,y] = 0 if hueImg[x,y]>(H-hueDiff) and hueImg[x,y]>(1) and hueImg[x,y]<( H+hueDiff): if satImg[x,y]>(S-satDiff) and satImg[x,y]<(S+satDiff): hueTrshld[x,y] = 255

hueTrshldErode = cv.CreateMat(hueImg.height, hueImg.width, cv.CV_8UC1) hueTrshldDilate = cv.CreateMat(hueImg.height, hueImg.width, cv.CV_8UC1)

kernel10 = cv.CreateStructuringElementEx(10,10,0,0, cv.CV_SHAPE_RECT) kernel8 = cv.CreateStructuringElementEx(8,8,0,0, cv.CV_SHAPE_RECT) kernel6 = cv.CreateStructuringElementEx(6,6,0,0, cv.CV_SHAPE_RECT) kernel4 = cv.CreateStructuringElementEx(4,4,0,0, cv.CV_SHAPE_RECT) cv.Erode(hueTrshld, hueTrshldErode, kernel6, 1) cv.Dilate(hueTrshldErode, hueTrshldDilate, kernel10, 1) cv.ShowImage("hueTrshldOr", hueTrshld) #original cv.ShowImage("hueTrshldDi", hueTrshldDilate) #dilated cv.ShowImage("hueTrshldEr", hueTrshldErode) #eroded if __name__ == '__main__': parser = OptionParser(usage = "usage: %prog [options] [filename|camera_index ]") parser.add_option("-c", "--cascade", action="store", dest="cascade", type="s tr", help="Haar cascade file, default %default", default = "haarcascades\haarcas cade_frontalface_alt.xml") (options, args) = parser.parse_args() # cascade = cv.Load(options.cascade) if len(args) != 1: parser.print_help() sys.exit(1) input_name = args[0] if input_name.isdigit(): capture = cv.CreateCameraCapture(int(input_name)) else: capture = None cv.NamedWindow("result", 1) cv.NamedWindow("resultFace", cv.CV_WINDOW_AUTOSIZE) cv.NamedWindow("hue", cv.CV_WINDOW_AUTOSIZE) cv.NamedWindow("hueImg", cv.CV_WINDOW_AUTOSIZE) cv.NamedWindow("hueTrshldOr", cv.CV_WINDOW_AUTOSIZE) cv.NamedWindow("hueTrshldEr", cv.CV_WINDOW_AUTOSIZE) cv.NamedWindow("hueTrshldDi", cv.CV_WINDOW_AUTOSIZE) if capture: frame_copy = None while True: t = cv.GetTickCount() #start timer frame = cv.QueryFrame(capture) if not frame: cv.WaitKey(0) break if not frame_copy: frame_copy = cv.CreateImage((frame.width,frame.height), cv.IPL_DEPTH_8U, frame.nChannels)

if frame.origin == cv.IPL_ORIGIN_TL: cv.Copy(frame, frame_copy) else: cv.Flip(frame, frame_copy, 0) detect_and_draw(frame_copy) if cv.WaitKey(10) >= 0: break #stop timer and show elapsed time t = cv.GetTickCount() - t print "detection time = %gms" % (t/(cv.GetTickFrequency()*1000.)) else: image = cv.LoadImage(input_name, 1) detect_and_draw(image, cascade) cv.WaitKey(0) cv.DestroyWindow("result") cv.DestroyWindow("resultFace") cv.DestroyWindow("hue") cv.DestroyWindow("hueImg") cv.DestroyWindow("hueTrshldOr") cv.DestroyWindow("hueTrshldEr") cv.DestroyWindow("hueTrshldDi")