Face Detection in JavaCV using haar classifier

OpenCV provides haar like feature detection algorithm which can be used for object detection. Wikipedia page http://en.wikipedia.org/wiki/Haar-like_features provides nice information about what are haar like feature.

OpenCV also provides haar training utility which can be used for training. It generates XML file from training samples which further can be used for fast object detection. Such XML file is provided with opencv package for face detection.

Gist below explains how to use haar classifier in JavaCV. Code loads classifier file haarcascade_frontalface_default.xml and uses cvHaarDetectObjects() to find faces in loaded image. More information about cvHaarDetectObjects() can be found at http://opencv.willowgarage.com/documentation/object_detection.html#haardetectobjects

Result of this code generates window which shows loaded picture and red rectangle around detected faces.

You can fork complete eclipse project at https://github.com/nikhil9/FaceDetection/

Hough Circle detection in Javacv

Opencv provides Hough circle Detection algorithm which can be used to detect circles. Some information about how algorithm works and its example using Opencv in cpp can be found in below link

http://opencv.itseez.com/doc/tutorials/imgproc/imgtrans/hough_circle/hough_circle.html

We will see how to use cvHoughCircle using Javacv. first we will have to process image to get grayscale or binary image. Using cvSmooth() helps most of the time for good detection however it depending upon kind of object  and background more image processing may be required.

First we will load image and then convert it to grayscale. Then use cvSmooth() to smooth the edges. cvHoughCircle() is used to detect circles and are stored in CvSeq. cvGetSeqElem() is used to extract each circle. We have to use each element in CvPoint3D32f. Center of circle is obtained in CvPoint type using cvPointFrom32f(). Obtained center and radius is used to draw circle on input image using cvCircle.

Following code is a demonstration of all the above processes.


Input Image:

Output Image: