Haar cascade classifiers is used for objection detection, especially for the the face detection.
Its algorithms needed a lot of positive & negative images from extracting features
Each features is a single values obtained by subtracting sum of pixels under the white rectangles from pixels under black rectangles
Each kernels is calculated gets a lot of the features find the sum of pixels under white & black rectangles ,that features some of relevant & other are irrelevant
Every features should be trained images and classify them positive, negative. May bi some mis classification or occurred,minimax the error rate, accurately classified face ,or non face images
Haar classifiers detecting the edges & lines, it help makes effective face detections.