1 What is it?
1.1 It’s a Feature-based AdaBoost Machine Learning technique proposed by Paul Viola and Michael Jones in their paper, “Rapid Object Detection using a Boosted Cascade of Simple Features” in 2001
2 How does it work?
2.1 Generate simple features like vertical or horizontal subtraction boxes. The feature is their subtracted values in the boxes.
3 How to solve the computation problem?
3.1 Firstly, transfer the detection window size so that the features to be calculated maintain an acceptable computation cost.
3.2 Secondly, use AdaBoost to select the most typical features among them.
3.3 At last, the algorithm applies a method called Cascade of Classifiers which means organizing the feature detection process in serval (38) stages. In the first serval stages, the algorithm tries to screen out irrelevant area. If the area is ROI, then proceed the detection to next stages. The detection can return False in each stage until it goes through all stages and the algorithm will return True.