Robust Object Detection Based on Decision Trees and a New Cascade Architecture

In this paper, we describe a robust object detection method using decision trees and a new cascade architecture. On the one hand, we design a weak classifier for multi-valued features on AdaBoost algorithm based on decision trees method, which directly reduces training time and increases the object detectionpsilas precision. On the other hand, the use of new cascade architecture is great helpful for the problem of minimizing the false accept rate and a cascadepsilas classification complexity. Experiments on MIT+CMU frontal face data sets and PETS2001 data sets show that the proposed method is comparable to other existing object detection method and outperforms the object detection method proposed by Ivan Laptev.