Reliable AdaBoost Classification in Joint Feature Spaces
Liu Tian-Jian, Xu Ping · 2012
Adaboost is used to select and combine weak classifiers from a very large pool of weak classifiers and it has been proven to be very successful for detecting faces. We follow the approach and applied it to detect rear views of cars. The detector was carefully examined and was expanded in a number of ways, such as feature quantization, learning algorithm in joint feature spaces. By using Adaboost in joint feature spaces, a reliable and fast classification is got. Experiment shows this classification perform good hitting rate and litter false positive rate than traditional AdaBoost algorithm.