Pedestrian detection based on improved HOG feature and robust adaptive boosting algorithm
Jiefa Wu -, Sheng Yang, Lingling Zhang · 2011
Feature extraction and statistical classification methods are widely used in the object detection procedure. In this paper, improved Histograms of Oriented Gradients (HOG) features are used to represent the edge information of images. After that, HOG and Haar features are extracted to illustrate the performance of different types of features. Furthermore, the decision tree for classification is trained by Gentle Adaboost algorithm which selects some weak learners. Finally, we employ a novel detection method to get an outstanding and visual output. Experiments show that the improved method gets a good performance.