Random Forests and Their Applications in Scene Understanding

Produktbeschreibung

In this thesis, ensemble learning methods, such as Random Forest and Adaptive Boosting, are applied to several tasks of computer vision problems. This dissertation is divided into two parts. The first part addresses the problem of recognizing human actions from video data by using spatio-temporal features. Moreover, a novel feature is proposed that eliminates several disadvantages of already existing spatio-temporal features. Finally, the Random Forest classifier is applied to the task of human action recognition from acceleration data using multiple inertial sensors. The second part of the thesis targets at improving the Random Forest algorithm by introducing several modifications that are motivated from other ensemble learning methods e.g. Adaptive Boosting and Support Vector Machines. These contributions result in a more robust and efficient algorithm, regarding several real-world scenarios.

Keywords: 384510, 84510, 845, Fortschritt-Berichte VDI, Reihe 10, Random Forest, Machine Learning, Spatio-temporal features, Scene Understanding, Object Detection, Object Recognition, Human Action Recognition,