Car-pose detection using Randomized WLD
Lei Lei, Yi Hu, Dae-Hwan Kim, Sung-Jea Ko · 2011
In both vehicle detection and vehicle tracking, the orientation of car will provide useful information to predict the trajectory. In this paper, we propose a method to determine the orientation of car in a still image. We train a set of Randomized Weber Local Descriptor (RWLD) based classifiers to overcome this problem. To make the system robust and fast, we also propose a tree structure to organize the classifiers to a pose estimator. We evaluate our method on a database consisting of more than 2000 vehicle images. The experimental results show that our method is effective. This pose estimator can be used for a variety of applications conveniently.