Advanced driver assistance system

In recent years, pedestrian detection (PD) plays a vital role in a variety of applications such as security cameras, automotive control and so forth. These applications require two essential features, i.e. high speed performance and high accuracy. Firstly, the accuracy is determined by how the image features are described. The image feature description must be robust against occlusion, rotation, and the change in object shapes and illumination conditions. A number of feature descriptors have been proposed. Previously histogram of oriented gradients(HOG) features were extensively used along with support vector machine (SVM) classifier for PD. HOG features and SVM classifier can achieve good performance for PD, but they are time consuming. To achieve high detection speed with good detection performance, a Two-step framework method was proposed. The Two-step framework consists of a full-body detection (FBD) step and a head-shoulder detection (HSD) step. Zhen Li proposed the fusion of Haar-like and HOG features to get better performance, and the HSD step utilizes edgelet features for classification and detection. This paper has limitations as low detection rate and less computation speed. In order to alleviate these limitations we propose here a new methodology for improving the detection rate and speed. Hence, the performance and accuracy of the detection can be improved by the combination of Haar-like and Triangular features for FBD and Edgelet/Shapelet for HSD. We expect an average 95% detection rate and 60% faster speed for the proposed method.