Pedestrian Detection using KPCA and FLD Algorithms

A pedestrian detection method by using kernel principle component analysis (KPCA) and Fisher linear discriminant (FLD) is presented in this paper. The basic idea of this method is to first utilize the KPCA algorithm to perform feature extraction, which obtains the nonlinear principle components in the high dimension feature space composed of haar wavelet coefficients, and then implement classification via the FLD algorithm in the KPCA-transformed space. The two-phase classification approach is also regarded as the essence of kernel Fisher discriminant (KFD) in other works. The effectiveness of the proposed method for detecting people is verified using the DaimlerChrysler pedestrian classification benchmark dataset.