Pedestrian Detection using Principal Components Analysis of Gradient Distribution

In this paper we proposed a new method forpedestrian detection in images and videos. Our method usesa sliding window to search through images. In order toextract the features, each window is divided intooverlapping cells and features are extracted from them. Thefeature that we extracted to describe each window is basedon analysis of gradient distribution of each cell. Aftergradient distribution of a cell computed, the PCA is appliedon it and using a mathematical expression that gauges theattitude of edges we got the feature of the cell. Putting thefeatures of the cells next to each other forms the featurevector of the window. Then, the extracted features areclassified using Support Vector Machine (SVM). Finally, thelearned SVM model tested on the INRIA pedestrian dataset.The proposed method was compared with Histograms ofOriented Gradient (HOG) approach and the results showthat our method has comparable detection accuracy as wellas having more robustness when facing with noise.

کلیدواژه ها:

Pedestrian Detection ، Gradient Distribution Histogram of Oriented Gradients ، Principal Component Analysis