Pedestrian detection using dense LDB descriptor combined with HOG

Pedestrian detection plays a vital role in numerous vision-based safety and security applications in recent days. Given an image, a pedestrian detector computes features from it and works on the features to classify if there is pedestrian. This paper presents a new feature set for pedestrian detection where a modified version of the local difference binary features are combined with the histogram of oriented gradients features. The linear support vector machine is used as the classifier. The performance of the proposed detector is presented in terms of miss-rate versus FPPI and miss-rate versus FPPW, and is compared with available pedestrian detectors of similar type. The computational efficiency of the detector is also studied.