Raw Bayer Pattern Image Synthesis on Histogram of Gradients for Pedestrian Detection

In recent days, there has been a vast implementation and emergency for many pedestrian detection methods. In the current scenario, the complexity of pedestrian detection has increased, and its identification is a challenging task due to the variety of scenes and the complexity of most pedestrian identification systems. The availability of challenging public datasets has fueled most of the progress made in recent years. To keep up with the rapid pace of innovation, the proposed work is a pedestrian detection approach based on the Bayer pattern and gamma correction that efficiently classifies pedestrians and non-pedestrians. The proposed method uses a Support Vector Machine (SVM) to ensure that the gradients recovered from Bayer pattern pictures are robust enough to be employed in pedestrian detection algorithms based on Histograms of Oriented Gradients (HOG). Benchmark datasets from MIT, Caltech, and INRIA were used to test the proposed technique. In comparison to other previous techniques, it is inferred that our proposed technique outperforms them by maintaining good accuracy.