Lightness Pretreatment with Pedestrian Detection Studies Using CH&K Algorithm for a Night Environment

Abstract

This paper presents a real-time pedestrian detection algorithm for a night environment. By first converting the nighttime image data to the L * a * b * color space, it can be extracted in the area L * with robust noise reduction and contrast adjustment. This data is used to generate pre-data through image subtraction. A background image is generated using the data, and the Cascade Histogram of Oriented Gradient &Kalman filter (CH & K) algorithm is proposed to track the movement of pedestrians. In addition, pre-processing algorithms and the proposed algorithm can replace Histograms of Oriented Gradients (HOGs) with rather heavy computations, and sensitive Haar-like features in the night can be used for real-time pedestrian detection and brightness.