Boosting the Accuracy of AdaBoost for Object Detection and Recognition
Zahid Mahmood, Nazeer Muhammad, Muhammad Arif, Imran Shahzad, Fahad Khan, Mazhar Ali, Uzair Ijaz Khan, Samee U. Khan · 2016
Recently, large number of object detection and recognition algorithms are playing a key role in several applications, such as security and surveillance. Although, these algorithms perform exceptionally well under normal lighting conditions, however their detection and recognition accuracy abruptly degrades under non-uniform illuminations, such as strong sunlight and bad lighting conditions. In this paper, we apply our own developed Multi-Scale Retinex (MSR) algorithm as a pre-processing module to boost the accuracy of the AdaBoost algorithm, which is considered to be state-of-the-art algorithm for robust object detection and recognition. Simulation results show that the MSR can be reliably and effectively used under non-uniform illuminations to boost the accuracy of the AdaBoost.