Real-Time Pedestrian Detection Based on Improved YOLO Model

To investigate accurate and real-time pedestrian detection results is a mainstream trend in the field of intelligent security. Avoiding too many external disturbances causing the errors and omissions, in order to obtain a reliable and discriminative detection. This paper proposes a deep learning method based on improved YOLO model to efficiently detect pedestrians. It addresses two necessary above issues: (1) leverage real-time saliency region detection through surveillance camera; and (2) extract more detail discriminative feature with human parsing. The results show that our deep real-time saliency and detail discriminative feature with human parsing based on improved YOLO model, successfully learn both spatial and temporal cues, making pedestrian detection further ensures the accuracy and timeliness in practical application scenes.