Vehicle Detection in ITS Based on Improved Background Modeling Method As An Advanced Pattern Recognition Strategy for Surveillance Purposes

• Main Authors: Alavianmehr, Mohammad Ali, Helfroush, Mohammad Sadegh, Danyali, Habibollah, Tashk, Ashkan
• Format: Conference Proceeding
• Language: English
• Subjects: Adaptation models; Cameras; Feature extraction; Gaussian Mixture Model (GMM); Image color analysis; Intelligent Transport System (ITS); Object detection; Statistical processing; Surveillance; Vehicle detection
• ISSN: 2049-3630
• DOI: 10.1109/PRIA.2019.8786019

Moving object detection from municipal surveillance cameras is an important issue for Intelligent Transportation Systems (ITS) purposes. This paper presents a moving object detection algorithm that is more robust than adaptive Gaussian mixture (GMM) model, and provides a novel and practical choice for intelligent video surveillance systems using static cameras. The proposed method comprises of three steps. In the first step, statistical bit maps (BMs) are constructed randomly in a block-wise manner. The second step is assigned to background image construction based on two types of comparisons named as intra and inter block comparative approaches. In the third step, based on the smooth or unsmooth states detection for each block of the original video frames, an update or unmatched process will be applied. The results of proposed method's implementation show its better and more efficient performance than other competitive methods like GMM for live and real time moving object detection from video images acquired by the means of municipal ITS surveillance cameras.