A Smart Traffic Control System Based on Pixel-Labeling and SORT Tracker

Autonomous vehicle detection and tracking are crucial for intelligent transportation management control systems. Although many techniques used to develop smart traffic systems, this article discusses using pixel-labeling real-time tracking. We propose a novel system that segments the image an Extreme Gradient Boost (XGBoost) classifier extract foreground objects. The proposed model is divided into following steps: 1) at first, all images preprocessed remove noise; 2) performed by XGBoost separate background from foreground; 3) pixels classified as was extracted converted binary image, then blob extraction technique localize each vehicle; 4) verify detected vehicles Intersection over Union (IoU) score ground truth calculated; 5) verified were subjected Visual Geometry Group (VGG) feature based on which unique identifier assigned of them enable multi-object across frames; 6) counted categorized stationary moving cars detecting motion in Farneback optical flow algorithm finally, Simple Online Real-time Tracker (SORT) outperforms existing state-of-the-art monitoring terms precision, achieving 0.86 0.92 with Karlsruher Institut Technology Aerial Image Sequences (KIT-AIS) dataset, 0.83 detection, 0.87 Vision Meets Drone Single Object-Tracking (VisDrone) dataset. can be several purposes, such identification traffic, density intersections, conditions road, providing pedestrian way.