An Integrated Real-time Road Surveillance System Using Yolov8 For Multi-class Object Detection

Abstract

The Growing Number Of Road Accidents Worldwide Necessitates The Deployment Of Intelligent Systems Capable Of Continuously Monitoring Roadway Conditions. This Paper Introduces A Comprehensive Computer Vision Framework Designed To Detect Various Roadway Features, Including Surface Irregularities, Traffic Control Devices, Elevation Changes, And Biological Entities. The System Utilizes The YOLOv8 Deep Learning Architecture, Which Is Recognized For Its Superior Performance In Balancing Detection Accuracy With Processing Speed. A Specialized Image Dataset Was Curated And Annotated To Train The Model For Recognizing Four Distinct Categories. The Trained System Processes Both Still Images And Continuous Video Feeds, Marking Identified Objects With Bounding Boxes And Confidence Values. Experimental Evaluation Demonstrates That The Proposed Framework Achieves A Mean Average Precision Of 0.93 While Maintaining Real-time Processing Capabilities. This Work Presents A Significant Advancement Over Conventional Approaches That Typically Address Only Single Object Categories, Offering A More Complete Solution For Automated Road Monitoring Applications.

Keywords