Litcius/Paper detail

Vehicle Detection in the Autonomous Vehicle Environment for Potential Collision Warning

Mario Gluhakovic, Marijan Herceg, Miroslav Popović, Jelena Kovačević

202027 citationsDOI

Abstract

In this paper, a method for the vehicles detection in the surroundings of an autonomous vehicle and warnings of potential collision with them is presented. The method, which consists of two parts, is implemented in robot operating system (ROS). The first part is used to detect vehicles in an autonomous vehicle environment, in which, YOLO v2 algorithm, trained on a newly created set of images, is used. The YOLO v2 algorithm is configured to detect four classes of objects: a car, a van, a truck, and a bus. The second part of the proposed method is the ROS node for distance assessment. In particular, two ROS nodes for distance assessment are created; one ROS node used for distance assessment in the Carla simulator, while the other ROS node is used for real-world distance assessment. The testing results of the proposed method show promising results.

Topics & Concepts

Node (physics)Computer scienceCollision avoidanceTruckCollisionSet (abstract data type)Real-time computingRobotCollision detectionArtificial intelligenceSimulationComputer visionAutomotive engineeringEngineeringComputer securityProgramming languageStructural engineeringAdvanced Neural Network ApplicationsAutonomous Vehicle Technology and SafetyVideo Surveillance and Tracking Methods
Vehicle Detection in the Autonomous Vehicle Environment for Potential Collision Warning | Litcius