Zpracování a analýza dat z lidaru pro detekci a identifikaci objektů v reálném čase

Abstract

Goal of this master’s thesis is analysis of lidar sensors, design of the lidar data processing, object detection and clasificaion algorithm. Frame structure for attaching lidar to the vehicle was designed and constructed. For the data processing purpose was used Python library Ouster-SDK. Object detection is ensured by the yolo11 neural network. Application was optimized for the Follow the Vehicle use case. Yolo network was retrained on custom lidar dataset in demand of better detection stability.