The monograph presents modern approaches to the analysis of the road environment images based on deep learning. The neural network architectures for solving problems of classification, segmentation of images, detection of objects on them are considered. Author analyzes in detail the deep neural network architectures for the detection of the road scene elements (vehicles, traffic lights) on the images from the on-board video camera. A solution is proposed for calculation of founded vehicle position. The approach to the use of convolutional neural networks of different architectures has been analyzed to detect the visibility loss of a video camera on the basis of recognition of images obtained from it. Author describes tools for creating datasets in the traffic scene recognition tasks. Author also presents examples of software and hardware implementation for these architectures using a graphics processor and NVidia CUDA technology. The publication is intended for scientists and engineers engaged in the development of machine vision systems using deep convolutional neural networks and may be useful for lecturers, students and postgraduates of relevant university specialties.
Bitte wählen Sie Ihr Anliegen aus.
Rechnungen
Retourenschein anfordern
Bestellstatus
Storno