Use of a tilted diffractive lens with an increased depth of field in decision-making systems based on computer vision
Alexey Dzyuba (ITMO University), Pavel Serafimovich (Image Processing Systems Institute), Svetlana Khonina (Image Processing Systems Institute, Samara National Research Univ.), Sergei Popov (Image Processing Systems Institute, Samara National Research Univ.)
Abstract
The paper proposes a computer vision system with an increased depth of field. The optical unit of this system contains a tilted diffractive lens, which provides an increase in the depth of field separately for each spectral channel. The decision-making block is built on the basis of convolutional neural networks, which are trained to process single-channel spectral images. The proposed approach has been tested on real-world data sets: images of house number plates and images of road signs. The results of the work can be used in computer vision systems for robots and unmanned aerial vehicles.
Speaker
Alexey Dzyuba
ITMO University
Russia
Discussion
Ask question
