In this work, training and recognition of the types of aberrations corresponding to individual Zernike functions have been carried out by the pattern of the intensity of the point scattering function (PSF) using convolutional neural networks. The PSF intensity patterns in the focal plane were modeled using the Fast Fourier Transform algorithm. When training a neural network, the learning coefficient and the number of epochs for a dataset of a given size were selected empirically. The average prediction errors of the neural network for each type of aberration were obtained for a set of 15 Zernike functions from a dataset of 15 thousand PSF pictures. As a result of training, for most types of aberrations, averaged absolute errors were obtained in the range 0.012–0.015, however, the determination of the aberration coefficient (magnitude) requires additional research and data, for example, calculating the PSF in the extrafocal plane.
The paper proposes a video surveillance scheme for compact placement of a system for railway rolling stock accounting. This design is based on the use of a tilted diffractive optical element and a tilted lens. Such an optical design makes it possible to significantly increase the depth of focus of the imaging system. This work considers the influence of the tilt of a diffractive lens on the shape and size of the focused area. Analytical relations describing the geometry of the focused region for various spectral channels are given. The possibility of increasing by several times the size of the zone of accurate image classification using a neural network has been demonstrated. The proposed approach has been tested on real-world dataset of images of house number plates.
The paper proposes using a two-zone different level Fresnel lens to increase the depth of field. On the one hand, such a diffractive optical element can reduce the weight of the device compared to, for example, cubic phase and binary axicon apodization. On the other hand, such an element has a simpler structure compared to a harmonic lens or free-form DOE. A neural network is used to restore the image. Optimization of the surface relief of the proposed two-zone lens is performed.
KEYWORDS: Video, Video surveillance, Video processing, Data processing, Signal processing, Distributed computing, Computing systems, Systems modeling, Multidimensional signal processing
Current trends in increasing the intelligence level of software systems for analyzing data flows under the constraints of processing time require the study of new technologies that take advantage of a distributed computing environment. In this paper, we implement a technology for distributed streaming processing of multidimensional optical signals based on the Apache Flink framework. The technology is studied on the task of transport video surveillance tracking unique objects captured by a system of geographically dispersed video cameras. We study the ability of proposed solution scale the processing depending on the amount of resources in the cloud, the quantity and quality of optical signals. The characteristics of processing processes of a set of frames of varying complexity on a computing cluster are investigated. NVIDIA AI City Challenge is used as test data sets.
KEYWORDS: Light sources and illumination, Cameras, Video, Control systems, Sensors, Imaging systems, Reliability, Video surveillance, Image processing, Video processing
In optical information-measuring systems structured lighting is an important part that largely influences and even determines both used algorithms and hardware. In many cases it is the addition of a structured lighting turns a vision system in measurement system and improves intellectual features of information system. I suggest using a structured lighting for detecting the presence of a train on a controlled track and fixing car-to-car gaps in vision systems on the railway. This approach eliminates the need of special sensors or additional equipment for the specified control and, in conjunction with other decisions of complex technology articulating design of all components of vision systems, allows to increase the maximum speed of trains passing through the checkpoints, to improve significantly the accuracy of the process of recognition of identification numbers of cars and containers, makes it possible to bring the camera closer to the track, to control all roads of the railway station simultaneously under the minimal distance between them.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.