48-ші ғылыми-әдiстемелiк конференциясының материалдары
жүктеу/скачать 5,84 Mb. Pdf көрінісі
|
3-книга
| Keywords: integrated circuits, neural network, commercialization of studies, scientific developments.
Over the past 3 years, application of digital informational technologies in various spheres of science, techniques and economy of Kazakhstan have become topical importance. According it, in 2016th the «Digital Kazakhstan» State Program for 2017-2020 was adopted. This program is aimed to improving quality of life of people and to competitiveness of economy of the country by development of digital technologies. According to international analytical agencies, it is shown that there is a primary link between education and commercialization of studies are the modern way of economic growth of the country. Therefore, in order to achieve goals in accordance with the «Informational Kazakhstan 2020» state program which adopted on October 13, 2006, our state carried out a big goal on creating a digital society and training specialists with business oriented skills. The revolution in computer vision has been brewing for many years. Machine vision is a scientific direction in the field of artificial intelligence, in particular robotics. Neural networks are the complement and development of classical expert and indistinct systems - the main directions of artificial intelligence theory applications. The main modern problem of creating effective neural networks is the improvement of biological neuron models, the approximation of neural networks work to the work of human brain. At the present time, scientific research on various branches of this leading problem is being intensively pursued: selection of optimal functional models of neurons, perfection of the multilayer neural networks topology, computational algorithms of training, circuitry and electronic implementation of neural networks on integrated circuits. However, in our opinion, the most important solution is the correct choice of the biological neuron's dynamics model. This model should describe the basic physical characteristics of the self-organization process of the action potential of the neuron itself. Then, due to the self-similarity of self-organization processes, it is possible to create an effective artificial neural network from such neurons. We proposed a general scale-invariant model of action potential of neurons. Our nonlinear model correctly describes the main patterns of neuronal dynamics, this is asymmetric intermittency, chaos, hierarchy and self- similarity, depolarization, phase reinstallation. On the basis of what a three-layer neural network was developed and created with analog and digital electronic elements generating signals with scale-invariant properties. To date, our neural networks with non-linear electronic elements are used to classify chaotic signals, in order to recognize the object (speakers by voice) and remove noise (in a signal with an unknown noise level). The algorithm that we will use when detecting a face is highly effective for searching objects on images and video streams in real time. According to preliminary estimates, our detector will have an extremely low probability of false face detection. The method works well and detects facial features even when observing an object at a slight angle. The accuracy of recognition using this method can reach values above 90%, which is a very good result. The product offered by us is a promising area of artificial intelligence application. Potential users of the product and scope: 1. Large and medium-sized educational centers are used to count applicants in the 217 admissions office in order to quickly respond to the shortage of consultants. Conducting the rating of teachers by the number of students in relation to the total number of participants. Monitoring and control centers for the area with the help of machine vision. 2. for monitoring personnel and statistics of the time and number of customers visited to improve service and meet market demand. And also the detection of suspicious situations in real time without the participation of the human factor 3. for monitoring of expensive equipment, registration of staff time. 4. Strategically important objects of the state are for monitoring of the underground, stations and the airport. Individual approach to each client will be carried out, taking into account each client's request up to the design of the user interface of the software. It is also planned to provide further product maintenance. Students and university graduates will be involved with basic knowledge of the principles of artificial intelligence. For this, the relevant disciplines have already been introduced into the educational process of the KazNU named after al-Farabi. Our state, like many others, are sending huge amounts of money in the sphere of application of artificial intelligence for security and military tasks. In this direction, the dominant factor of our product is domestic development. It is possible to execute additional orders based on the main product core. The machine vision with the latest generation neural network we have proposed is affordable from an economic point of view, an ecological and domestic product. жүктеу/скачать 5,84 Mb. Достарыңызбен бөлісу:
|