STUDY OF THE EFFICIENCY OF PROCESSING BIG DATA IN VIDEO SURVEILLANCE SYSTEMS

Анотація

Abstract. The paper presents an overview of methods of intelligent data processing in security video surveillance systems. For efficient operation of such systems should use modern data processing technologies and algorithms for information mining. An important task in video surveillance systems is the processing of moving images. It is also

increasingly necessary to perform statistical analysis of data obtained from video cameras, to determine the exact location of objects of observation, their number, to determine the characteristics of the obtained data for security, marketing, analytics, еtс. A comparison of methods for traffic control in modern video surveillance systems is given. It is proposed to use distributed systems for big data processing where all the data that need to be processed are distributed between different computing devices in the system, processed in parallel, and sent to the main device, where all parts of the tasks are collected and the final result is obtained. The software model for distributed data processing in video surveillance systems is developed in the work. An analysis of the main data processing technologies: Hadoop and Spark. In Hadoop, data processed by devices in a distributed system is stored in non-volatile memory. After processing, the data is deleted from the non- volatile memory of the computing device. The process of recording and deleting data takes some time and significantly slows down data processing systems. The paper experimentally proves the efficiency of using Spark distributed data processing technology. This technology uses RAM to store data on computing devices. This allows faster processing of information in video surveillance systems and analysis of data from video cameras. This approach opens opportunities for further research of video surveillance systems, improving their characteristics for better and more accurate data processing.