DETECTION AND CLASSIFICATION OF MOVING OBJECTS IN A VIDEO SEQUENCE 

Shepelev Kirill Valer'evich, Master’s degree student, Penza State University (40 Krasnaya street, Penza, Russia), shepelevkir@gmail.com

004.932.2

10.21685/2072-3059-2017-4-5

Background. The current accuracy of moving objects detection via open source software does not exceed 60 %. This problem has several aspects. The first aspect is a lack of algorithms and software for import substitution, the second is a high cost of creating a separate video surveillance system with moving object detecting and classifying functions.
Materials and methods. For the purpose of detecting and classifying moving objects in a video sequence the author uses a motion detection algorithm by calculating static pixel values over a mixture of normal distributions and an algorithm for classifying images. The software implementation of the system is performed by means of OPENCV computer vision library.
Results and conclusions. The work suggests a model of video sequence representation in motion detection problems based on a three-dimensional dynamic information array of pixel brightness levels. It is distinguished by an updated zero motion mask frame and allows increasing the accuracy of object detection by dynamically refining the static part of a video sequence using a low-pass filter. The author has developed and implemented a motion detection algorithm based on estimating the probability density of static pixels falling out in a video sequence. The algorithm is distinguished by the use of a mix of normal distributions of brightness levels of
information field elements. A combined method for detecting and classifying moving objects in a video sequence is offered. The accuracy of detection of objects based on the combined method increases by 15 % compared to the accuracy of a simple motion detector and by 5 % relative to the accuracy of the Haar cascade classification

detection and classification of moving objects in a video sequences, video analysis, video surveillance, computer vision

1. Shepelev K. V. SCIENCE TIME. Mezhdunarodnyy elektronnyy nauchnyy zhurnal [International scientific e-journal]. 2016, no. 4, p. 241.
2. Bradsky G., Kaehler A. LA, O’Reilly. Cambridge, 2008, 140 p.
3. De Maesschalck R., Jouan-Rimbaud D., Massart D. L. Chemometrics and Intelligent Laboratory Systems. 2000, vol. 50 (1), pp. 1–18.
4. Nalimov V. V., Chernova N. A. Statisticheskie metody planirovaniya ekstremal'nykh eksperimentov [Statistical methods of extreme experiments planning]. Moscow: Nauka, 1965, p. 64.
5. Shepelev K. V., Shashkov B. D. Estestvennye i tekhnicheskie nauki [Natural and engineering sciences]. 2017, no. 4, p. 243.
6. Shepelev K. V., Makarychev P. P., Shepeleva D. S. OPEN INNOVATION: sb. st. Mezhdunar. nauch.-prakt. konf. [proceedings of an International scientific and practical conference]. Penza: Nauka i Prosveshchenie, 2017, p. 298.
7. Makarychev P. P. Novye informatsionnye tekhnologii i sistemy: sb. nauch. st. XII Mezhdunar. nauch.-tekhn. konf. [New information technologies and systems: proceedings of XII International scientific and technical conference]. Penza, 2015, pp. 24–26.
8. Popov D. I. Izvestiya vysshikh uchebnykh zavedeniy. Povolzhskiy region. Tekhnicheskie nauki [University proceedings. Volga region. Engineering sciences]. 2016, no. 2 (38), pp. 83–92.