Search academic texts

Information × Registration Number 0209U010523, 0109U004834 , R & D reports Title Biometric identification of the person in video observation systems popup.stage_title Head Sachenko Anatoliy, Registration Date 14-12-2009 Organization Ternopol National Economic University popup.description2 The research objects are video surveillance systems. The goal of the project is to improve the motion detection methods and algorithms in illumination conditions with the usage of background subtraction method as well as image splitting to the quadtree for reducing of the processing time of each frame. Also to develop face detection methods for video surveillance systems by using the combined cascade of neural network classifiers for the validity and speed increasing in the conditions of the limited computational resources. The main results of this project are the following: analyzed existent methods and algorithms for motion detection in video stream in different illumination conditions and selected a directions of their improvement; improved method and algorithms of motion detection from video stream by using hierarchical data structure and integral image. This allowed to significantly improve the performance of detection process. Also it was proposed an adaptive background updating model depending if the current pixel belongs to foreground or background. This allowed to detect motion in the frame during a long time and with scene illumination changes; it is developed the generalized informational face detection model which uses the multilevel combined cascade of classifiers. This model allowed improving face detection methods for grayscale and color images processing; it is developed face detection method using the combined cascade of neural network classifiers which includes the additional level for face-candidates verification represented by the convolutional neural network. This allowed increasing the detection validity; face-candidates verification method is improved using the processing of candidate's image by convolutional neural network at once. This improvement allowed increasing the speed of the combined cascade of neural network classifiers; face detection information technology is developed on the base of abovementioned methods. It is implemented in Microsoft Visual Studio Team System 2008 using С++ and also utilizes Intel Open Computer Vision Library and Intel Integrated Performance Primitives. Product Description popup.authors Куриляк Юрій Орестович Лешко Тарас Миколайович Палій Ігор Орестович popup.nrat_date 2020-04-02 Close