This paper presents a novel algorithm for detection certain types of emergencies relating to fire, smoke and explosions by processing the data recorded from the camera monitoring, based on cascaded approach. First, the combination of Adaboost and Local binary pattern (LBP) are using for getting Region of Interest (ROI) and reducing time complexity. Next, to alleviate common problems of vulnerable such as false positive, we propose to use Convolutional Neural Network (CNN). The final experimental results showed that the accuracy rate of this method for emergencies detection could reach 95.2%.

In the paper the method of single-frame image super-resolution based on the singular decomposition of matrix operator of the convergence square matrix operator is proposed. The characteristic vectors-features are obtained by using the Moore-Penrose pseudoinverse of matrix operator, which are used for enlarged image synthesis. The series of computational experiments based on images with fluctuation of intensity function are performed. The comparison results with others methods have confirmed the effectiveness of developed approach. The main advantages of proposed method for different enlargement coefficients are considered.

The new learning-based image super-resolution method is described in this article. The process of increasing the resolution of video frames or images from a set according to the method is based on the weight coefficients of synaptic connections. These coefficients are obtained by the learning neural-like structure on a pair of images of low and high resolution. The dimension influence of the training set on the generalization properties of the neural-like structure is investigated. The comparison of work effectiveness of the proposed method to existing ones is analyzed.

This thesis is devoted to development of visitor's queue density analysis and registration method for a retail videosurveillance systems. Developed method of foreground segmentation is based on initial background modeling, selective temporal median filter and local binary patterns. Based on the literature review, a problem statement has been examined and investigated. Basic methods and their limitations were also analyzed. Software application, that was developed to implement segmentation method and test results shows that method increases an objects of interest segmentation quality and also has an improved segmentation efficiency by minimizing an impacts of illumination and micromovements in an area of interest.

This article considers two algorithms which are designed for taking into consideration such events as the light changing in the video frame, the background micro-movements and occlusion of an object. These events determine the negative influences on the segmentation process, because they often arise in the eyeshot of surveillance cameras that are operated outdoors.