computer vision

Зменшення кількості хибних викликів під час розв’язання задачі детектування полум'я у відеопотоці з використанням глибоких згорткових нейронних мереж

In this paper, we develop a new approach for detecting fire in images based on convolutional neural networks. Cascade structure, which provides improved efficiency of recognition in images with low resolution and objects that can visually resemble flames, was proposed. We have performed an experimental comparison with the modern method of objects detecting Faster R-CNN. As a result of the experiments, it was found that performance of fire recognition improved on average by 20%.

Time-frequency analysis of heart sounds in spline basis

Splines have been used for the solution of the considered problems. This allows getting a single model for the three tasks and combine flexibility of the model with ease of calculations. For filtering and segmentation of acoustic signals spline filters that are similar to Savitsky-Golay filters have been used. Various widths of spline fragments provides a possibility to have different smoothness and select fragments of varying detalization.

BOX method for identification straight line on the raster image

The method BOX transformation that reflects lines on raster images for the points of intersection circumscribing the square image. Displaying is carried out by lines with distinct pairs of crossing points and has N2 complexity. The number of pairs of points lying on a straight line accumulates at a point of reflection. This selection allows you to direct the number of dots on them and perform filtering of individual points. The algorithms of direct and inverse conversion and the examples of images conversion have been demonstrated.