Browsing by Author "Mukhina, M. P."

Now showing 1 - 7 of 7

ACCURACY RESEARCH METHOD OF THE MODIFIED ALGORITHM FOR DETECTING LINEAR LANDMARKS
(Київ «Освіта України», 2018-06) Mukhina, M. P.; Tkachenko, O. Yu.; Barkulova, I. V.
A search algorithm for the most extended landmark by which unmanned aerial vehicle can be followed by and implemented flight correction was proposed. The software was developed based on the Python language. The functionality of this software is to detect the linear landmarks from images of geophysical field, received from unmanned aerial vehicle in real time. Images were processed by Hough Line Transform method. As a result, obtained visualization of the object detection with the greatest length, as linear landmark, which allows to estimate unmanned aerial vehicle location. The visual analysis of the effectiveness of this algorithm for inertial navigation system correction shown that the algorithmic software is appropriate for use on unmanned aerial vehicle board and due to applying computer vision systems, gives as correct results of location determining as possible.
ALGORITHM OF VARIATIVE FEATURE DETECTION AND PREDICTION IN CONTEXT-DEPENDENT RECOGNITION
(Київ «Освіта України», 2018-03) Mukhina, M. P.; Barkulova, I. V.
Application of context-dependent classification for recognition tasks is proposed. In the context-free classification, the starting point was the Bayesian classifier. Morphological features such as object form, area, and eccentricity were considered through context-dependent classification. As result, dependences which can be used for object recognition have been obtained, and further they can be used together with interesting point detectors. The procedure of prediction of object variative features was developed.
ANALYSES OF MARGINALIZED PARTICLE FILTERING BLOCK OF NAVIGATION DATA
(«Освіта України», 2017-06) Mukhina, M. P.; Prymak, A. P.
The particle filter offers a general numerical tool to approximate the posterior density function for the state in nonlinear and non-Gaussian filtering problems. While the particle filter is fairly easy to implement and tune, its main drawback is that it is quite computer intensive, with the computational complexity increasing quickly with the state dimension. One remedy to this problem is to marginalize out the states appearing linearly in the dynamics. The result is that one Kalman filter is associated with each particle. Filtering block has been developed with the help of which navigation data received from UAV is filtered. UAV motion with camera on board has been conducted and photos have been captured from it. Photos have been processed by OpenSurf method, with the help of which feature points has been detected, filtered and compared with previous image. Result of research shows that with help of comparing of two neighboring images we can reconstruct relief above which UAV flew.
COMBINATION OF HOUGH TRANSFORM AND CANNY EDGE DETECOR FOR IMPROVEMENT OF LINEAR OBECT DETECTION RESULTS
(Київ «Освіта України», 2018-12) Mukhina, M. P.; Derkach, V. Yu.; Prymak, A. P.
Hough transform together with Canny edge detector are proposed to use for road lines detection at different weather conditions at various places. Canny operator is used for edge detection on image, then Hough transformation is applied for the detection of extended objects. Proposed testing of such combination of two methods is done on datasets of frames from the unmanned aerial vehicle benchmark. The goal of this research is to obtain comparative results of Hough transform use effectiveness at different conditions and to develop set of recommendations for the implementation of the proposed software for automatic detection of lines in the image in order to identify the roadway and improve the effectiveness of further recognition of ground moving objects. The research was implemented with the help of means of Python 3.6 language in the environment of Anaconda, Skicit image library.
COMPARISON OF BRIEF AND ORB BINARY DESCRIPTORS
(Київ «Освіта України», 2018-12) Mukhina, M. P.; Trach, Yu. V.; Prymak, A. P.
A great deal of features detectors and descriptors are proposed nowadays for various computer vision applications. The task of image processing which is invariant to all weather conditions (such as rain, fog, smoke, unfavorable lights, camera rotation) is presented. The influence of weather conditions on the number of determined key points in the image is analyzed, and how certain unfavorable conditions influence the tracking of these points from frame to frame. Two binary descriptors are considered for finding special points of the image; BRIEF (Binary robust independent elementary features) and ORB (Oriented FAST and rotated BRIEF) descriptors, as well as their modifications, to explore the most efficient descriptor which can be used in applications that run in real time. The result of the study shows that the descriptors have high stability characteristics, working with different types of images and rotation angles, using the recommendations for the use of their modifications.
PERFORMANCE AND SPEED COMPARISON OF SURF AND ORB DESCRIPTORS
(Київ «Освіта України», 2018-03) Mukhina, M. P.; Yeremeieva, T. A.
Fast and robust image processing and matching is a very important task with various applications in computer vision and robotics. In this paper, we compare the performance of two different image matching techniques, i.e., by speed up robust features and by rotated robust independent elementary features, against different kinds of transformations and deformations such as scaling, rotation, noise, fisheye distortion, and cropping. For this purpose, we manually apply different types of transformations on original images and compute the matching evaluation parameters such as the number of key points in images, the matching rate, and the execution time required for each algorithm and we will show that which algorithm is the best more robust against each kind of distortion.
STRUCTURE OF AIDED CLASSIFICATION OF GROUND OBJECTS BY VIDEO OBSERVATION
(«Освіта України», 2017-12) Mukhina, M. P.; Barkulova, I. V.
Analysis of classification structure by video observation has been done. It was formulated, that for feature extraction and their classification, normalized hypothesis for object feature detection, taking into account camera orientation and flight height, have being obtained. The system with aided classification based on probabilistic models, such as Bayesian classifier and Markov chain model, is proposed. The applied algorithm was used for detection by only two features related to Binary Large Objects (BLOB) analyses. Classification was done by two main feature parameters: area and center of mass. Feature vector contains the most informative components and allows the minimization of decision risks. Results have proven the reliability of classification during a number of video frames in the condition of non-full data descriptive space.