Category: Research

The presented paper benchmarks the performance of state-of-the-art methods of objects detection in the particular case of airplanes on the ground identification detection in aerial images taken from unmanned aerial vehicles or satellites. There were tested two popular single-stage neural networks YOLO v.3 and Tiny YOLO v.3 based on the “You Only Look Once” approach. The considered artificial neural network architectures for objects detection has been trained and applied over the specifically created image database. Experimental verification proves their high detection ability, location precision and real-time processing speed using modern graphics processing unit. That approach can be easily applied for detection of many different classes of ground objects.

In the paper, two examples of embedded vision modules are described. Firstly, it is demonstrated how fiducial marker tracking algorithm can be adopted for operation on Raspberry Pi. Usage of proposed ideas allows to achieve around 60fps speed of binary marker tracking. Secondly, we describe the problem of text detection and recognition in outdoor environment. Experimental results indicate on acceptable results and good potential to provide low-cost and efficient embedded vision system for this purpose. Technical details of both embedded vision modules are comprehensively discussed.

The competition on the world market of smartphones and tablets between the acknowledged leaders on one side and the numerous newcomers on the other makes them all look for new solutions that open additional opportunities for the developers and customers without the growth of the price. The progress with new opportunities turns possible when it happens simultaneously in the software and the hardware. The brightest one example of the above statement can be observed for the sensors of mobile devices. It is totally impossible to imagine modern smart devices having no sensors, as the progress of last decade (SLAM, face ID, OCR, pattern recognition etc.) was achieved thanks to considerable improvements of sensors and the algorithms for their processing. The paper addresses the questions of
characteristics analysis of such mobile sensors as accelerometer, magnetometer and gyroscope from the point of view of their application in indoor navigation field. Signals of BLE beacons and their processing methods are investigated as well. The sensor fusion task is briefly discussed and several practical examples are given.

Augmented reality is popular and rapidly growing direction. It is successfully used in medicine, education, engineering and entertainment. In the paper, basic principles of typical augmented reality system are described. An efficient hybrid visual tracking algorithm is proposed. The approach is based on combining of the optical flow technique with direct tracking methods. It is demonstrated that developed technique allows to achieve stable and precise results. Comparative experimental results are included.

In the paper, a system for real-time positioning is proposed. Developed signal processing algorithms for precise user localization and navigation are described. It is demonstrated that proper calibration and received signal filtering leads to improvement of positioning accuracy. Peculiarities of Bluetooth Low Energy beacons as signal sources are considered. Key components of the created software development kit are described. Experimental results of testing on mobile platforms are given.

Nowadays automatic methods based on artificial intelligence are rapidly growing. In the paper, a problem of automatic target recognition in synthetic aperture radar images is described. It is demonstrated, that two different machine learning instruments can provide very high classification accuracy. In particular, support vector machines with proper optimization and developed local feature set gives competitive results. Secondly, a novel architecture of convolutional neural network is proposed. Important practical aspects of both methods are analyzed. Experimental results for MSTAR are given.

SAR is a very popular instrument for imaging of the ground surface. Possibility of high-resolution image formation makes it superior tool for various information extraction tasks. In the paper, a problem of automatic target recognition is comprehensively analyzed. An idea of azimuth and range target profiles fusion is proposed. It is demonstrated, that usage of a proper image preprocessing with appropriate feature extraction steps allow to achieve a competitive recognition accuracy while keeping a low-dimensionality of feature vectors. Experimental results are discussed for a publicly available MSTAR dataset.

Real-time user positioning and navigation services are widely used daily by millions of people. The challenge is that common global positioning systems fail in indoor environment or in scenes with a limited sky view. In the paper, the indoor navigation framework based on the Bluetooth beacons is proposed. Such system allows user to obtain the position in GPS denied environment on a real-time basis. Key steps for precise localizing and main components of the developed framework are considered. Initial experimental results are provided