Zbigniew Mikrut

Seamless Tracing of Human Behavior Using Complementary Wearable and House-Embedded Sensors

This paper presents a multimodal system for seamless surveillance of elderly people in their living environment. The system uses simultaneously a wearable sensor network for each individual and premise-embedded sensors specific for each environment. The paper demonstrates the benefits of using complementary information from two types of mobility sensors: visual flow-based image analysis and an accelerometer-based wearable network. The paper provides results for indoor recognition of several elementary poses and outdoor recognition of complex movements. Instead of complete system description, particular attention was drawn to a polar histogram-based method of visual pose recognition, complementary use and synchronization of the data from wearable and premise-embedded networks and an automatic danger detection algorithm driven by two premise- and subject-related databases. The novelty of our approach also consists in feeding the databases with real-life recordings from the subject, and in using the dynamic time-warping algorithm for measurements of distance between actions represented as elementary poses in behavioral records. The main results of testing our method include: 95.5% accuracy of elementary pose recognition by the video system, 96.7% accuracy of elementary pose recognition by the accelerometer-based system, 98.9% accuracy of elementary pose recognition by the combined accelerometer and video-based system, and 80% accuracy of complex outdoor activity recognition by the accelerometer-based wearable system.

Video Detection Algorithm Using an Optical Flow Calculation Method

The article presents the concept and implementation of an algorithm for detecting and counting vehicles based on optical flow analysis. The effectiveness and calculation time of three optical flow algorithms (Lucas-Kanade, Horn-Schunck and Brox) were compared. Taking into account the effectiveness and calculation time the Horn-Schunck algorithm was selected and applied to separating moving objects. The authors found that the algorithm is effective at detecting objects when they are subject to binarisation using a fixed threshold. Thanks to the specialized software the results obtained by the algorithm were compared with the manual ones: the total vehicle detection and counting rate achieved by the algorithm was 95,4%. The algorithm is capable to analyse about 8 frames per second (Intel Core i7 920, 2.66 GHz processor, Win7x64).

A method of linear "star-sections" applied for object separation in ERCP images

The aim of the present study is to analyze ERCP images (endoscopic retrograde cholangiopancreatography). This means images of a contrasted pancreatic duct obtained through an X-ray method. More precisely, this involves creating an algorithm performing the first stage of the analysis/separation of the object of interest.

Combining pattern matching and optical flow methods in home care vision system

The article presents the structure and working of the system supervising the convalescent or elder person at home. Images acquired from a suitably mounted camera are analyzed to determine the pose and activity of the observed person. Extensive configuration module allows to define zones of rest and obstructing objects. Situations of long immobility are detected in places where it should not happen. The activity of observed person is computed using two independent methods: by counting the number of frames in which the active poses are detected and by counting the number of frames, in which the dominant component of the optical flow histogram exceeded the threshold value. By keeping methods of image analysis as simple as possible the processing time was achieved close to the real-time.

Complete Scanpaths Analysis Toolbox

This paper presents a complete open software environment for control, data processing and assessment of visual experiments. Visual experiments are widely used in research on human perception physiology and the results are applicable to various visual information-based man-machine interfacing, human-emulated automatic visual systems or scanpath-based learning of perceptual habits. The toolbox is designed for Matlab platform and supports infra-red reflection-based eyetracker in calibration and scanpath analysis modes. Toolbox procedures are organized in three layers: the lower one, communicating with the eyetracker output file, the middle detecting scanpath events on a physiological background and the one upper consisting of experiment schedule scripts, statistics and summaries. Several examples of visual experiments carried out with use of the presented toolbox complete the paper

Recognition of objects normalized in log-polar space using Kohonen networks

In the paper, an algorithm is presented for the construction of representations of 18 object classes, which can be later recognized by a hybrid neural network. The preprocessing took place in log-polar space and it included: object centering, binarization, edge detection, normalization of angular position and scaling. After the normalization and log-Hough transformation, the maxima have been projected onto respective axes. In the paper, results have been discussed of neural network learning using such constructed representations, and the map of features for the Kohonen layer has been analyzed.

Analysis of Images of Bronchial Tree Ramifications for Summarization of Bronchoscopic Videos

Intensive endoscopic examinations result in an increasing number of records in video format. Storing whole videos is ineffective when considering mass storage devices usage. Additionally, reviewing these videos is a very time-consuming task.

INSIGMA: an intelligent transportation system for urban mobility enhancement

Intelligent Transportation Systems (ITS) aim to improve safety, mobility and environmental performance of road transport. The INSIGMA project provides a fresh look at the possible innovations in this field, by enhancing the functionality and accuracy of ITS in urban environments. This paper describes the architecture, sensors, processing algorithms, output modules and advantages of the developed system. A comparison of existing ITS systems has been provided as background. Special attention has been given to performance and privacy issues, as the system includes social aspects such as location monitoring.

Contribution of the INSIGMA Project to the Field of Intelligent Transportation Systems

With the growing number of vehicles traveling on public roads, traffic congestion has become a serious problem, resulting in more unpredictable travel times, increased fuel consumption and pollution. Intelligent Transportation Systems (ITS), already being developed by several countries, aim to improve safety, mobility and environmental performance. The goal of the INSIGMA project is to develop a system providing functionality of a typical ITS: real-time traffic monitoring, route planning and traffic control. In this paper we discuss the concepts and solutions developed within the project: dynamic map, sensors – videodetector and GPS tracker – as well as advanced route planning and traffic control algorithms.

Detection and Recognition of Selected Class Railway Signs

Abstract The paper aims at presentation of results of research on detection and recognition of selected class railway signs (W11p). When conducting the research, the authors have proposed their own algorithm, which achieved about 90% effectiveness at detecting W11p signs and 98% effectiveness at classifying them. The processes of localisation, segmentation and recognition of W11p signs were considerably simplified thanks to the application of backpropagation neural network. The authors believe that two non-standard methods related to the use of the network deserve attention: the application of an interactive method of generating the training set, owing to which also pixels highly diversified in terms of their colours could be included, and the use of a full spectrum of neural network responses, which made it possible to accomplish a feedback. It consisted in an automatic adjusting of the network responses’ threshold to the results of segmentation and recognition.