Árpád Barsi

Neural self-organization using graphs

The self-organizing feature map (SOFM) algorithm can be generalized, if the regular neuron grid is replaced by an undirected graph. The training rule is furthermore very simple: after a competition step, the weights of the winner neuron and its neighborhood must be updated. The update is based on the generalized adjacency of the initial graph. This feature is invariant during the training; therefore its derivation can be achieved in the preprocessing. The newly developed self-organizing neuron graph (SONG) algorithm is applied in function approximation, character fitting and satellite image analysis. The results have proven the efficiency of the algorithm.

The Effect of Tear Supplementation on Ocular Surface Sensations during the Interblink Interval in Patients with Dry Eye.

Purpose To investigate the characteristics of ocular surface sensations and corneal sensitivity during the interblink interval before and after tear supplementation in dry eye patients. Methods Twenty subjects (41.88±14.37 years) with dry eye symptoms were included in the dry eye group. Fourteen subjects (39.13±11.27 years) without any clinical signs and/or symptoms of dry eye were included in the control group. Tear film dynamics was assessed by non-invasive tear film breakup time (NI-BUT) in parallel with continuous recordings of ocular sensations during forced blinking. Corneal sensitivity to selective stimulation of corneal mechano-, cold and chemical receptors was assessed using a gas esthesiometer. All the measurements were made before and 5 min after saline and hydroxypropyl-guar (HP-guar) drops. Results In dry eye patients the intensity of irritation increased rapidly after the last blink during forced blinking, while in controls there was no alteration in the intensity during the first 10 sec followed by an exponential increase. Irritation scores were significantly higher in dry eye patients throughout the entire interblink interval compared to controls (p<0.004). NI-BUT significantly increased after HP-guar (p = 0.003) but not after saline drops (p = 0.14). In both groups, either after saline or HP-guar the shape of symptom intensity curves remained the same with significantly lower irritation scores (p<0.004), however after HP-guar the decrease was significantly more pronounced (p<0.004). Corneal sensitivity to selective mechanical, cold and chemical stimulation decreased significantly in both groups after HP-guar (p<0.05), but not after saline drops (p>0.05). Conclusion Ocular surface irritation responses due to tear film drying are considerably increased in dry eye patients compared to normal subjects. Although tear supplementation improves the protective tear film layer, and thus reduce unpleasant sensory responses, the rapid rise in discomfort is still maintained and might be responsible for the remaining complaints of dry eye patients despite the treatment.

Detecting road junctions by artificial neural networks

Road junctions are important objects for all traffic related tasks, and are essential e.g. for vehicle navigation systems. They also play a major role in topographic mapping. For automatically capturing road junctions from images models are needed, which describe the main aspects. This paper presents an approach to road junction detection based on raster and vector information. The raster features are similar to the ones used in classification approaches. The vector features are derived from a road junction vector model containing edges as road boarders. The whole feature serves as input to an artificial neural network. The neural classifier decides for search window, whether its central pixel is a part of a road junction or not. The developed junction operator was tested on several black-and-white medium resolution orthoimages. The achieved results demonstrate that such junction models can successfully identify three- and four-arms road junctions.

Macular Changes Correlate with the Degree of Acute Anterior Uveitis in Patients with Spondyloarthropathy

Abstract Purpose: To evaluate the correlation between anterior chamber inflammation and the central foveal thickness of the retina in a homogenous uveitic group (seronegative spondyloarthropathy (subgroup: ankylosing spondylitis)) of patients with the first acute anterior unilateral uveitic attack. Methods: Central foveal thickness (FT) and perifoveal retinal thickness were recorded by optical coherence tomography (Optovue RTVue-100), and the difference between the two eyes was calculated. Statistical analysis was performed by Mathworks Matlab software. Anterior chamber inflammation was measured by laser flare photometry (Kowa FM-600). Results: A statistically significant (p < .05) increase was found in retinal thickness in all OCT subfields in acute anterior uveitic eyes compared to healthy fellow eyes. There was a linear correlation between the degree of inflammation (laser flare photometry values) and central foveal thickness (r = .900, p < .001). Conclusions: We could demonstrate close correlation between macular thickness and inflammation in anterior uveitic patients with spondyloarthropathy.

Pedestrian detection by profile laser scanning

Detecting pedestrian flow parameters is of high priority in many future applications. Besides conventional solutions, such as special gates that allow only entering one-by-one or registering individuals by cards, novel remote sensing methods are able to support pedestrian sensing and detection. Mechanical solutions are expensive and slow down pedestrian traffic. Camera systems are able to collect sensitive information that require careful and complicated data handling and privacy management. Besides, acquiring 3D information (e.g. height of humans) is complicated and technical capabilities of optical sensors seriously decrease in case of dark or foggy/smoky environment. This paper discusses a method based on profile laser scanning technology and it gives a brief overview on the high accuracy and data rate laser scanning technology and on the primary application fields. In the proposed method, devices deployed in pedestrian corridors are able to continuously monitor and detect the traffic flow within the range of the sensor. The scanners are mounted in oblique position and scan perpendicular to the pedestrian traveling direction. In addition to the height and width of humans, their position and velocity also can be derived. Based on the collected data, useful information can be obtained for multiple purposes: directed pedestrian count/volume, traffic classification, flow distribution, velocity determination and additional information, e.g. on carried baggage. The proposed method does not involve collecting sensitive information about human beings, thus no privacy issues raised. The suggested technique is safe and has absolutely no hindering effect on the traffic flow, needs only info-communication support and power supply. The developed method has been validated by multiple field measurements. Reference data has been acquired by video that enabled the verification of the count, traveling direction and velocity estimation. The paper presents some statistics of the comparative analysis that prove the accuracy and reliability of the data obtained.

Comparison of spatial reconstruction software packages using DSLR images

Point cloud based modeling has become widely used in recent years in the engineering practice. This paper presents several point clouds produced by different software packages using pixel based multi view spatial reconstruction. The purpose of the investigation is to give statistical quality measures beyond the apparent differences between the point clouds. Reference measurements have been carried out by terrestrial laser scanning. There are proprietary and free software solutions among the tested programs.

Graph based neural self-organization in analyzing remotely sensed images

The Self-Organizing Neuron Graph (SONG) algorithm generalizes the Kohonen feature maps. The technique is proved in analyzing different aerial and satellite images. The base of the method is a given graph; its neurons detect the positions of the data points (pixels), which are derived by image processing functions.

Creating OpenCRG Road Surface Model from Terrestrial Laser Scanning Data for Autonomous Vehicles

The OpenCRG road surface model as part of an open available data format is originally designed for simulation purposes, where the micro level information about the road surface is stored in this format. The paper introduces the whole open data format family (OpenDRIVE, OpenCRG and OpenSCENARIO) and then demonstrates that an efficient road surveying technology, namely the terrestrial laser scanning (TLS) is excellent data source to create realistic description about the road surface. The TLS captured point cloud stores very fine details about the surface, including height irregularities, potholes, cracks, manholes. Because the consortium being behind the developers contain car manufacturers (e.g. Daimler, BMW), mapping agencies (e.g. Here) and technology companies (e.g. Fraunhofer), the usage of the OpenCRG format is expected not only in laboratory circumstances, but later in real life applications. Autonomous vehicles have extreme requests for detailed information about the road and its neighborhood, and the developed open format seems to have excellent support to self-driving vehicle control.

AN OFFLINE PATH PLANNING METHOD FOR AUTONOMOUS VEHICLES

Abstract Driving a road vehicle is a very complex task in terms of controlling it, substituting a human driver with a computer is a real challenge also from the technical side. An important step in vehicle controlling is when the vehicle plans its own trajectory. The input of the trajectory planning are the purpose of the passengers and the environment of the vehicle. The trajectory planning process has several parts, for instance, the geometry of the path-curve or the speed during the way. Furthermore, a traffic situation can also determine many other parameters in the planning process. This paper presents a basic approach for trajectory design. To reach the aim a map will be given as a binary 2204 x 1294 size matrix where the roads will be defined by ones, the obstacles will be defined by zeros. The aim is to make an algorithm which can find the shortest and a suitable way for vehicles between the start and the target point. The vehicle speed will be slow enough to ignore the dynamical properties of the vehicle. The research is one of the first steps to realize automated parking features in a self-drive car.

The Evolution of Central Retinal and Choroidal Thickness in Acute Anterior Uveitic Patients with Spondyloarthropathy

Purpose To describe and correlate the degree of anterior segment inflammation with central retinal and choroidal thickness throughout the treatment period (in the course of follow-up) in the eyes affected with acute anterior uveitis in the patients with seronegative spondyloarthropathy (subgroup: ankylosing spondylitis). Methods Thirty eyes of 30 consecutive Caucasian patients with HLA-B27-associated acute anterior uveitis were included in this study. The flare, AC cell number, and central retinal/choroidal thickness were determined at each visit by optical coherence tomography and laser flare photometry. Treatment consisted of topical corticosteroids. Statistical analysis was performed by MathWorks Matlab software. Results In the follow-up period, central retinal thickness was increased in the first 9-10 days and then decreased until stabilization (after 5-6 weeks). The flare and AC cell number decreased rapidly at the beginning of the treatment, in the first 10 days, and thereafter, slower decrease could be observed until complete resolution of inflammation. Statistically significant, positive correlation was found between initial laser flare value and maximal central retinal thickness (r=0.881, p < 0.001). Positive correlation between flare and retinal thickening was observable throughout the treatment period. Central choroidal thickness was decreased also significantly during the follow-up (p < 0.001). Conclusions The retina and choroid may play a biomarker function in the anterior segment inflammation of the eye in the patients with seronegative spondyloarthropathy.