Papers by Emmanuel Karlo Nyarko
Electronics
Chronic wounds are a heavy burden on medical facilities, so any help in treating them is most wel... more Chronic wounds are a heavy burden on medical facilities, so any help in treating them is most welcome. Current research focuses on wound analysis, especially wound tissue classification, wound measurement, and wound healing prediction to assist medical personnel in wound treatment, with the main goal of reducing wound healing time. The first phase of wound analysis is wound segmentation, where the task is to extract wounds from the healthy tissue and image background. In this work, a standard feedforward neural network was developed for the purpose of wound segmentation using data from the MICCAI 2021 Foot Ulcer Segmentation (FUSeg) Challenge. It proved to be a simple yet efficient method for extracting wounds from images. The proposed algorithm is part of a compact system that analyzes chronic wounds using a robotic manipulator, RGB-D camera and 3D scanner. The feedforward neural network consists of only five fully connected layers, the first four with Rectified Linear Unit (ReLU) ...
Sensors, 2021
Chronic wounds, or wounds that are not healing properly, are a worldwide health problem that affe... more Chronic wounds, or wounds that are not healing properly, are a worldwide health problem that affect the global economy and population. Alongside with aging of the population, increasing obesity and diabetes patients, we can assume that costs of chronic wound healing will be even higher. Wound assessment should be fast and accurate in order to reduce the possible complications, and therefore shorten the wound healing process. Contact methods often used by medical experts have drawbacks that are easily overcome by non-contact methods like image analysis, where wound analysis is fully or partially automated. This paper describes an automatic wound recording system build upon 7 DoF robot arm with attached RGB-D camera and high precision 3D scanner. The developed system presents a novel NBV algorithm that utilizes surface-based approach based on surface point density and discontinuity detection. The system was evaluated on multiple wounds located on medical models as well as on real pate...
Chua’s model of nonlinear coil in a ferroresonant circuit obtained using Dommel’s method and grey box modelling approach
Nonlinear Dynamics, 2016
Chua’s model of the nonlinear coil is often used in ferroresonance research. In this paper, two m... more Chua’s model of the nonlinear coil is often used in ferroresonance research. In this paper, two methods are used to determine such a model: usual Dommel’s method and a grey box modelling approach, which is a novel way to model the nonlinear coil in a ferroresonant circuit. The parameters of the grey box model are determined by nonlinear regression using a metaheuristic optimization algorithm. The obtained models are evaluated by comparing the steady-state types obtained by measurements with those obtained using simulation. Comparing the obtained Chua’s models, results indicate that the grey box modelling approach is more accurate than Dommel’s method for lower coil nominal voltage. However, the main conclusion is that Chua’s model in general cannot be used to predict all steady-state types typical for ferroresonance, i.e. it can be used in a highly restricted manner for appearance of certain steady-state types only.
Materials
Replacing a specified quantity of cement with Class F fly ash contributes to sustainable developm... more Replacing a specified quantity of cement with Class F fly ash contributes to sustainable development and reducing the greenhouse effect. In order to use Class F fly ash in self-compacting concrete (SCC), a prediction model that will give a satisfactory accuracy value for the compressive strength of such concrete is required. This paper considers a number of machine learning models created on a dataset of 327 experimentally tested samples in order to create an optimal predictive model. The set of input variables for all models consists of seven input variables, among which six are constituent components of SCC, and the seventh model variable represents the age of the sample. Models based on regression trees (RTs), Gaussian process regression (GPR), support vector regression (SVR) and artificial neural networks (ANNs) are considered. The accuracy of individual models and ensemble models are analyzed. The research shows that the model with the highest accuracy is an ensemble of ANNs. T...
European Conference on Mobile Robots, 2009
An approach for registration of sparse feature sets detected in two stereo image pairs taken from... more An approach for registration of sparse feature sets detected in two stereo image pairs taken from two different views is proposed. Analogously to many existing image registration approaches, our method consists of initial matching of features using local descriptors followed by a RANSAC-based procedure. The proposed approach is especially suitable for cases where there is a high percentage of false initial matches. The strategy proposed in this paper is to modify the hypothesis generation step of the basic RANSAC approach by performing a multiplestep procedure which uses geometric constraints in order to reduce the probability of false correspondences in the hypothesis. The algorithm needs approximate information about the relative camera pose between the two views obtained e.g. by odometry. However, the uncertainty of this information is allowed to be rather high. The presented technique is evaluated using both synthetic data and real data obtained by a stereo camera system.
European Conference on Mobile Robots, 2011
A range image segmentation approach is presented. A range image obtained by a 3D sensor is transf... more A range image segmentation approach is presented. A range image obtained by a 3D sensor is transformed to a 2.5D triangle mesh. This mesh is then segmented into approximately convex surfaces by an iterative procedure based on the incremental convex hull algorithm and region growing. The proposed approach is designed as a fast mid-level tool whose task is to provide suitable geometric primitives for a high-level object recognition or robot localization algorithm. The presented approach is experimentally evaluated using 3D data obtained by a Kinect sensor.
Natural Hazards, 2016
Forecasting and operational routing flood requires accurate forecasts on proper feed time, to be ... more Forecasting and operational routing flood requires accurate forecasts on proper feed time, to be able to issue suitable warnings and take suitable emergency actions. Floodrouting problem is one of the most complicated matters in hydraulics of open channels and river engineering. Flood routing is the process of computing the progressive time and shape of a flood wave at successive points along a river. To get an approximate solution of the flood-routing problem, different techniques are used. This paper describes an approach to train artificial neural network (ANN) using social-based algorithm (SBA). The approach illustrates feed-forward neural network optimization for the flood-routing problem of Kheir Abad River called FF-SBA. To this end, the number and effective time lag of input data in ANN models are initially determined by means of linear correlation between input and output time series; subsequently, the weights of the feed-forward network is optimized by
10th IFAC Symposium on Robot Control, 2012
A fast robot pose tracking algorithm based on planar segments extracted from range images is pres... more A fast robot pose tracking algorithm based on planar segments extracted from range images is presented. A range image obtained from a 3D sensor is transformed to a 2.5D triangle mesh from which planar segments are extracted. Using information provided by each planar segment based on its size and orientation, a directed search hypothesis generation algorithm using a tree structure is presented. The presented approach is experimentally evaluated using 3D data obtained by a Kinect sensor mounted on a mobile robot. Results indicate that the proposed method is much faster than similar previously proposed methods.
Detection of Dominant Planar Surfaces in Indoor Environments Using Stereo Vision and Image Segmentation
28th International Conference …, 2010
Pregled bibliografske jedinice broj: 537044. Zbornik radova. Autori: Nyarko, Emmanuel Karlo; Cupe... more Pregled bibliografske jedinice broj: 537044. Zbornik radova. Autori: Nyarko, Emmanuel Karlo; Cupec, Robert; Delalija, Aleksandar. Naslov: Detection of Dominant Planar Surfaces in Indoor Environments Using Stereo Vision and Image Segmentation. ...
An approach for registration of sparse feature sets detected in two stereo image pairs taken from... more An approach for registration of sparse feature sets detected in two stereo image pairs taken from two different views is proposed. Analogously to many existing image registration approaches, our method consists of initial matching of features using local descriptors followed by a RANSAC-based procedure. The proposed approach is especially suitable for cases where there is a high percentage of false initial matches. The strategy proposed in this paper is to modify the hypothesis generation step of the basic RANSAC approach by performing a multiplestep procedure which uses geometric constraints in order to reduce the probability of false correspondences in the hypothesis. The algorithm needs approximate information about the relative camera pose between the two views obtained e.g. by odometry. However, the uncertainty of this information is allowed to be rather high. The presented technique is evaluated using both synthetic data and real data obtained by a stereo camera system.
Automatika: Journal for …, 2009
An approach for registration of sparse feature sets detected in two stereo image pairs taken from... more An approach for registration of sparse feature sets detected in two stereo image pairs taken from two different views is proposed. Analogously to many existing image registration approaches, our method consists of initial matching of features using local descriptors followed by a RANSAC-based procedure. The proposed approach is especially suitable for cases where there is a high percentage of false initial matches. The strategy proposed in this paper is to modify the hypothesis generation step of the basic RANSAC approach by performing a multiple-step procedure which uses geometric constraints in order to reduce the probability of false correspondences in generated hypotheses. The algorithm needs approximate information about the relative camera pose between the two views. However, the uncertainty of this information is allowed to be rather high. The presented technique is evaluated using both synthetic data and real data obtained by a stereo camera system.
The image processing described in this paper is used for visual quality control in ceramic tile p... more The image processing described in this paper is used for visual quality control in ceramic tile production. The tiles surface quality depends on the surface defects. The described image processing is based on the neural network approach. The described diagnostic algorithm is presented to detect surface failures on white ceramic tiles. The tiles are scanned and the digital images are pre-processed and classified using neural networks. Pre-processing of the image data is used to keep the number of inputs of the neural networks performing the classification relatively small. It is important to reduce the amount of input data with problem specific pre-processing. The auto-associative neural network is used for feature generation and selection while the probabilistic neural network is used for classification. Simulation was performed in Matlab using the Neural Network Toolbox. The algorithm is evaluated experimentally using the real tile images. The analysis of the detection capabilities...
The image processing described in this paper is used for visual quality inspection in ceramic til... more The image processing described in this paper is used for visual quality inspection in ceramic tile production. The tiles surface quality depends on the surface defects and has the influence to the tile quality classification. The described image processing is based on the neural network approach. The described diagnostic algorithm is presented to detect surface failures on white ceramic tiles. The tiles are scanned and the digital images are pre-processed and classified using neural networks. Pre-processing of the image data is used to keep the number of inputs of the neural networks performing the classification relatively small. It is important to reduce the amount of input data with problem specific pre-processing. Statistical methods are used in the pre-proccesing of the image data. For classification purposes, a probabilistic neural network and a standard feedfoward neural network are used and the results obtained are compared. The analysis of the detection capabilities is done...
LeArEst — The software for border and area estimation of data measured with additive error
2017 International Symposium ELMAR, 2017
This paper proposes a solution to the problem of estimating object dimensions from a noisy image.... more This paper proposes a solution to the problem of estimating object dimensions from a noisy image. Image noise can be produced by the physical processes of imaging, or can be caused by the presence of some unwanted structures (e.g. soft tissue captured in X-ray images of bones). We suppose that the data are drawn from uniform distribution on the object of interest, but contaminated by an additive error having normal or Laplace distribution. The software for border estimation of an object registered in such manner has been developed, and brief description of its key functions is given. The software is able to estimate the borders of an object on a given line that intersects it, as well as to estimate its area. Its input data may be numerical data, as well as images in JPEG format.
Statistical estimation of the object’s area from the image contaminated with additive noise
2020 15th Iberian Conference on Information Systems and Technologies (CISTI)
Area estimation of circular or ellipsoidal object on an image is a current issue in computer visi... more Area estimation of circular or ellipsoidal object on an image is a current issue in computer vision. Several methods that address this problem have been previously presented, but it turned out that they do not give satisfactory results when dealing with noisy images. As part of the research presented in this paper, a statistical model for estimating the width of uniform distribution for data contaminated with additive error was applied in order to approach the mentioned problem in an innovative manner. Initially, a method for length estimation of intersection of an object with an arbitrary line has been developed. It is possible to estimate the set of object’s edge points using this method. Further, a circle or an ellipse is fitted in that set of points and its area is calculated, which approximates the area of the object itself. It is also possible to estimate the area of a circular or ellipsoidal object represented by a set of points in the plane.The presented method was implemented and publicly released as a package for the programming language R. The method has been extensively tested on the problem of estimating the area of objects recorded using RGB-D camera. Different noise levels were added to the captured images, and estimation results were compared with the ones obtained by several established methods. The test results showed that the method presented in this paper gives qualitatively the best results of area estimation when dealing with noisy images.
Place Recognition Based on Planar Surfaces Using Multiple RGB-D Images Taken From the same Position
2019 European Conference on Mobile Robots (ECMR), 2019
This paper considers indoor place recognition based on matching of planar surfaces and straight e... more This paper considers indoor place recognition based on matching of planar surfaces and straight edges extracted from depth images obtained by an RGB-D camera. The idea of using planar surfaces as landmarks for robot localization has already been investigated. In this paper, the advantage of using multiple RGB-D images acquired from the same viewpoint by a camera mounted on a pan-tilt head is addressed. This simple straightforward method of expanding the field of view of a standard RGB-D camera allows 3D models of the observed place to be built, which contain information about relative positions of geometric features that are not contained within a single camera FoV. A high recognition rate is achieved indicating the practical applicability of the investigated approach. A publicly available dataset for the evaluation of place recognition methods is created. Using this dataset, the ability of recognizing places from viewpoints that differ from those from which the model is built can b...
Osnove automatskog upravljanja: priručnik za laboratorijske vježbe
In this paper we present the current progress and results from the filtering of Croatian Meteor N... more In this paper we present the current progress and results from the filtering of Croatian Meteor Network video meteor detections using soft computing methods such as neural networks and support vector machines (SVMs). The goal is to minimize the number of false-positives while preserving the real meteor detections. This is achieved by pre-processing the data to extract meteor movement parameters and then recognizing patterns distinct to meteors. The input data format is fully compliant with the CAMS meteor data standard, and as such the proposed method could be utilized by other meteor networks of the similar kind.
A range image segmentation approach is presented. A range image obtained by a 3D sensor is transf... more A range image segmentation approach is presented. A range image obtained by a 3D sensor is transformed to a 2.5D triangle mesh. This mesh is then segmented into approximately convex surfaces by an iterative procedure based on the incremental convex hull algorithm and region growing. The proposed approach is designed as a fast mid-level tool whose task is to provide suitable geometric primitives for a high-level object recognition or robot localization algorithm. The presented approach is experimentally evaluated using 3D data obtained by a Kinect sensor.
Uploads
Papers by Emmanuel Karlo Nyarko