Arama Sonuçları
Listeleniyor 1 - 7 / 7
Yayın Driver recognition using gaussian mixture models and decision fusion techniques(Springer-Verlag Berlin, 2008) Benli, Kristin Surpuhi; Düzağaç, Remzi; Eskil, Mustafa TanerIn this paper we present our research in driver recognition. The goal of this study is to investigate the performance of different classifier fusion techniques in a driver recognition scenario. We are using solely driving behavior signals such as break and accelerator pedal pressure, engine RPM, vehicle speed; steering wheel angle for identifying the driver identities. We modeled each driver using Gaussian Mixture Models, obtained posterior probabilities of identities and combined these scores using different fixed mid trainable (adaptive) fusion methods. We observed error rates is low as 0.35% in recognition of 100 drivers using trainable combiners. We conclude that the fusion of multi-modal classifier results is very successful in biometric recognition of a person in a car setting.Yayın Annular CMUT arrays for side looking intravascular ultrasound imaging(IEEE, 2007) Zahorian, Jaime; Güldiken, Rasim Oytun; Gürün, Gökçe; Qureshi, Muhammad Shakeel; Balantekin, Müjdat; Değertekin, Fahrettin Levent; Carlier, Stephane; Şişman, Alper; Karaman, MustafaAlthough side looking intravascular ultrasound (SL-IVUS) imaging systems using single element piezoelectric transducers set the resolution standard in the assessment of the extent of coronary artery disease, improvements in transducer performance are needed to perform harmonic imaging and high resolution imaging of vulnerable plaque. With their small channel count; annular arrays exploiting the inherent broad bandwidth of CMUTs and electronic focusing capability of integrated electronics provide a path for desired SL-IVUS imaging catheters. In this paper, we first describe the design, low temperature fabrication of an 8401 mu m diameter, 8 element CMUT annular array. Testing of the individual elements in oil shows a uniform device behavior with 100% fractional bandwidth around 20MHz without including the effects of attenuation and diffraction. We also present linear scan imaging results obtained on wire targets in oil, tissue and tissue mimicking phantoms using both unfocused and dynamically focused transducers. The results for axial and lateral resolution are in agreement predicted by the simulations and show the feasibility of this approach for high resolution SL-IVUS imaging.Yayın Feature extraction in shape recognition using segmentation of the boundary curve(Elsevier Science BV, 1997-10) Özuğur, Timuçin; Denizhan, Yağmur; Panayırcı, ErdalWe present a new method for feature extraction of two-dimensional shape information based on segmentation of the boundary curve. This approach partitions closed shapes into segments and finds their angular spans. The number of segments and the angular spans form the first two feature parameters of a given shape. Fourier coefficients of all segments constitute the final feature parameters. The algorithm renders the shapes independent of scale, rotation and translation, The main advantage of this method is to speed up substantially the recognition process of the shapes, mainly because it is possible to design the classification rule in a hierarchical way. It is therefore suitable for objects to be sorted in a factory environment where the silhouette boundary supplies sufficient information for identification.Yayın Kernel likelihood estimation for superpixel image parsing(Springer Verlag, 2016) Ateş, Hasan Fehmi; Sünetci, Sercan; Ak, Kenan EmirIn superpixel-based image parsing, the image is first segmented into visually consistent small regions, i.e. superpixels; then superpixels are parsed into different categories. SuperParsing algorithm provides an elegant nonparametric solution to this problem without any need for classifier training. Superpixels are labeled based on the likelihood ratios that are computed from class conditional density estimates of feature vectors. In this paper, local kernel density estimation is proposed to improve the estimation of likelihood ratios and hence the labeling accuracy. By optimizing kernel bandwidths for each feature vector, feature densities are better estimated especially when the set of training samples is sparse. The proposed method is tested on the SIFT Flow dataset consisting of 2,688 images and 33 labels, and is shown to outperform SuperParsing and some of its extended versions in terms of classification accuracy.Yayın Imaging of rough surfaces by RTM method(IEEE, 2024) Sefer, Ahmet; Yapar, Ali; Yelkenci, TanjuAn electromagnetic imaging framework is implemented utilizing a single frequency reverse time migration (RTM) technique to accurately reconstruct inaccessible two-dimensional (2D) rough surface profiles from the knowledge of scattered field data. The unknown surface profile, which is expressed as a 1D height function, is either perfectly electric conducting (PEC) or an interface between two penetrable media. For both cases, it is assumed that the surface is illuminated by a number of line sources located in the upper medium. The scattered fields, which should be collected by real measurements in practical applications, are obtained synthetically by solving the associated direct scattering problem through the surface integral equations. RTM is subsequently applied to generate a cross-correlation imaging functional which is evaluated numerically and provides a 2D image of the region of interest. A high correlation is observed by the functional in the regions where the transitions between two media occur. Hence, it results in the acquisition of the unknown surface profile at the sites where the functional attains its highest values. The efficiency of the proposed method is comprehensively tested by numerical examples covering various types of scattering scenarios.Yayın Evaluation of feature selection and encoding methods for superpixel image parsing(Işık Üniversitesi, 2017-12-14) Sünetci, Sercan; Ateş, Hasan Fehmi; Işık Üniversitesi, Fen Bilimleri Enstitüsü, Elektronik Mühendisliği Yüksek Lisans ProgramıThis thesis is about image parsing which is one of the important problems in computer vision. The goal of image parsing is segmentation of object and labeling of each object. Recently, a popular way of image segmentation and classifcation is superpixels. Image is segmented into visually logical small regions by using superpixel algorithm and then, superpixels are parsed into diferent classes. Classifcation performance is signifcantly afected by the properties of superpixel algorithm and parametric settings. SuperParsing is one of the superpixel-based image parsing algorithm and provides a succesful nonparametric solution for image segmentation and classifcation problem without any need for classifer training. SuperParsing labels each superpixel based on feature matching between the superpixel and a subset of the training superpixels. The training subset is determined by global matching between the test image and the training set. For superpixel matching the method makes use of a rich set of superpixel features. Class conditional log-likelihood is computed based on these matched features. The main objective of this thesis is to show improvements in labeling accuracy percentage by using feature encoding and selection methods, including learned features from Convolutional Neural Network (CNN) models. We perform two different encoding methods to selected features of superpixels and show that feature encoding improves parsing accuracy. The applied feature encoding methods are locality-constrained linear encoding (LLC) and kernel codebook encoding (KCB). LLC encoding method gives us 2:6% improvement on per-pixel accuracy for SIFT Flow dataset and 6:8% improvement on per-pixel accuracy for 19-class LabelMe dataset. KCB encoding method gives us 3:6% improvement on per-pixel accuracy for SIFT Flow dataset and 6:2% improvement on per-pixel accuracy for 19-class LabelMe dataset. All these results are overall improvement which are computed over original SuperParsing. Most recent studies about image segmentation and classifcation use CNN tiioi improve their accuracy percentage. Features extracted from pre-trained networks, which are trained on large image databases, can be used in addition to handcrafted features in image segmentation. Last layer of these CNN models give the best features for classifcation. We test learned CNN features together with KCB or LLC encoding methods. We use CNN features both for global matching and superpixel matching. These tests give us 7:3% overall improvement over originalSuperParsing on SIFT Flow dataset and 10:3% overall improvement over original SuperParsing on 19-class LabelMe dataset.Yayın Segmentation based classification of retinal diseases in OCT images(Institute of Electrical and Electronics Engineers Inc., 2024) Eren, Öykü; Tek, Faik Boray; Turkan, YaseminVolumetric optical coherence tomography (OCT) scans offer detailed visualization of the retinal layers, where any deformation can indicate potential abnormalities. This study introduced a method for classifying ocular diseases in OCT images through transfer learning. Applying transfer learning from natural images to Optical Coherence Tomography (OCT) scans present challenges, particularly when target domain examples are limited. Our approach aimed to enhance OCT-based retinal disease classification by leveraging transfer learning more effectively. We hypothesize that providing an explicit layer structure can improve classification accuracy. Using the OCTA-500 dataset, we explored various configurations by segmenting the retinal layers and integrating these segmentations with OCT scans. By combining horizontal and vertical cross-sectional middle slices and their blendings with segmentation outputs, we achieved a classification a ccuracy of 91.47% and an Area Under the Curve (AUC) of 0.96, significantly outperforming the classification of OCT slice images.
