Wegdan Abdelsalam

Shape representation and description using the Hilbert curve

In this paper, a novel linear-time approach to shape representation and description is presented. The object shape is captured by scanning the object image using a space-filling curve (SFC). The resulting vector is smoothed, using wavelet approximation, and sampled. In addition, the concept of key feature points (KFPs) is introduced to utilize a priori information about the classification of the images in the database in optimizing the representation of the objects within each class. The proposed technique achieves a recognition rate of 88.3% on the MPEG-7 core experiment part B. On the Kimia-99 and Kimia-216 datasets, a precision average of 95.6% is attained. Retrieval rates of 94.2% and 95.6% are achieved on the gray-scale and binary versions of the ETH-80 dataset, respectively.

Proposing a hybrid tag-camera-based identification and navigation aid for the visually impaired

In this paper a barcode-based system to help the visually impaired identity objects in the environment and navigate through unknown territories is introduced. The system is based on the idea of tagging the different objects with 2D barcodes. With the aid of a portable camera and a computing device, the system can recognize and relay the barcode content to the user. We present the advantages of the proposed system compared to those of existing technologies. The different steps for recognizing and extracting the barcodes are described and applied to a sample image.

An Efficient Shape Representation and Description Technique

In this paper, we present a novel approach to shape representation and description based on the combination of the Hilbert space filling curve and wavelet analysis. Our objective is to capitalize on the localization-preserving nature of the Hilbert space filling curve and the approximation power of the wavelet transform. The object image is scanned using the Hilbert curve and the resulting vector is smoothed using the wavelet transform and sampled. The technique is O(N) for both representation and comparison. We present some experimental results on the MPEG-7 dataset, Kimia-99 dataset, ETH-80 dataset, and a logo dataset.

A Template-Based Shape Representation Technique

In this paper we present a novel approach to shape representation based on correlating a set of object Regions of Interest (RoI) with a set of shape templates. The resultant correlations are the shape features used to build a Template-based Shape Feature Vector (TSFV) that represents the shape of the object. For each class of objects, a set of Main Shape Features (MSFs) is determined so that only the most descriptive features are used when comparing shapes. The proposed technique is tested on two benchmark databases, Kimia-99 and Kimia-216 and is shown to produce competitive results.

A Roving User Modeling Framework for Location Tracking Applications

With the advent of GPS systems, location tracking has become an essential component of fleet management systems. Reliable location estimation of the location of each member of the fleet is critical for optimum operation of organizations such as the police force, taxi-cab, and trucking companies. In this paper a framework for modeling drivers in such environments for location tracking purposes is proposed. The frameworks goal is to increase the LTSs ability to accurately answer queries about the whereabouts of its fleet members in the past, present, and the future

Significantly improving scan-based shape representations using rotational key feature points

In a previous paper we have presented the idea of representing the shape of a 2D object by scanning it following a Hilbert curve then performing wavelet smoothing and sampling. We also introduced the idea of using only a subset of the resulting signature for comparison purposes. We called that set the Key Feature Points (KFPs). In this paper we introduce the idea of taking the KFPs over a number of views of the original shape. The proposed improvement results in a significant increase in recognition rates when applied to the MPEG-7 and ETH-80 data sets when the Hilbert scan is used. Similar improvement is achieved when the raster scan is used.

Shape Matching Using a Novel Warping Distance Measure

This paper presents a novel distance measure, the Minimum Landscape Distance (MLD). MLD is a warping distance measure that provides a non-linear mapping between the elements in one sequence to those of another. Each element in one sequence is mapped to that with the highest neighborhood structural similarity (landscape) in the other sequence within a window. Different window sizes are tested on a number of datasets and a linear relationship between the window size and the sequence size is discovered. Experimental results obtained on the Kimia-99 and Kimia-216 datasets show that MLD is superior to the Euclidean, correlation, and Dynamic Time Warping (DTW) distance measures.

A view-based 3D object shape representation technique

In this paper we present a novel approach to 3D shape representation and matching utilizing a set of shape representations for 2D views of the object. The proposed technique capitalizes on the localizationpreserving nature of the Hilbert space filling curve and the approximation capabilities of the Wavelet transform. Each 2D view of the object is represented by a concise 1D representation that can be used to search an image database for a match. The shape of the 3D image is represented by the set of 1D representations of its 2D views. Experimental results, on a subset of the Amsterdam Library of Object Images (ALOI) dataset, are provided.

Managing uncertainty in location based applications

In previous work we introduced the idea of user modeling as a means of reducing uncertainty in location tracking of human-controlled moving objects. We used the name roving users (RU for short) to refer to this subset of moving objects. In this paper we discuss the issue of reducing the complexity of the user model caused by variables with high number of possible values. We show how self organizing maps (SOM) could be used to classify values of a certain variable such that the classes are used - rather than actual values - in calculating the conditional probabilities of child variables. We support our proposed technique with some experimental results

A user modeling approach to improving estimation accuracy in location-tracking applications

Summary form only given. Location tracking systems are discrete in nature location information about each moving object (MO) is sampled at certain points in time. To determine the location of a MO between location reports or sometime in the future, we have to estimate the location at that point in time using the location reports we already have. The sampling frequency could affect the systems estimation accuracy as well as operating costs. Poor estimation accuracy could also carry a cost. The objective is to maximize the estimation accuracy while minimizing the operating cost. In this paper, we introduce the novel idea of using user modeling to improve the estimation accuracy of both the route and speed of a MO without the need to increase the sampling rate. We focus on a subset of moving objects we call roving users, or RUs for short. A RU is a human or human-controlled MO. The idea is that humans are creatures of habit. Knowing how a RU behaved in the past could help us estimate what he/she will do in the future. This knowledge could help us estimate the user location more accurately.