Alper Yildirim

Classifier-based offline feature selection and evaluation for visual tracking of sea-surface and aerial targets

An offline feature selection and evaluation mechanism is used in order to develop a robust visual tracking scheme for sea-surface and aerial targets. The covariance descriptors, known to constitute an effi- cient signature set in object detection and classification problems, are used in the feature extraction phase of the proposed scheme. The per- formance of feature sets are compared using support vector machines, and those resulting in the highest detection performance are used in the covariance based tracker. The tracking performance is evaluated in dif- ferent scenarios using different performance measures with respect to ground truth target positions. The proposed tracking scheme is observed to track sea-surface and aerial targets with plausible accuracies, and the results show that gradient-based features, together with the pixel locations and intensity values, provide robust target tracking in both surveillance scenarios. The performance of the proposed tracking strategy is also compared with some well-known trackers including correlation, Kanade- Lucas-Tomasi feature, and scale invariant feature transform-based track- ers. Experimental results and observations show that the proposed target tracking scheme outperforms other trackers in both air and sea surveil- lance scenarios. C

Automated fixed-point precision optimization

An algorithm which minimizes the hardware resources of fixed-point operations for a given accuracy is presented. For range analysis, forward propagation is used in order to determine the ranges of the intermediate signals. For the precision analysis, a search algorithm is used. Analytical quantization error models are used with the precision search algorithm. The main novelty of this work is the use of the weighted search algorithm which can minimize the hardware resources for a specific FPGA. By using weights, the search algorithm is modified to produce an optimum result for the target FPGA for a specific hardware resource such as multipliers or blok rams.

A clustering approach for radar warning receivers

In this study a mission data base based clustering approach that can be used in radar warning receivers for the purpose of deinterleaving is suggested. Cell based deinterleaving technique, which is widely used at the present time, utilizes the information of direction of arrival, frequency and pulse width. In this study, different from this approach used in the literature, frequency, direction of arrival and pulse amplitude parameters are utilized for deinterleaving. With this technique it is shown that accurate results can be obtained by simulation.

Unsupervised classification of multispectral Landsat images with multidimensional particle swarm optimization

This article proposes a novel unsupervised classification approach for automatic analysis of multispectral Landsat images. The automatic classification of the information in multidimensional (MD) Landsat data space by dynamic clustering is addressed as an optimization problem and two recently proposed heuristic techniques based on Particle Swarm Optimization (PSO) are applied to determine the optimal (number of) clusters in a given input data space: distance metric and a proper validity index function. The first technique, the so-called MD-PSO, re-forms the native structure of swarm particles (agents) in such a way that they can make inter-dimensional passes with a dedicated dimensional PSO process. Fractional global best formation (FGBF) basically collects all promising dimensional components and fractionally creates an artificial global best (aGB) agent that has the potential to be a better ‘guide’ than the swarm’s native global best position (gbest) agent. In this study, the proposed dynamic clustering approach based on MD-PSO and FGBF techniques is applied to automatically classify the colour-coded representations of the multispectral (MD) Landsat data. The approach has been applied to real-world multispectral data and it provided quite encouraging results compared to the traditional K-means and ISODATA (iterative self-organizing data analysis) clustering methods. The proposed unsupervised technique determines the true number of classes within Landsat data for optimal classification performance while preserving spatial resolution and textural information in the classification map.

Deinterleaving for radar warning receivers

Modern radar warning receivers form the clusters by processing the incoming pulses using multiple parameter deinterleaving. Validation and identification of formed sequences obtained from formed clusters are made by time-of-arrival (TOA) deinterleaving. Especially, validation and identification of the sequences utilizing as minimum number of pulses as possible are of utmost importance. The improved method presented in this paper focuses mainly on identifying the sequence having small number of pulses with a possibility of missed pulse. The algorithm uses the predefined pulse repetition interval (PRI) range during TOA deinterleaving to tackle the aforementioned situations.

FPGA-based match filter implementation in frequency domain using an overlap-add method

In this research, a real time matched filter is implemented on FPGA using an overlap-add method. The matched filter that increases the signal-to-noise ratio (SNR) for pulse compression and low probability intercept (LPI) radars is implemented in digital domain. This design is implemented on Xilinx Virtex-5 based processing board that samples in intermediate frequency (2.5 GHz). In the overlap-add method, we propose to design the matched filter by using two parallel FFT cores. Furthermore, the matched filter results are presented for different intra-pulse modulations.

Electronic attack techniques validation environment

In this article, the development of the electronic attack (EA) techniques validation environment is presented. EA techniques validation environment is a signal based closed hardware in loop which operates in intermediate frequency. By the development of this environment, radar and EA signals are processed on real time FPGA and the impact of radar and EA systems to each other are tested. Laboratory test environment provides cost/time advantage on both increasing the resistivity of radar systems to electronic attacks and the effectiveness of the electronic attack on the radar systems.

Salient point region covariance descriptor for target tracking

Abstract. Features extracted at salient points are used to construct a region covariance descriptor (RCD) for target tracking. In the classical approach, the RCD is computed by using the features at each pixel location, which increases the computational cost in many cases. This approach is redundant because image statistics do not change significantly between neighboring image pixels. Furthermore, this redundancy may decrease tracking accuracy while tracking large targets because statistics of flat regions dominate region covariance matrix. In the proposed approach, salient points are extracted via the Shi and Tomasi’s minimum eigenvalue method over a Hessian matrix, and the RCD features extracted only at these salient points are used in target tracking. Experimental results indicate that the salient point RCD scheme provides comparable and even better tracking results compared to a classical RCD-based approach, scale-invariant feature transform, and speeded-up robust features-based trackers while providing a computationally more efficient structure.

Target range estimation based on a non-homogenous poisson process model

In this study, the author analyses target range estimation errors in matched filtering-based detection performed in high range resolution (HRR) radars. Conventional radar signal processors use point target detectors, where extended target responses are put through a point detection process by windowing and thresholding. The author demonstrates through simulations that the performance of degradation under the point target assumption can be significant for HRR radars, where targets extend across several detection cells. The author modelled the reflections for stationary and moving extended target scenarios by using three target signal models (TSMs). A non-homogenous Poisson process (NHPP) is provided to model the signal at the output of the target detector, which includes reflections from targets and clutter. The corresponding maximum likelihood (ML) estimator is derived as a range estimation technique. The author simulated a target detection process and made comparisons between the classical- and NHPP-based peak estimator performances for each of the TSMs. Furthermore, the ML estimation (MLE) algorithm is extended for multiple targets. The author demonstrates that the ML estimator significantly reduces target range estimation errors compared with the classical point target estimators.

Real-time frequency parameter extraction for electronic support systems

In this research, real time automatic frequency parameter extraction methods for electronic support systems are implemented on FPGA. The frequency parameter is extracted by using digital instantaneous frequency measurement (DIFM) and fast fourier transform (FFT) methods. Estimation performances of these two different methods for different type of radars at different signal to noise ratios (SNR) are analyzed. This design is implemented on Xilinx Virtex-6 based processing board that samples in 2.5 GHz. Experimental results of the design for FPGA implementation are presented.