Alexander N. Dolia

The minimum volume covering ellipsoid estimation in kernel-defined feature spaces

Minimum volume covering ellipsoid estimation is important in areas such as systems identification, control, video tracking, sensor management, and novelty detection. It is well known that finding the minimum volume covering ellipsoid (MVCE) reduces to a convex optimisation problem. We propose a regularised version of the MVCE problem, and derive its dual formulation. This makes it possible to apply the MVCE problem in kernel-defined feature spaces. The solution is generally sparse, in the sense that the solution depends on a limited set of points. We argue that the MVCE is a valuable alternative to the minimum volume enclosing hypersphere for novelty detection. It is clearly a less conservative method. Besides this, we can show using statistical learning theory that the probability of a typical point being misidentified as a novelty is generally small. We illustrate our results on real data.

Data fusion and processing for airborne multichannel system of radar remote sensing: methodology, stages, and algorithms

Methodology and stages of data processing in multichannel airborne radar imaging systems are considered. It is shown that data fusion in such systems requires special techniques, algorithms, and software for image processing and information retrieval. Some approaches and methods are proposed. The results are demonstrated for simulated and real images.

Dichoptic motion perception limited to depth of fixation

When counterphase spatio-temporal flicker is presented to the left and right eye continuous directional motion can be perceived. Here, we investigate whether this type of dichoptic motion can be observed at different depth planes. Four observers indicated direction of motion for dichoptic motion stimuli, presented in a context containing crossed and uncrossed disparity information in different conditions. Our results show that despite the presence of disparity cues in the stimulus, discrimination of motion direction remained maximal at interocular phase offsets that correspond to binocular motion perception at zero disparity. This constraint brings into question perception of dichoptic motion as the result of an early binocular motion system. We compared our results with predictions of a computational stereo-motion model [Qian, N. (1994). Computing stereo disparity and motion with known binocular cell properties. Neural Computations, 6, 390-404; Qian, N., & Andersen, R. A. (1997). A physiological model for motion-stereo integration and a unified explanation of Pulfrich-like phenomena. Vision Research, 37, 1683-1698]. In contrast to our empirical results, simulations of cell activation in this hybrid energy model predict maximal activation at non-zero disparities. It is concluded that perception of dichoptic motion is a by-product of early interocular combination at low contrasts rather than the result of a dedicated stereo-motion system.

Neural networks for local recognition of images with mixed noise

A new approach to neural network (NN) application for local recognition of images with mixed noise is put forward. Although some pixels in images can be corrupted by spikes the proposed technique permits to eliminate uncertainty observed in this case and to correctly recognize the pixels that, in fact, correspond to an edge, a homogeneous region or an small-sized object. For this purpose a procedure of recognition of one among four basic hypotheses and additional determination of spike properties within the scanning window is proposed. This recognition task is performed for two groups of outputs (classes) of one common NN. The problems of NN learning and structure selection for this case are discussed. The performance of neural network classifier is analyzed for different input data types and for various characteristics of noise. An improvement in correct recognition is shown for the proposed approach in comparison to previous work.

Mitigation of image impairments for multichannel remote sensing data fusion

Whilst for the majority of applications image quality depends on sensor accuracy and principles of image formation, in remote sensing systems information is also degraded by communication errors. To improve image fusion results in the presence of communication and sensor impairments we propose a two-stage approach. Preliminary nonlinear locally-adaptive image processing is applied at the first stage for mitigating impairments produced in image sensors and communication systems, and fusion algorithms are used at the second stage. The efficiency of the proposed algorithms is demonstrated for satellite remote sensing images and simulated data with similar characteristics and distortions. The influence of image distortions and the effectiveness of mitigation are estimated for an image fusion architecture for low-level image classification based on artificial neural networks. Experimental results are presented providing quantitative assessment of the proposed algorithms.

Multiple objective optimization for active sensor management

The performance of a multi-sensor data fusion system is inherently constrained by the configuration of the given sensor suite. Intelligent or adaptive control of sensor resources has been shown to offer improved fusion performance in many applications. Common approaches to sensor management select sensor observation tasks that are optimal in terms of a measure of information. However, optimising for information alone is inherently sub-optimal as it does not take account of any other system requirements such as stealth or sensor power conservation. We discuss the issues relating to developing a suite of performance metrics for optimising multi-sensor systems and propose some candidate metrics. In addition it may not always be necessary to maximize information gain, in some cases small increases in information gain may take place at the cost of large sensor resource requirements. Additionally, the problems of sensor tasking and placement are usually treated separately, leading to a lack of coherency between sensor management frameworks. We propose a novel approach based on a high level decentralized information-theoretic sensor management architecture that unifies the processes of sensor tasking and sensor placement into a single framework. Sensors are controlled using a minimax multiple objective optimisation approach in order to address probability of target detection, sensor power consumption, and sensor survivability whilst maintaining a target estimation covariance threshold. We demonstrate the potential of the approach through simulation of a multi-sensor, target tracking scenario and compare the results with a single objective information based approach.

Correction of misclassifications using a proximity-based estimation method

An estimation method for correcting misclassifications in signal and image processing is presented. The method is based on the use of context-based (temporal or spatial) information in a sliding-window fashion. The classes can be purely nominal, that is, an ordering of the classes is not required. The method employs nonlinear operations based on class proximities defined by a proximity matrix. Two case studies are presented. In the first, the proposed method is applied to one-dimensional signals for processing data that are obtained by a musical key-finding algorithm. In the second, the estimation method is applied to two-dimensional signals for correction of misclassifications in images. In the first case study, the proximity matrix employed by the estimation method follows directly from music perception studies, whereas in the second case study, the optimal proximity matrix is obtained with genetic algorithms as the learning rule in a training-based optimization framework. Simulation results are presented in both case studies and the degree of improvement in classification accuracy that is obtained by the proposed method is assessed statistically using Kappa analysis.

Correction of misclassifications in primary local image recognition using a nonlinear graph-based estimation technique

A description of an approach to primary local image recognition is given. The motivation for its application and its characteristics are discussed. Then a method for correction of misclassifications that occur in primary local image recognition is proposed. This method uses a graph-based estimation technique that uses information contained in supplementary classes in order to remove misclassifications and/or confirm the correct recognition of pixel hypotheses. In addition, the method is able to remove the supplementary classes after they are no longer needed. The particular features of the considered approach are that it is iterative and uses structures similar to those of center weighted median filters. The numerical simulation results are presented to illustrate the efficiency of the proposed technique.

Dichoptic motion within Panum's fusional area?

Purpose. Recent computational and neurophysiological findings suggest the existence of binocular complex cells that jointly encode motion and stereoscopic depth. Dichoptically presented gratings flickering in spatio-temporal quadrature can induce directional motion perception but the existence of an early binocular motion system is a debated issue. Here we investigated perception of dichoptic motion for stimuli defined on different depth planes. Methods. Stimuli were presented to the left and right eye on a calibrated CRT display with a refresh rate of 120 Hz using a split-screen Wheatstone configuration. On each trial Ss verged on a fixation-cross flanked by nonius lines before two vertical sine-wave gratings of 1.1 c/deg were presented in dichoptic view. The gratings were displayed in a circular aperture for 200 msec flickering in temporal quadrature. After each presentation Ss indicated whether direction of motion was left or right. In three conditions disparity of the stimulus was set to −8, 0, or +8 arc min whereas spatial phase offset between gratings varied between pi/2 and 3pi/2 in randomly intermixed trials. Spatial and temporal phase of the gratings were randomised across trials. Results. Although stimuli appeared on different depth planes psychometric functions shifted systematically with disparity. Discrimination of motion direction was best when phase offset together with horizontal disparity resulted in spatial quadrature. No reliable discrimination performance was obtained for stimuli with static pedestal, flicker frequencies beyond 4 Hz, and disparities outside Panums fusional area. Conclusions. The present result does not support the notion of an early binocular motion system that can integrate dichoptic motion at different depth planes.

Mitigation of sensor and communication system impairments for multichannel image fusion and classification

The impairments produced in image sensors and communication channels degrade image quality and introduce significant errors in the results of data fusion. To improve image fusion results in the presence of different types of impairments we propose a two-stage approach. New multistage nonlinear locally-adaptive image processing algorithms are developed and applied at the first stage to mitigate image impairments such as geometric distortions due to communication system synchronization errors, narrowband frequency interferences and sensor noise. Image fusion and classification algorithms based on artificial neural networks and support vector machines are used at the second stage. Experimental results are presented for real satellite remote sensing images and simulated data providing quantitative assessment of the proposed algorithms.