Huaitie Xiao

Active Learning With Gaussian Process Classifier for Hyperspectral Image Classification

Gaussian process (GP) classifiers represent a powerful and interesting theoretical framework for the Bayesian classification of hyperspectral images. However, the collection of labeled samples is time consuming and costly for hyperspectral data, and the training samples available are often not enough for an adequate learning of the GP classifier. Moreover, the computational cost of performing inference using GP classifiers scales cubically with the size of the training set. To address the limitations of GP classifiers for hyperspectral image classification, reducing the label cost and keeping the training set in a moderate size, this paper introduces an active learning (AL) strategy to collect the most informative training samples for manual labeling. First, we propose three new AL heuristics based on the probabilistic output of GP classifiers aimed at actively selecting the most uncertain and confusing candidate samples from the unlabeled data. Moreover, we develop an incremental model updating scheme to avoid the repeated training of the GP classifiers during the AL process. The proposed approaches are tested on the classification of two realworld hyperspectral data. Comparison with random sampling method reveals a better accuracy gain and faster convergence with the number of queries, and comparison with recent active learning approaches shows a competitive performance. Experimental results also verified the efficiency of the incremental model updating scheme.

A new multiple extended target tracking algorithm using PHD filter

Abstract A new multiple extended target tracking algorithm using the probability hypothesis density (PHD) filter is proposed in our study, to solve problems on tracking performance degradation of the extended target PHD (ET-PHD) filter under the nonlinear conditions and its intolerable computational requirement. It is noted that with the current Gaussian mixture implement of ET-PHD filter satisfying tracking performance could only be obtained under linear and Gaussian conditions. To extend the application of ET-PHD filter for nonlinear models, our study has derived a particle implement of ET-PHD (ET-P-PHD) filter. Our study finds that the main factors influencing the computational complexity of the ET-P-PHD filter are the partition number of measurement set and the calculation of non-negative coefficients of cells in partitions. With the pretreatment of measurements and application of a new K -means clustering based measurement set partition method, we have successfully decreased the partition number. In addition, a gating method for target state space, which is based on likelihood relationship between target state and measurement, is proposed to simplify the calculation of non-negative coefficients. Simulation results show that the algorithms proposed by our study could satisfyingly deal with multiple extended target tracking issues under nonlinear conditions, and lead to significantly lower computational complexity with tiny effect on tracking performance.

An MRF Model-Based Active Learning Framework for the Spectral-Spatial Classification of Hyperspectral Imagery

Hyperspectral image classification has attracted extensive research efforts in the recent decades. The main difficulty lies in the few labeled samples versus high dimensional features. The spectral-spatial classification method using Markov random field (MRF) has been shown to perform well in improving the classification performance. Moreover, active learning (AL), which iteratively selects the most informative unlabeled samples and enlarges the training set, has been widely studied and proven useful in remotely sensed data. In this paper, we focus on the combination of MRF and AL in the classification of hyperspectral images, and a new MRF model-based AL (MRF-AL) framework is proposed. In the proposed framework, the unlabeled samples whose predicted results vary before and after the MRF processing step is considered as uncertain. In this way, subset is firstly extracted from the entire unlabeled set, and AL process is then performed on the samples in the subset. Moreover, hybrid AL methods which combine the MRF-AL framework with either the passive random selection method or the existing AL methods are investigated. To evaluate and compare the proposed AL approaches with other state-of-the-art techniques, experiments were conducted on two hyperspectral data sets. Results demonstrated the effectiveness of the hybrid AL methods, as well as the advantage of the proposed MRF-AL framework.

Analysis of Quantum Radar Cross Section and Its Influence on Target Detection Performance

Quantum radar is a new detection technology based on the mechanism of quantum physics and is promising for enhancing radar target detection capability. In this letter, the quantum radar equation is established based on the traditional radar equation; meanwhile, how the quantum radar cross section (QRCS) influences target detection performance and how QRCSs for some typical targets distribute are investigated via simulations. Simulation results demonstrate that the signal-to-noise ratio increases with the QRCS, the peak values of side-lobes of the QRCS for the cylinder surface fluctuate little, and the QRCS for corner reflectors almost exhibits no variations in a wide range of incident angles. The results listed above can benefit the signal design and performance evaluation of quantum radar as well as the design of quantum stealthy targets.

Spatial contextual classification of remote sensing images using a Gaussian process

ABSTRACT Gaussian process (GP) classifiers represent a powerful and interesting theoretical framework for the Bayesian classification of remote sensing images. However, the integration of spatial information in GP classifier is still an open question, while researches have demonstrated that the classification results could be improved when the spatial information is used. In this context, in order to improve the performance of the traditional GP classifier, we propose to use Markov random fields (MRFs) to refine the classification results with the neighbourhood information in the images. In the proposed method (denoted as GP-MRF), the MRF model is used as a post-processing step to the pixelwise results with GP classifier which classifies each pixel in the image separately. Therefore, the proposed GP-MRF approach promotes solutions in which adjacent pixels are likely to belong to the same class. Experimental results show that the GP-MRF could achieve better classification accuracy compared to the original GP classifier and the state-of-the-art spatial contextual classification methods.

Joint multi-target filtering and track maintenance using improved labeled particle PHD filter

As is known, the most prominent advantage of Finite Set Statistics (FISST) based multi-target tracking algorithms is it could cope with complicated tracking problems arising from special events such as target birth, target death and tracks crossing without complicated data association. Through improving the existing labeled particle Probability Hypothesis Density (L-P-PHD) filter, an improved labeled particle PHD (IL-P-PHD) filter is proposed in this paper. Simulation experiment shows that the tracking performance of IL-P-PHD filter is much better than L-P-PHD filter on complicated multi-target tracking problems, IL-P-PHD filter could extract target track information while efficiently detecting target birth and disappearance and stably estimating target state.

A Perturbation-Based Approach for Compressed Sensing Radar Imaging

This letter is concerned with two-dimensional radar imaging via compressed sensing. First, a perturbation-based sparse representation dictionary is established to alleviate the basis mismatch effect, reduce the memory requirement, and accelerate the imaging process. Then, an iterative algorithm based on variational Bayesian inference is developed for sparse recovery that is user-parameter-free and suitable for imposing a priori constraint. Experimental results based on both synthetic and real data verify the performance of the proposed approach.

Imaging targets moving in formation using parametric compensation

When conventional motion compensation algorithms that are fit for a single target are applied to cooperative targets imaging, a well-focused image cannot be obtained due to the low correlation between adjacent returned signals. In this paper, a parametric compensation method is proposed for the imaging of cooperative targets. First, the problem of the imaging is formulated by analyzing the translational motion of the target moving along a rectilinear fight path and by assuming a signal model of cooperative targets imaging. A bulk image is then obtained by parametric compensation of the linear and quadratic phase terms, which is performed by means of estimating the translational motion parameters through the fractional Fourier transform. Next, the number of targets in the bulk image is estimated by clustering number estimation, and the segmented images from the bulk image are separated by the normalized cuts. Finally, well-focused images are obtained by refined parametric compensation of the residual quadratic and cubic phase terms, which is carried out by estimating the parameters through maximizing the image contrast. Simulation results demonstrate the effectiveness of the proposed method.

Joint tracking and identification of the unresolved towed decoy and aircraft using the labeled particle probability hypothesis density filter

For the distance and velocity deception from the new towed decoy, echo signal from the target and decoy appear as one target on time domain and frequency domain because of aliasing. Therefore, independent measurements for the decoy and aircraft are unavailable for conventional algorithm, neither identification and tracking. In this paper is proposed a new algorithm for joint identification and tracking of the decoy and aircraft which are unresolved within radar beam, innovations for which include: First, construction of echo signal model in the three interfering stages. Once stage decided with the decoy presence detection algorithm, the proposed particle filter based measurement generating algorithm sequentially estimates the aircraft and decoy character parameters on different stages, separating the aircraft signal and decoy signal. Secondly, based on the improved labeled particle probability hypothesis density (IL-P-PHD) filter, an algorithm for joint identification and tracking of vertically the unresolved aircraft and decoy is proposed, realizing real time identification and sequential estimation of movement state. Simulation experiment demonstrates that the proposed algorithm behaves in a manner consistent with our expectations.

A supervised method for nonlinear dimensionality reduction with GPLVM

Dimensionality reduction is a very important task in many artificial applications. Among the current researches, Gaussian process latent variable model (GPLVM), which is a nonlinear dimensionality reduction method, has become a hot topic in recent years. In this paper, a supervised version of GPLVM with the prior characterized by the determinantal clustering process (DCP) prior is proposed. The combination of GPLVM with the DCP prior over the low dimensional positions has the ability of preservation the discriminative property between different classes. Experiments were conducted on two data sets, and the results demonstrated the better performance of the proposed method with respect to the original GPLVM.