Huadong Meng | Tsinghua University (original) (raw)
Papers by Huadong Meng
Computing Research Repository, 2010
Space-time adaptive processing (STAP) is an effective tool for detecting a moving target in space... more Space-time adaptive processing (STAP) is an effective tool for detecting a moving target in spaceborne or airborne radar systems. Statistical-based STAP methods generally need sufficient statistically independent and identically distributed (IID) training data to estimate the clutter characteristics. However, most actual clutter scenarios appear only locally stationary and lack sufficient IID training data. In this paper, by exploiting the intrinsic
2014 IEEE 80th Vehicular Technology Conference (VTC2014-Fall), 2014
IEEE 2002 International Conference on Communications, Circuits and Systems and West Sino Expositions, 2002
A method for evaluating the robustness of constant false alarm rate (CFAR) detectors is presented... more A method for evaluating the robustness of constant false alarm rate (CFAR) detectors is presented, which is based on the powerful methodology of influence function (IF) developed in the literature on robust statistics. It can evaluate and compare the robustness of different kinds of CFAR detectors by calculating the first derivative of false alarm probability (FAP) at an underlying distribution, which is named FAP-IF. The FAP-IF of cell averaging (CA) and ordered statistics (OS) CFAR detectors are presented. The computational simulation of radar detection in exponential clutter is also presented to prove the theoretical result.
2003 IEEE International Conference on Acoustics, Speech, and Signal Processing, 2003. Proceedings. (ICASSP '03)., 2003
The traditional CFAR processors are based on the sliding-window concept, which have substantial p... more The traditional CFAR processors are based on the sliding-window concept, which have substantial performance degradation under nonhomogeneity. Owing to temporal processing and the exploitation of the local homogeneity of the map cell, the clutter-map procedure acquires enhanced robustness with little CFAR losses. In this paper, a Gaussian biparametric clutter-map constant false alarm rate (GBCM-CFAR) processor is proposed which merges the
International Journal of Modern Physics C, 2011
Science China Information Sciences, 2012
2010 International Waveform Diversity and Design Conference, 2010
2001 CIE International Conference on Radar Proceedings (Cat No.01TH8559), 2001
Two methods for estimating the center frequency and bandwidth, separately called centroid method ... more Two methods for estimating the center frequency and bandwidth, separately called centroid method (CM) and energy integration method (EIM), are proposed. A comparison of the complexity and the performance between the two methods and the general ones is presented. It is shown that the two methods are effective for short-time clutter data in practical application
2007 IEEE Intelligent Transportation Systems Conference, 2007
2010 2nd International Conference on Signal Processing Systems, 2010
Compressed sensing is currently one of the most popular fields in signal processing, which enable... more Compressed sensing is currently one of the most popular fields in signal processing, which enables us to acquire “sparse” signals by sampling under Nyquist's frequency. In this paper, a novel hardware architecture based on Optimized CoSaMP algorithm is presented for sparse recovery of compressed sensing. Due to the crucial calculation of least squares in the algorithm, the proposed architecture develops a least squares module (LS module) with linear array structure, which achieves good efficiency as well as flexibility and scalability. The running results obtained from the implementation on a field-programmable gate array (FPGA) show that our design possesses excellent performance and high accuracy.
2010 International Conference on Electrical and Control Engineering, 2010
2011 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2011
2007 IEEE Intelligent Transportation Systems Conference, 2007
This paper presents a novel method for cast vehicle shadow segmentation in a sequence of traffic ... more This paper presents a novel method for cast vehicle shadow segmentation in a sequence of traffic images taken from a stationary camera mounted on a tall building or pedestrian crossings bridge. Based on a simplified vehicle/shadow joint contour model, which is represented by a polygon, we propose a two-step method: first, estimate the environmental parameters of current traffic scene through
2007 IEEE Intelligent Transportation Systems Conference, 2007
2011 IEEE RadarCon (RADAR), 2011
ABSTRACT The probability hypothesis density (PHD) filter is a practical alternative to the theore... more ABSTRACT The probability hypothesis density (PHD) filter is a practical alternative to the theoretically optimal multi-target Bayesian filter based on random finite sets (RFS) for multi- target tracking. In this paper, we propose Track PHD (TPHD) filter based on a track state space consisted of target position history and it propagates the multi-target intensity function of track RFS. The new filter provides the estimates of target track states and makes it easy to confirm identities. Simulation results demonstrate TPHD filter is effective in estimating multi-target states and providing target identities even when targets are in close proximity. Multi-target Bayesian filter based on random finite sets (RFS) is a theoretically optimal approach to multi-target tracking (MTT) involving the joint estimation of a time- varying number of targets and their states from a sequence of noisy measurements in the presence of data association uncertainty, clutter and miss detection (1). The probability hypothesis density (PHD) filter (2) is a suboptimal and approximate approach to RFS multi-target Bayesian filter by propagating the first order moment of state RFS over time. The peaks of the PHD function give the identity-free estimates of target states. The Gaussian-Mixture PHD (GMPHD) tracker (3) and Data association schemes for particle-PHD filter (4) have been proposed to implement temporal association of state estimates over time. For PHD target state consists of position and its high-order derivatives such as speed and acceleration. The practical transition of these high-order parameters is difficult to be well described in Markov time-varying model and estimates for them are generally unstable. So, based on such states it's not easy to correctly provide identities of close targets for PHD filter (3). Without considering the unstable high-order parameters of position in traditional state space, we propose Track PHD (TPHD) filter based on a track state space consisted of target position history. The new filter propagates the posterior intensity function or the first order moment of track RFS. There are three advantages for TPHD filter. Firstly, it's easy for TPHD filter to confirm temporal association of state estimates over time and provide the identities of individual targets, because the track states consists of the history trajectories. Secondly, track state can be utilized to more accurately predict the motion tendency, which will be beneficial for improvement of tracking accuracy. Besides, close targets are easily distinguished from each other for TPHD recursion in the track state space. For illustration purposes, we demonstrate the ability of TPHD filter to estimate multi-target positions and correctly provide identities. The performance of TPHD tracker is benchmarked against that of GMPHD tracker.
IEEE 10th INTERNATIONAL CONFERENCE ON SIGNAL PROCESSING PROCEEDINGS, 2010
2006 IEEE Intelligent Transportation Systems Conference, 2006
In most present frontal collision warning systems (FCWS), the warning algorithm mainly depends on... more In most present frontal collision warning systems (FCWS), the warning algorithm mainly depends on simple combination of linearly predicted vehicle-motion parameters. Such systems suffer from high false-alarm-rate due to the incapability of identifying different transportation scenarios. Scenario parsing deals with such problems by analyzing transportation scenarios and applying specific threat assessment algorithm to each scenario. In this paper, approaches of
2008 9th International Conference on Signal Processing, 2008
Computing Research Repository, 2010
Space-time adaptive processing (STAP) is an effective tool for detecting a moving target in space... more Space-time adaptive processing (STAP) is an effective tool for detecting a moving target in spaceborne or airborne radar systems. Statistical-based STAP methods generally need sufficient statistically independent and identically distributed (IID) training data to estimate the clutter characteristics. However, most actual clutter scenarios appear only locally stationary and lack sufficient IID training data. In this paper, by exploiting the intrinsic
2014 IEEE 80th Vehicular Technology Conference (VTC2014-Fall), 2014
IEEE 2002 International Conference on Communications, Circuits and Systems and West Sino Expositions, 2002
A method for evaluating the robustness of constant false alarm rate (CFAR) detectors is presented... more A method for evaluating the robustness of constant false alarm rate (CFAR) detectors is presented, which is based on the powerful methodology of influence function (IF) developed in the literature on robust statistics. It can evaluate and compare the robustness of different kinds of CFAR detectors by calculating the first derivative of false alarm probability (FAP) at an underlying distribution, which is named FAP-IF. The FAP-IF of cell averaging (CA) and ordered statistics (OS) CFAR detectors are presented. The computational simulation of radar detection in exponential clutter is also presented to prove the theoretical result.
2003 IEEE International Conference on Acoustics, Speech, and Signal Processing, 2003. Proceedings. (ICASSP '03)., 2003
The traditional CFAR processors are based on the sliding-window concept, which have substantial p... more The traditional CFAR processors are based on the sliding-window concept, which have substantial performance degradation under nonhomogeneity. Owing to temporal processing and the exploitation of the local homogeneity of the map cell, the clutter-map procedure acquires enhanced robustness with little CFAR losses. In this paper, a Gaussian biparametric clutter-map constant false alarm rate (GBCM-CFAR) processor is proposed which merges the
International Journal of Modern Physics C, 2011
Science China Information Sciences, 2012
2010 International Waveform Diversity and Design Conference, 2010
2001 CIE International Conference on Radar Proceedings (Cat No.01TH8559), 2001
Two methods for estimating the center frequency and bandwidth, separately called centroid method ... more Two methods for estimating the center frequency and bandwidth, separately called centroid method (CM) and energy integration method (EIM), are proposed. A comparison of the complexity and the performance between the two methods and the general ones is presented. It is shown that the two methods are effective for short-time clutter data in practical application
2007 IEEE Intelligent Transportation Systems Conference, 2007
2010 2nd International Conference on Signal Processing Systems, 2010
Compressed sensing is currently one of the most popular fields in signal processing, which enable... more Compressed sensing is currently one of the most popular fields in signal processing, which enables us to acquire “sparse” signals by sampling under Nyquist's frequency. In this paper, a novel hardware architecture based on Optimized CoSaMP algorithm is presented for sparse recovery of compressed sensing. Due to the crucial calculation of least squares in the algorithm, the proposed architecture develops a least squares module (LS module) with linear array structure, which achieves good efficiency as well as flexibility and scalability. The running results obtained from the implementation on a field-programmable gate array (FPGA) show that our design possesses excellent performance and high accuracy.
2010 International Conference on Electrical and Control Engineering, 2010
2011 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2011
2007 IEEE Intelligent Transportation Systems Conference, 2007
This paper presents a novel method for cast vehicle shadow segmentation in a sequence of traffic ... more This paper presents a novel method for cast vehicle shadow segmentation in a sequence of traffic images taken from a stationary camera mounted on a tall building or pedestrian crossings bridge. Based on a simplified vehicle/shadow joint contour model, which is represented by a polygon, we propose a two-step method: first, estimate the environmental parameters of current traffic scene through
2007 IEEE Intelligent Transportation Systems Conference, 2007
2011 IEEE RadarCon (RADAR), 2011
ABSTRACT The probability hypothesis density (PHD) filter is a practical alternative to the theore... more ABSTRACT The probability hypothesis density (PHD) filter is a practical alternative to the theoretically optimal multi-target Bayesian filter based on random finite sets (RFS) for multi- target tracking. In this paper, we propose Track PHD (TPHD) filter based on a track state space consisted of target position history and it propagates the multi-target intensity function of track RFS. The new filter provides the estimates of target track states and makes it easy to confirm identities. Simulation results demonstrate TPHD filter is effective in estimating multi-target states and providing target identities even when targets are in close proximity. Multi-target Bayesian filter based on random finite sets (RFS) is a theoretically optimal approach to multi-target tracking (MTT) involving the joint estimation of a time- varying number of targets and their states from a sequence of noisy measurements in the presence of data association uncertainty, clutter and miss detection (1). The probability hypothesis density (PHD) filter (2) is a suboptimal and approximate approach to RFS multi-target Bayesian filter by propagating the first order moment of state RFS over time. The peaks of the PHD function give the identity-free estimates of target states. The Gaussian-Mixture PHD (GMPHD) tracker (3) and Data association schemes for particle-PHD filter (4) have been proposed to implement temporal association of state estimates over time. For PHD target state consists of position and its high-order derivatives such as speed and acceleration. The practical transition of these high-order parameters is difficult to be well described in Markov time-varying model and estimates for them are generally unstable. So, based on such states it's not easy to correctly provide identities of close targets for PHD filter (3). Without considering the unstable high-order parameters of position in traditional state space, we propose Track PHD (TPHD) filter based on a track state space consisted of target position history. The new filter propagates the posterior intensity function or the first order moment of track RFS. There are three advantages for TPHD filter. Firstly, it's easy for TPHD filter to confirm temporal association of state estimates over time and provide the identities of individual targets, because the track states consists of the history trajectories. Secondly, track state can be utilized to more accurately predict the motion tendency, which will be beneficial for improvement of tracking accuracy. Besides, close targets are easily distinguished from each other for TPHD recursion in the track state space. For illustration purposes, we demonstrate the ability of TPHD filter to estimate multi-target positions and correctly provide identities. The performance of TPHD tracker is benchmarked against that of GMPHD tracker.
IEEE 10th INTERNATIONAL CONFERENCE ON SIGNAL PROCESSING PROCEEDINGS, 2010
2006 IEEE Intelligent Transportation Systems Conference, 2006
In most present frontal collision warning systems (FCWS), the warning algorithm mainly depends on... more In most present frontal collision warning systems (FCWS), the warning algorithm mainly depends on simple combination of linearly predicted vehicle-motion parameters. Such systems suffer from high false-alarm-rate due to the incapability of identifying different transportation scenarios. Scenario parsing deals with such problems by analyzing transportation scenarios and applying specific threat assessment algorithm to each scenario. In this paper, approaches of
2008 9th International Conference on Signal Processing, 2008