Search

A marginalized particle filtering (MPF) approach is proposed for target tracking under the background of passive measurement. Essentially, the MPF is a combination of particle filtering technique and Kalman filter. By making full use of marginalization, the distributions of the tractable linear part of the total state variables are updated analytically using Kalman filter, and only the lower-dimensional nonlinear state variable needs to be dealt with using particle filter. Simulation studies are performed on an illustrative example, and the results show that the MPF method leads to a significant reduction of the tracking errors when compared with the direct particle implementation. Real data test results also validate the effectiveness of the presented method.

The authors present an algorithm for point-target detection in infrared imagery based on the statistical threshold technique and continuousness of point-target motion trajectory for detecting a point-target from a sequence of images. The algorithm has been successfully tested on computer simulation images that include single or several point-targets with different SNR characteristics. >

The use of neural networks for radar ship target recognition is discussed. A practical approach to recognizing radar ship targets with similar video echoes while maintaining identification is proposed and simulated. Preprocessing is performed using Fourier-Mellin transform and coding transform, after which the frequencies of occurrence of the specific binary words, such as diagrams and trigrams, are chosen as features. A smaller neural network can be trained to identify the radar ship target by extracted features. >

This paper presents an algorithm based on TOA (time of arrival) and azimuth measurements. The location principle and feasibility of the technique are analysed in detail. For some bistatic radar systems with receivers which locate and track 3-D moving targets by means of a passive method, it is sometimes difficult to measure the pitch angle of targets. The algorithm proposed is of great importance to the practicability of bistatic passive location and tracking systems. With the help of computer simulation, the performance of this algorithm is evaluated for typical target paths according to Monte Carlo methods. >

Under the condition of jamming the number of usable measurements in a T/R-R bistatic system is variable. Therefore it is necessary to study some location and tracking algorithms concerning different sets of measurement data. The multi-stage tracking technique proposed makes full use of measurement data in T/R and R stations under different condition of jamming, as a result the tracking precision of 3-D moving targets is increased. The performance of these algorithms is evaluated with the help of computer simulation for typical target paths.

Multiple-model Kalman Filter Bank (KFB) is the conventional method in the area of maneuver target tracking. However, the KFB method is often faced with the dilemma between noise sensitivity and the accuracy of the temporal and spatial location of the maneuver. The difficulty, of dealing with this problem is due to the fact that the maneuver information of a target is often submerged in the measurement noise, especially when the measurements are acquired in low SNR case. With the wavelet packets analysis theory, the primitive measurements can be decomposed into two parts in a resolution level One is the "trend" of the measurements which is related to the true course of the target's moving track. The other is the "fluctuation" of the measurement which is related to the maneuver information of the target and the measurement noise. In the case of conventional KFB, measurements are processed in a single resolution level, thus, the noise components, the maneuver information components and the target location components of the measurements are weighted as the same to the KFB. With the wavelet packets theory, we can decompose the measurements into their trends and fluctuations in different resolution levels, thus KFB is applied respectively to the trends and the fluctuations So the measurement noise is easily attenuated with little or without attenuation of the location information anti maneuver information. Simulation tests and comparison between conventional KFB method and wavelet packets analysis method are also discussed in the paper to demonstrate practicality and advantage of the wavelet packets method.

A fusion algorithm is presented for asynchronous multi-sited-sensors target tracking based on extended pseudo measurement filter. Even in a jamming environment it can preserve the tracking of target. The difference between the algorithm and MGEKF is discussed. Afterwards, in conjunction with a simplified joint probabilistic data association, the tracking of two cross moving targets in clutter was implemented.

The paper emphatically studies the way for solving the problem of track confusion in multiple targets tracking by using message fusion technique. The discussion is on the basis of measurement data of practical bistatic systems. At first the paper introduces the association algorithm used for multiple targets tracking in bistatic systems and points out that track confusion easily happens if dynamic state parameters are the only ones used for association process. Then it gives several association methods fusing data of target state, Doppler frequency (F/sub ds/) and radar cross-section (RCS). After analyzing characteristics of F/sub ds/ and RCS data, the paper suggests two fusion modes-cascade fusion and parallel fusion. Both modes can decide whether or not to use F/sub ds/ data according to its change. From one hundred statistical results it is clear that track confusion frequency drops greatly when using data fusion techniques.

The paper presents a 3-D location and tracking algorithm based on the emitter's distance difference between two measurements and azimuth message. The location principle and feasibility of the method are analysed in detail. Under the condition that an emitter moves linearly at a constant speed, if we can locate and track an emitter only by using azimuth and distance difference without any angle of elevation, the equipment in the receiver will be simplified and the independence of the receiver may be improved as well. Then the algorithm proposed in the paper is of great importance to the practicability of the single observer passive location and tracking system. Performance of this algorithm is evaluated with the help of computer simulation test for three typical target paths. >

An attempt is made to develop a high-speed detection scheme for moving and stationary point targets registered in an IR (infrared) image sequence. An iterative approximate 3-D line searching algorithm (for non-manoeuvering targets in space) based upon the geometric representation of lines is derived. The convergence of the algorithm is proved. An analysis of the algorithm yields the theoretical detection performance. Statistical experiment results show the effectiveness and the computational efficiency of the algorithm in the case of low signal-to-noise ratio. The algorithm may be employed to satisfy the real-time processing requirement of an IR homing. >

The authors present the detailed derivation of the WMEKF (weighted modified extended Kalman filter) algorithm for a bistatic system, in which the passive observer measures the azimuth, elevation, and time difference. A unique multistatic system composed of netted-bistatic systems is proposed. The processed results at each bistatic system are sent over a communication network to a central processor and processed with the WLS (weighted least sequence) algorithm. The tracking performance of the multistatic system is evaluated by means of the Monte Carlo simulation technique. >

A method for point target recognition is presented which is based on fusing information from thermal and visual images. Usually targets and decoys are made of different materials, and there are inside heat sources in targets. Thus, their emission, absorption and reflection characteristics in infrared band and in visual band are different. The differences between targets and decoys in the two mentioned bands are used to distinguish targets from decoys. By using information from thermal and visual images, one can acquire information which can not be obtained by using thermal images alone or visual images alone. >

In a frequency-agile magnetron radar, the cavity frequency tracking system of the tuning magnetron is used to control the transmitted pulse sequences and carry out the auto-track of the local oscillator. An experimental scheme of a small steady-state error tracking system is proposed, and the principle of the tracking system is described. The time-domain and frequency-domain performance of the tracking system is discussed. Experimental results indicate that this system has a high tracking accuracy. >

In this paper, the problem of noisy image moving point target recognition is discussed. A recognition method is proposed. This method is composed of the three processes in cascade which are: process A, preprocessing of the noisy image moving point target sequence; process B, choice and extraction of features; and process C, classification of the extracted features by using a second-order neural network, respectively. A simulation experiment is carried out for three kinds of image moving point targets in which the signal to noise level ratio (SNR) is specified. The experimental results indicate that the recognition system with the proposed method has both high and reliable identification rate. >

Tracking a maneuvering target in a nonlinear interference environment has been discussed in some literature, and this paper considers the problems of tracking in the non-Gaussian noise environment and then tracking with multiple passive sensors. The method of target state estimation in this paper is a dynamic programming approach. Unlike the traditional method, the estimating problem is reduced to a multiple hypothesis-testing problem, and then is solved by using the dynamic programming algorithm. The simulation results show the superiority of the new method to EKF.

The accurate debiasing processing for the bias in the classical conversion mainly depends on the first and second statistics of the cosine of the bearing measurement errors. An unbiased conversion is presented in this paper. Some comparison between the new unbiased conversion with the previously offered debiased conversion is made.

One of the most important problem must be resolved in multitarget tracking with multisensor is the data association. It includes two folds: associating each observation of every sensor with each target, named scan association; associating-the track with track formed by different sensors, track association. There exist some difficulties in obtaining the models and in computing burden while tracking multitarget in a high density clutter environment. This paper presents an efficient method for tracking multitarget with multisensor, and the new method applied integer programming to associate the incoming measurements to established tracks. This paper assumes no new target appears and no target eliminates while the tracking goes on. Theory analysis and Monte Carlo simulations show the potential of this algorithm.

The improved and generalized modified gain functions for target tracking are presented. Simulation results are presented which demonstrate the superiority of the new modified gain functions.

Coordinate transformation is always concerned with data processing of bistatic/multistatic systems. The influence of coordinate transformation upon measurement errors of the bistatic system is discussed in the paper. Detail derivations of the relation equation between the original coordinate system and the transformed new one, quantitative analysis of the influence and a number of curves are also given in the paper.

Conventional target association and tracking techniques such as PDA and JPDA have very fine performance when the measurement acquired from sensors are perfect. However, when jamming and stealth techniques are widely used, it is very difficult for sensors to gain perfect measurements. Though a single sensor in a distributed sensors system might fail to acquire continually perfect measurements of low-RCS (stealth) targets under jamming environment, the distributed sensors system might gain relatively perfect measurements by integrating measurement hits or fractional trajectories of targets from every sensor in the system. The paper discusses association and tracking techniques under the circumstances mentioned above. Special attention is paid to track initiation and data association. Monte Carlo simulation test results are also presented in the paper.

When the communication link of a T/R-R bistatic radar system is affected by external interference, the number of usable measurements in it is variable. Therefore it is necessary to study some tracking algorithms concerning with different sets of measurement data. The multi-stage tracking technique proposed in the paper makes full use of measurement data in T/R and R station under different condition, as a result the tracking precision of 3-D moving targets is increased. The performance of these algorithms is evaluated with the help of computer simulation for typical target paths.

This paper presents an algorithm for multiple targets tracking. The bistatic radar system proposed in the paper contains a T/R station and a R station. Generally the T/R station provides data such as azimuth angle and distance of a target while the R station has measurement of azimuth angle and distance sum of the target. After coordinate conversion and target locating the measurement data is converted into 3-D coordinate position of the target. In order to track multi-targets it is important to select data association method. In the paper according to conditions of real-time system we accomplish multiple targets tracking in two stages. In the first stage the locating data of a target is used to do association with every target data stored in last tracking cycle. It is necessary to finish this stage as soon as we get the target's locating data due to the demand of real-time system. However association error sometimes happens in stage one. Therefore we decide to test the association result of stage one. When R station passes the observation field in every cycle, the second tracking stage begins so as to correct the association result of stage one. It has proved that the performance of T/R-R system m multiple targets tracking is greatly improved because of the new tracking algorithm discussed in this paper.

Automatic recognition of point targets in the same orbit is presented. It is impossible to recognize point target using shape characteristics. Some studies of point target recognition are based on the difference of motion characteristics. However, the targets discussed in this paper move accompanied by each other in the same space orbit and therefore have almost the same motion characteristics. Some new characteristics for the recognition are expected. Computation methods of the radiation in visible band and IR band are considered at first. After that, the radiation difference between the targets, with the same contour but different thickness of the shell, are discussed in detail. It is found that the IR differences are very useful for the recognition of targets with the same or similar radiation in visible band. On the other hand, the visible radiation feature can be used for the recognition of the targets with the same or similar IR irradiation.