Search

This work is concerned with the problem of multiple target track detection in heavy clutter. Using the 'modified high order correlation' (HOC) [1-4] process and a track scoring mechanism a new method is developed to perform data association and track identification in the presence of heavy clutter. Using this new scheme any number of very close, crossing or splitting target tracks can be resolved without increasing the computational complexity of the algorithm. The applicability of the method for continuous detection of target tracks that can originate and terminate at any scan is also demonstrated. In addition, the operating characteristics as a function of the clutter density are also provided. Simulation results on all the cases are presented.

This paper presents a new method for clutter rejection and dim target track detection from infrared (IR) satellite data using neural networks. A high order correlation method is developed which recursively computes the spatio-temporal cross-correlations between data of several consecutive scans. The implementation of this scheme using a connectionist network is also presented. Several important properties of the high order correlation method are established which indicate that the resultant filtered images capture all the target information. The simulation results using this approach show at least 93% clutter rejection. Further improvement in the clutter rejection rate is achieved by modifying the high order correlation method to incorporate the target motion dynamics. The implementation of this modified high order correlation using a high order neural network architecture is demonstrated. The simulation results indicate at least 97% clutter rejection rate for this method. A comparison is also made between the methods developed here and the conventional frequency domain three-dimensional (3-D) filtering scheme, and the simulation results are provided.

A fast algorithm for the learning process of multilayer neural networks is presented using the recursive least squares method. This algorithm minimizes the global sum of the squared errors between the actual and the desired output values iteratively. Lastly, this method is compared with the standard back-propagation and the accelerated back-propagation learning algorithms and results show that the proposed method has fewer iterations and uses less computational time.