Controlled Computational Load Scheme for Information Processing in Closely Packed Object Environment.

The paper presents a simulation study on the performance of a target tracker using controlled selection paring for a track splitting filter algorithm. In a typical track splitting filter all the observation which fall inside a likelihood ellipse are used for update, however, in our proposed controlled selection paring procedure track splitting filter (TSF) algorithm uses maximum of two observations for track update. This kind of approach is particularly important if the computational and memory requirement is limited in comparison to the amount of input data (number of objects) that can saturate the system. Much of the previous performance work has been done on specific (deterministic) scenarios [1]. One of the reasons for considering the specific scenarios, which were normally crossing targets, was to test the efficiency of the track splitting algorithm for different situations. However, this approach only gives a measure of performance for a specific, possibly unrealistic, scenario and it was felt appropriate to develop procedures that would enable a more general performance assessment. Therefore, a random target motion scenario is adopted. Its implementation in particular for testing the proposed controlled track splitting algorithm using Kalman filters is investigated through a number of performance parameters which gives the activity profile of the tracking scenario. This kind of performance evaluation can provide a guide line for developing appropriate tracking systems of various categories. The complete proto type system is implemented using a TMS320C6416 DSP. [ABSTRACT FROM AUTHOR]