The tabu search particle filter is proposed in this research based on the integration of the modified tabu search metaheuristic optimization and the genetic particle filter. Experiments with this algorithm in real time human tracking applications in unconstrained environments show that it is more robust, accurate and faster than a number of other existing metaheuristic filters, including the evolution particle filter, particle swarm filter, simulated annealing filter, path relink filter and scatter search filter. Quantitative evaluation illustrates that even with only ten particles in the system, the proposed tabu search particle filter has a success rate of 93.85% whereas the success rate of other metaheuristic filters ranged from 68.46% to 17.69% under the same conditions. The accuracy of the proposed algorithm (with ten particles in the tracking system) is 2.69 pixels on average, which is over 3.85 times better than the second best metaheuristic filters in accuracy and 18.13 times better than the average accuracy of all other filters. The proposed algorithm is also the fastest among all metaheuristic filters that have been tested. It achieves approximately 50 frames per second, which is 1.5 times faster than the second fastest algorithm and nineteen times faster than the average speed of all other metaheuristic filters.

Furthermore, a unique colour sequence model is developed in this research based on a degenerated form of the hidden Markov model. Quantitative evaluations based on rigid object matching experiments illustrate that the successful matching rate is 5.73 times better than the widely used colour histogram. In terms of speed, the proposed algorithm achieves twice the successful matching rate in about three quarters of the processing time consumed by the colour histogram model.

Overall, these results suggest that the two proposed algorithms would be useful in many applications due to their efficiently, accuracy and ability to robustly track people and coloured objects.