A hybrid approach of gravitational search algorithm and ant miner plus for fingerprint recognition

The gravitational search algorithm (GSA) is an eminent heuristic algorithm inspired by the laws of gravity and motion. It possesses an independent physical model in which the mass agents are guided by gravitational force to quickly achieve the convergence. Although the GSA is proven to be efficient for science and engineering problems, the mass agents can be trapped in premature convergence due to the heaviness of masses in the later iterations. The occurrence of premature convergence impedes the agents' further exploration of the search space for a better solution. Here, the ant miner plus (AMP) variant of the ant colony optimization (ACO) algorithm is utilized to avoid the trapping of agents in local optima. The AMP algorithm extends the exploration ability of the GSA algorithm by using the attributes of pheromone updating rules generated by best ants and a problem-dependent heuristic function. The AMP variant adheres to the attributes of the ACO algorithm and is also a decision-making variant which determines the problem solution more efficiently by constructing a directed acyclic graph, considering class-specific heuristic values, and including weight parameters for the pheromone and heuristic values. In this research, this hybridization of GSA and AMP (GSAMP) algorithms is presented, and it is utilized for the decision-making application of fingerprint recognition. Here, fingerprint recognition is conducted for complete as well as latent fingerprints, which are poor quality partial fingerprints, mostly acquired from crime scenes by law enforcement agencies. The experiments are performed for the complete fingerprint dataset of FVC2004 and the latent fingerprint dataset of NIST SD27, using the proposed GSAMP approach and the individual algorithms of Ant Miner (AM) and AMP. The experimental evaluation indicates the superiority of the proposed approach compared to other methods.