A parameter estimation method of the simple PCNN model for infrared human segmentation. (February 2019)

• Record Type: Journal Article
• Authors: He, Fuliang; Guo, Yongcai; Gao, Chao
• Is Part Of: Optics & laser technology. Volume 110(2019)
• Page Start: 114
• Page End: 119
• Publication Date: 2019-02
• Subjects: PCNN -- Cuckoo search -- Kent chaotic map -- KD-Tree -- Human segmentation; Optics -- Periodicals; Lasers -- Periodicals; Electronic journals; 621.366
• DOI: 10.1016/j.optlastec.2018.05.042
• Languages: English
• ISSNs: 0030-3992

Abstract

Highlights: The appropriate model for the infrared human segmentation. An efficient and effective multi-parameter estimation method of PCNN model. Strengthening the local search ability by Kent chaotic map. Overpassing the search blindness by k dimensional-tree. Abstract: In order to address the parameter estimation problem of pulse coupled neural network (PCNN) in the infrared human segmentation, this paper proposes an improved search method of optimal parameters. First, based on a simple PCNN model, the cuckoo search (CS) algorithm, instead of the existing swarm intelligent methods, is introduced to deal with this multi-parameter optimization problem. Next, the Kent chaotic map is applied to the local search process of CS algorithm, so as to enhance search precision. Furthermore, to cope with the search blindness, a guidance mechanism by the k dimensional-tree (KD-Tree) is presented to accelerate the convergence speed of parameter search based on the current search history. The proposed method is applied to a set of benchmark test functions and a series of infrared human segmentation experiments, and the experimental results demonstrate that the simple PCNN Model with this proposed parameter optimization method is superior to other PCNN segmentation models with existing swarm intelligent optimization algorithms both in the visual segmentation effect and the objective quantitative evaluations of accuracy and computing speed.