Premium
Deterministic crowding introducing the distribution of population for template matching
Author(s) -
Sato Junya,
Akashi Takuya
Publication year - 2018
Publication title -
ieej transactions on electrical and electronic engineering
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.254
H-Index - 30
eISSN - 1931-4981
pISSN - 1931-4973
DOI - 10.1002/tee.22591
Subject(s) - matching (statistics) , transformation (genetics) , template matching , population , simple (philosophy) , computer science , crowding , rotation (mathematics) , algorithm , dimension (graph theory) , division (mathematics) , focus (optics) , mathematical optimization , artificial intelligence , mathematics , image (mathematics) , statistics , physics , neuroscience , biology , biochemistry , chemistry , demography , philosophy , arithmetic , epistemology , sociology , pure mathematics , optics , gene
This paper explains deterministic crowding (DC), introducing the distribution of population for template matching. We apply a simple genetic algorithm (GA) to template matching because this approach is effectively able to optimize geometric transformation parameters, such as parallel transformation, scaling, and in‐plane rotation. However, since the simple GA can obtain only one global optimum, detecting multiple objects is difficult. This is not of practical use. In order to detect multiple objects, we focus on DC, which is a multimodal optimization method and able to obtain multiple global and local solutions. In DC, there is a drawback where many individuals converge to one object and, hence, some objects cannot be detected. In order to solve this problem, the proposed method introduces the distribution of population. In experiments, the proposed method, DC, and crowding are applied to template matching and compared. The results confirm that the proposed method is better when an optimal threshold, which is used to create a cluster, is set. © 2017 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.