Target Recognition Based on Feature Weighted Intuitionistic FCM
To the issue of categorical attributes data of target recognition tending to false well-proportioned weight, a renewed technique for feature weighted intuitionistic fuzzy c means (FWIFCM) is presented, whose validity are checked by utilizing a practical e
- PDF / 428,388 Bytes
- 10 Pages / 439.37 x 666.14 pts Page_size
- 21 Downloads / 187 Views
Key Laboratory of CAPF for Cryptology and Information Security, Electronics Department, Engineering University of Armed Police Force, Xi’an 710086, People’s Republic of China [email protected] 2 Department of Information Engineering, Engineering University of Armed Police Force, Xi’an 710086, People’s Republic of China
Abstract. To the issue of categorical attributes data of target recognition tending to false well-proportioned weight, a renewed technique for feature weighted intuitionistic fuzzy c means (FWIFCM) is presented, whose validity are checked by utilizing a practical experiment for categorical attributes data. Finally, classifying function of additional feature weighted is analyzed and compared by providing an explicit experiment on 20 typical targets, and FWIFCM algorithm is well applied to typical target recognition on air. Simulation experiments prove that the technique proposed is promising and effective, while satisfactory results verify their applicability greatly. Keywords: Intuitionistic fuzzy sets · Fuzzy c-means clustering · Feature weighted · Target recognition
1
Introduction
Clustering analysis [1-3] is a general technique for diversified statistical analysis, which is also a significant branch of non-surveillance pattern classification in statistic pattern recognition. In nature, clustering can be used to diversified statistical analysis according to a natural rule “Like attracts like” classify data. There is a focus on the reasonable classification on the basis of data objections’ characteristic in clustering. In particular, the same class of data must have more similarity, and conversely the different classes of data must have more diversity. Clustering analysis originates in the many fields, commonly including mathematics, computer science, statistic, biology, economics and so forth. Furtherly, more and more clustering analysis algorithms[4,5] emerging currently have been made full use of in various application fields, such as voice recognition[6,7], image segmentation[8-10], data compression[11-13], network information security[14,15] and so on. In addition, it is such an important role on research about other fields such as biology [16], psychology, archeology, geology, social intercourse network [17] and marketing management. In the application of target recognition, target characteristic information getting from various kinds of sensors can effectively be fused and reasoning so that we can © Springer-Verlag Berlin Heidelberg 2015 H. Zha et al. (Eds.): CCCV 2015, Part I, CCIS 546, pp. 76–85, 2015. DOI: 10.1007/978-3-662-48558-3_8
Target Recognition Based on Feature Weighted Intuitionistic FCM
77
catch the accurate description of target attributes. In general, as for every dimension characteristic of the samples’ vectors in the traditional or classical algorithms, their construction degree to classification is well-proportioned. In contrast, owing to every dimension’ attributions consisting of vectors of samples’ features coming from different sensors, dimension, preciseness and r
Data Loading...
