NSGA DBSCAN: An Efficient Clustering Technique

Abstract

Clustering is one of the significant streams useful for determining groups and identifying significant distributions in the underlying data. Remote sensing images are utilized to automatically detect the high-resolution images. However, it demands accurate descriptions of the characteristics of the objects. Density-based spatial clustering of applications with noise (DBSCAN) evaluates clusters of arbitrary shape relying on a density-based notion of clusters. Additional implementation contains KD-Trees to save the data that allow efficient retrieval of data and bring down the time complexity from (n2) to (n log n). Therefore, improving the computational speed of the DBSCAN is the main motivation behind this research work. Because majority of existing clustering techniques used for remote sensing images suffer from noise issue and parameter tuning issue, which may degrade the performance of remote sensing vision systems. Therefore, to overcome this issue in this research work, a novel adaptive density-based clustering (NSGA DBSCAN) technique with noise is designed which is used to tune the parameters of DBSCAN based clustering technique for remote sensing images. Initially, random population is generated. Thereafter, for each random solution DBSCAN is implemented for clustering process. The solution that has accurate cluster with lesser noise is selected as non-dominated solutions. Thereafter, selection, mutation and crossover operator are used to explore the proposed technique further. After, getting the termination condition, tuned parameters for DBSCAN are obtained. Extensive experiments are carried out by considering benchmark remote sensing images (i.e., obtained from satellite sensors such as QUICKBIRD, IKONOS, MODIS, SPOT etc.). From visual and quantitative analysis, it is found that the proposed technique outperforms existing techniques in terms of Accuracy and Root mean square error. Therefore, the proposed technique is more applicable to real-time imaging systems.