ENHANCED FROST FILTER AND COSINE TANIMOTO CLASSIFICATION BASED NATURAL DISASTER MANAGEMENT WITH SATELLITE IMAGES

Abstract

Natural disasters are utmost incidents inside the earth''''s system that lead to sudden demise or bruise to humans, and destruction of precious materials, involving buildings, conveyance systems, farming land, forest and natural environment. Occurrences of economic losses due to natural disasters have resulted owing to the escalated susceptibility of the society globally and also due to weather-related disasters. Satellite image sensing remains the hypothetical instrument for disaster management as it provides information spanning wide-reaching areas and also at short time period. In this work we plan to develop a method called, Enhanced Frost Filter and Tanimoto Similarity Classification (EFF-TSC) for efficient disaster management using satellite images is proposed. The EFF-TSC method for disaster management is split into three steps. They are pre-processing, segmentation and classification. With the input image collected from satellite image database, first preprocessing is performed to preserve important features at the edges and remove the noisy pixel by means of an Enhanced Frost Filter Preprocessing model. Second, to the pre-processed satellite image, Threshold Pixel Segmentation is applied to partition into multiple segments. Finally, to the partitioned images, Tanimoto Similarity Classification is applied to classify the segmented image into two types, namely disastrous image and non-disastrous image. With this, an efficient disaster management is carried out with better accuracy and minimal time consumption. The application of the study is demonstrated using the Disaster image data set collected from Kaggle during the 2017. The results show the capability of the proposed EFFTSC method for disaster management across time and space from different images with considerable accuracy by also reducing peak signal to noise ratio with considerable time. The findings also suggest that the potential for forensic analysis of disasters using pixel segmentation and classification based on collected images can be utilized to several locations affected by disasters.

Keywords

• disaster management
• frost filter
• threshold pixel segmentation
• tanimoto similarity classification