Zeitschrift für Fernerkundung und GIS

Abstrakt

Data Identification of a Temporal Data for Soft classifiers FCM & PCM

Ranjana Sharma, P. K. Garg, R.K.Dwivedi, Mohan Vishal Gupta

In general, multispectral classifiers provide a complete suite of options for image classification using supervised, unsupervised or fuzzy based approaches. The image processing falls into 10 categories: restoration of image, image enhancement, image transformation, signature development of image, hard classifiers and soft classifiers for image , hardeners and hyper spectral analysis of image and accuracy assessment of result. Hard classifiers are commonly used in image classification, where a pixel has a membership value of either 0 or 1, thus it is considered as a pure pixel. The nature of pixel in soft classifier is mixed. The pixel of soft classifiers belongs to multiple classes. By theory of fuzzy set we can resolve the problem of multiple belongingness pixel of image. The ranges of membership value in fuzzy set are 0 and 1 where the value between 0 and 1 defines the proportion of occurrence of information within a pixel. This concept has been used in many applications, such as sensor signal analysis, uncertainty minimization. In this study, fuzzy soft classifiers and hybrid fuzzy based classifier with entropy, entropy based noise clustering have been used to learn the result of accuracy method(entropy) on classifiers output for multi-spectral data sets at pixel level. But any classification is considered to be incomplete without assessment of its accuracy. Various commercial companies have introduced variety of image processing tool which offer a related module to data input, visualization, enhancements, transformations, classification, accuracy assessment and output coupled with other GIS based modules. Some of the leading GIS software which have well defined image processing module are ERDAS Imagine, IDRISI, ENVI, and ER Mapper but the assessment of accuracy is not support by these software for the evaluation of soft classified output. So, a tool has been developed to handle such problems in this study. This tool mainly focuses soft classification algorithm. It has been named as Fuzzy Based Image classifier Tool (FBICET) which incorporates Entropy. The satellite image has been successfully classified with good accuracy using the FBICET.