ABSTRACT
Bangladesh is an agricultural country. It is susceptible to flood-driven agricultural losses and suffers more during the rainy season. Thus, the agricultural production of the country gets affected which directly influences the economy of the country. Therefore, it is essential to determine the severity of the flood for the well-being of agriculture and human life. This is very essential for the economic development of an agriculture-based developing country. For this reason, we propose a method to detect the flood area by applying techniques namely K-Means Clustering Algorithm and Color Probability. The method performs the background subtraction with the dynamic K-Means Clustering Algorithm to separate the background from the input flood image. Then connected component labelling will extract blobs. After that morphological closing will fill small background color holes and color probability will find out the watercolour pixels to find out the flooded region from the input flood image. Following this technique, we have achieved 96% accuracy.
