Analysis of rain fall and the temperature of Coimbatore District using land use and land cover change detection by image segmentation

Abstract

Image segmentation is a process has done for the classification of high resolution remote sensing images in the present research work. The segmentation results are capable of influencing the subsequent process effects. An image can be partitioned into a number of disjoint segments which is used to represent the image structures. It is found that it is more compact to represent an image and the low level and high structures can be combined. There are different types of methods to segment an image namely, threshold-based, edge-based and region-based. Region growing approach is image segmentation methods in which the neighboring pixels are examined and merged with the class region in case of no edges are detected. The iteration is done for every pixel boundary. Unlike gradient and Laplacian methods, the edges of the region are found by the region growing and it is perfectly their region. The images are determined by the LANDSAT TM satellite data. The remote sensing technique is used for collecting information about the Coimbatore district. The sensed data is a key to many diverse applications. The contribution of this work for Coimbatore district is to find the change of the Land used and Land covered in the entire region and also to find the changes in the green lands, vegetation and Land surface utilized for urban area. The neighboring regions are taken into account and the similarities are checked in the growing process. No single region is allowed to dominate the entire proceedings. A certain number of regions are allowed to grow at a time. Comparable regions will gradually combine into expanding regions. The Control of these methods may be quite complicated but efficient methods have been developed. The directions of growing pixels are easy and efficient to implement on parallel computers. The threshold-based segmentation is completely depending on the gray level images which regards the reflectivity of the featured images. It determines a threshold based on brightness of the ground objects. It is purely from the image background. But it is rapid and its uncertainty is significant. It is not convenient to process multi-spectral images.