Detecting floods using an object based change detection approach

In change detection analysis, it is important to reduce the influence of image misalignment in order to produce image changes that are relevant to the user. The accuracy of change detection solely depends on the image registration accuracy yet image misalignment is still a major challenge in change detection analysis. In change detection analysis, if change detection is performed on misaligned images, problems such as false changes and missed changes may occur which strongly affect the actual change detection accuracy. Object based change detection has been reported as one the best way to reduce the influence of effects of image misalignment. This is because of its capability to deal with misregistration errors that result into false change and missed changes. In this paper, a tiled object based change detection technique is proposed to reduce the influence of image misalignment on the tile pixel based change detection. Threshold level based fuzzy c-mean clustering is adopted during segmentation and classification, as well as image differencing is used to obtain a change image. Experiments show that the proposed method significantly reduces the false changes (commission error (from 16.67% to 0.7059%, 40.63% to 1.3881%, 31.58% to 2.1034%) in tile pixel based change detection meaning the method is robust to noise. The experimental results show that the method is capable of producing a high accuracy rate.