Change detection - siamese based framework to detect changes over the earth's surface (CD-CSNN).

Recognizing the change over the earth's surface in the context of remote sensing identities needs to find the differences between the images. The manual execution of this task will be erroneous and incompetent to scale. To automate this process, we propose a novel framework that uses a customized Siamese-based neural network pipeline (CSNN) to detect the changes over the earth's surface. Based on the spatiotemporal analysis, the proposed work leverage's the ancient divide and conquers strategy. The image is divided into sub-image, and the feature maps are extracted using convolution layers that detect the fine-grained changes which occur at the sub-image level. The proposed methodology uses the LEVIR-CD dataset, which has 637 images. The performance of the proposed change detection method increases, as the input image size gets reduced, this is experimentally proved by decreasing the original image size into sub-images of size 2 x 2. The best performance achieved at 2 x 2 sized sub-image is 94.15%, 87.0%, 92.0%, 89.40% for accuracy, precision, recall, and F1-score respectively, which outperforms the results of the baseline algorithm. Further, the proposed framework performs well on different terrains, with varying amounts and types of changes in the satellite image. [ABSTRACT FROM AUTHOR]