Dual adaptive model for change detection in multispectral images

The change detection (CD), resembles as basic issues in Earth tracking attains the major research concern over the past few decades. There is a considerable enhancement in the CD resource data in view of the rapid evolution in the satellite sensors in the current years, which provides very-high-resolution multispectral image with copious change evidences. However, localizing the precise varying area is considered as the real challenge. Hence, this research attempts to develop the Dual adaptive model to precisely locate the real changed areas. The pixel evaluation is done by the fusion network that hybrid the pre-trained model like segnet, U-net, ResNet and Fc-densenet. The pre-trained model is hybridized by the fusion parameter that is productively trained by using the adaptive optimization. The experimental result exhibits that the Dual adaptive model exceeds the competent model considering accuracy, precision, recall and F1-measure.