Global And Local Feature Extraction Methods In Multi-Source Remote Sensing Images

As an important kind of spatial information data, remote sensing images are of great significance on the aspects of defense, the national economy and social development. With the rapid development of sensor technology, more and more rich information is contained in the remote sensing image acquired by different types of sensors. However, this situation brings about two problems. Firstly, how to make full use of the information contained in the multi-source remote sensing image, especially the complementary information among them. Secondly, how to analysis, deal with and interpret these huge amount of data in the most efficient and high quality way. These are the two urgent problems in the remote sensing application.This paper explores to make full use of the respective characteristics of multispectral, panchromatic and SAR image from different kinds of sensors and presents a method to analysis multi-source remote sensing image in a global and local way. Due to multi-spectral images are with richer color information, panchromatic images have higher resolution, scattering characteristic in SAR images is significant. Especially, the man-made objects in the SAR images usually appear as highlighting points. The global and local feature extraction methods are designed for the multi-spectral, panchromatic, and SAR images.The paper uses the three images, extracts a kind of global feature—color features respectively in multi-spectral images to divide land and water, then extracts two kinds of local features—context feature for dock and geometric feature for ship in panchromatic images, extracts another kind of local feature—statistical intensity features in SAR images. Finally, realize the segmentation of sea dock and ship combined with multiple features. The experimental results show the method is effectively to segment dock and ship...