Small Target Segmentation Of Remote Sensing Image Based On Multi-Scale Information Fusion

With the rapid development of remote sensing technology,remote sensing image processing has gradually achieved many practical implementations and has received extensive attention.Semantic segmentation of remote sensing images is an important research direction of remote sensing image processing,which can be applied in scenarios such as land exploration.Different from natural image-oriented semantic segmentation,remote sensing images are often accompanied by the problem of small objects.In addition,since remote sensing images have a variety of data types,and different data types have their unique research value,it is necessary to broaden the research vision to multiple data types.This paper aims at the small target problem of remote sensing images,based on the idea of multi-scale information fusion,and based on the indepth analysis of the potential problems of the three data types of visible light images,hyperspectral images,and optical SAR images,and puts forward corresponding solutions Solve the algorithm,and finally develop a remote sensing image semantic segmentation system based on the Web framework.The main work results of the thesis are summarized as follows:(1)Aiming at the high-resolution problem of visible light images,a cross-resolution semantic segmentation network is proposed.By sampling at multiple resolution granularities,the trade-off between global features and local features of high-resolution images is solved.(2)Aiming at the problem of low samples in hyperspectral remote sensing images,based on the active learning method,a spectral spatial j oint segmentation network is proposed,which improves the network performance by retaining both spectral information and spatial information,and solves the semantic segmentation problem under a few samples.(3)Aiming at the multimodal problem of optical SAR images at the same time,a multi-scale and multi-modal semantic segmentation network is proposed to make better use of optical images and SAR images by deeply combining features of different scales in different modalities Complementary information when present at the same time.The methods proposed in this paper are all tested on the corresponding data sets to verify the effectiveness of the algorithm.In addition,in order to show the effect of the algorithm more intuitively,this paper designs a remote sensing image semantic segmentation system based on the Web framework,integrates several algorithms proposed in this paper,and provides technical support for the remote sensing image semantic segmentation task.