Remote sensing image thin and weak target segmentation method

Abstract

The invention provides a remote sensing image thin and weak target segmentation method. Firstly, data enhancement and corresponding preprocessing are carried out on an original remote sensing image, U-net is improved by means of the dense connection thought of DenseNet, and a Dense-Unet network structure is provided. Dense convolution is used in a network structure, the cascade relation between convolution channels is enhanced, through a symmetric structure and a jump connection thought, the connection between features of all layers is further tighter, and thin and weak target features can belearned more effectively. In order to ensure the real-time performance of final network identification and reduce the parameter quantity, a bottleneck layer and a batch normalization layer are introduced behind each dense block. And the objective function is adjusted by using the cost-sensitive vector weight, so that the problem of unbalanced segmentation target categories is solved, and the segmentation precision is further improved. And finally, a plurality of independent models are trained by using an ensemble learning method, the independent models are combined, and target category information is jointly predicted in the picture.

Description

technical field [0001] The invention relates to a digital image processing method, in particular to a method for accurately segmenting faint targets in remote sensing images. Background technique [0002] Remote sensing image target segmentation is an important technical method for remote sensing image target recognition, which is widely used in many fields such as environmental assessment, traffic planning, and automatic driving. Semantic segmentation of images is the key to understanding image information. The basic principle is to segment pixels into different regions according to the semantic meaning expressed in the image, that is, to recognize the image at the pixel level and mark the object category of each pixel. With the rapid development of remote sensing technology, high-resolution remote sensing satellite images have the characteristics of wide observation range, containing more object information, and difficult to extract information features. When traditional ...

AI Technical Summary

Problems solved by technology

[0003] However, the above deep learning networks all require large-scale data sets and accurate manual labeling, and all face the problems of indistinct features of subtle targets in remote sensing images, unbalanced categories, and large background interference. Poor target recognition accuracy

Images