Remote sensing image semantic segmentation method and device based on full convolutional neural network

Abstract

The invention discloses a remote sensing image semantic segmentation method and device based on a full convolutional neural network, and the method comprises the steps: constructing the full convolutional neural network composed of a convolution module BasicConv and a bottleneck layer Bottleneck, and defining each dense layer in Dense Blocks as a composite operation of the bottleneck layer and the convolution module with the convolution kernel size of 3; under the condition that the sizes of the feature maps in the dense blocks are consistent, designing Down Sampling and Up Sampling modules to perform down-sampling and up-sampling; connecting different steps by down-sampling or up-sampling, and connecting the downward passage and the upward passage of the same step by jumping connection; and constructing a similarity measurement function, and realizing semantic segmentation of the remote sensing image based on the function. The device comprises a processor and a memory. The parameter quantity is remarkably reduced at the cost of a small amount of memory space overhead, the neural network model is compressed, and feature information of different levels and scales can be effectively integrated.

Description

technical field [0001] The invention relates to the field of remote sensing images, in particular to a method and device for semantic segmentation of remote sensing images based on a fully convolutional neural network. Background technique [0002] Image segmentation, an important component of many visual understanding systems, involves segmenting an image (or video frame) into segments or objects. Its development has experienced from the earliest algorithms such as thresholding, histogram-based grouping, region growth, k-means clustering, and watershed, to more advanced algorithms such as active contours, graph cuts, conditional and Markov random fields, and sparseness. . However, in the past few years, deep learning algorithms have produced a new generation of image segmentation models with significantly improved performance. [0003] With the development of deep learning, many excellent algorithms have emerged in the segmentation task. At present, the public and popula...

AI Technical Summary

Problems solved by technology

[0004] In the field of semantic segmentation of remote sensing images, due to the particularity of remote sensing images, the use of existing semantic segmentation algorithms will result in an output whose predicted resolution is lower than the input resolution, and will greatly occupy memory and increase computing overhead. Semantic segmentation algorithm to obtain ideal segmentation results, it is necessary to design a semantic segmentation algorithm that meets the task according to the particularity of the field

Images