Aerial image segmentation method based on hierarchical context network

Abstract

The invention discloses an aerial image segmentation method based on a hierarchical context network. The method comprises the steps of firstly designing and constructing a pixel point-pixel point sub-network, then designing and constructing a pixel point-object sub-network, and forming a hierarchical context network according to the constructed pixel point-pixel point sub-network and pixel point-object sub-network, obtaining hierarchical context information, and then completing the segmentation operation of the aerial image by using the obtained hierarchical context information. According to the method, hierarchical context information of two granularities of semantics and details is constructed, so that the category of a target object is better helped to be judged, the spatial detail information of the target object is described, category feature representation is directly learned from an image by using an unsupervised clustering method, the classification relevance implied by feature representation is utilized, convolutional features are further helped to construct hierarchical context information, and the finally proposed hierarchical context network obtains the optimal segmentation performance on two public competition data sets and GF-2 satellite data.

Description

technical field [0001] The invention relates to the technical field of image segmentation, in particular to an aerial image segmentation method based on a hierarchical context network. Background technique [0002] High-resolution aerial image segmentation is crucial for many applications, such as urban change detection, disaster relief and refined agriculture. The purpose of this task is to determine the category of each pixel in the image; in high-resolution scenes , the heterogeneous appearance of objects like buildings, streets, trees, and cars can easily lead to differences within large classes and between small classes; exploring contextual information has been widely recognized as an effective way to solve this task problem, and in the past few In the middle of the year, convolutional neural networks are an optimal choice for capturing contextual information; early methods based on convolutional neural networks (such as FCN-8s) try to learn contextual information thro...

AI Technical Summary

Problems solved by technology

[0002] High-resolution aerial image segmentation is crucial for many applications, such as urban change detection, disaster relief and refined agriculture. The purpose of this task is to determine the category of each pixel in the image; in high-resolution scenes , the heterogeneous appearance of objects like buildings, streets, trees, and cars can easily lead to differences within large classes and between small classes; exploring contextual information has been widely recognized as an effective way to solve this task problem, and in the past few In the middle of the year, convolutional neural networks are an optimal choice for capturing contextual information; early methods based on convolutional neural networks (such as FCN-8s) try to learn contextual information through an encoder-decoder structure, although these methods can successfully use convolutional neural network Convolutional kernels capture contextual information, but their performance is still limited by the size of their receptive field

[0003] Almost all current segmentation methods try to distinguish different objects through the pixel-pixel relationship; however, there is a probability that there are pixels with similar appearance in different categories of object regions, for example, gray vehicles and gray roofs in aerial images Looking down from the air is very similar, which will further lead to the pixel-pixel relationship is easy to deduce wrong segmentation results, and it is difficult to distinguish confusing objects; therefore, it is necessary to design an aerial image segmentation method based on a hierarchical context network

Images