High-resolution remote sensing image classification method based on parallel hybrid convolutional network

Abstract

The invention discloses a high-resolution remote sensing image classification method based on a parallel hybrid convolutional network, and the method comprises the specific steps: inputting high-resolution remote sensing images and corresponding sample label data, including a training sample data set and a test sample data set; building a three-dimensional convolutional neural network and a two-dimensional convolutional neural network in parallel, and building an information fusion conversion network to realize spatial-spectral feature information fusion and deep extraction; inputting training sample data sets in batches to train a network, constructing a cross entropy loss function and a stochastic gradient descent algorithm to optimize the network, and updating parameters until the network converges; and inputting a test sample data set into the hybrid network model, and outputting a test sample label prediction value to complete high-resolution image classification. According to the method, the spatial features and the spectral features of the high-resolution remote sensing images can be extracted at the same time, high-efficiency and high-precision classification of the images is achieved through feature fusion, and the method plays an important role in research of natural resource monitoring, geographic national condition census, urban planning, climate change and the like.

Description

technical field [0001] The invention belongs to the technical field of image processing, and in particular relates to a high-resolution remote sensing image classification method based on a parallel hybrid convolution network. Background technique [0002] With the rapid development of remote sensing satellite technology, the resolution of remote sensing images has gradually increased. my country's high-resolution remote sensing technology is developing rapidly. At present, a major special project of high-resolution earth observation system has been launched to improve the ability to obtain high-resolution remote sensing images and promote the progress of spatial information technology. The release of sub-meter-level resolution remote sensing images of Gaofen-2 satellite marks that my country's remote sensing earth observation has entered the sub-meter-level era, and its sub-satellite point spatial resolution can reach 0.8 meters. High spatial resolution remote sensing imag...

AI Technical Summary

Problems solved by technology

[0006] Aiming at the shortcomings of the existing high-resolution remote sensing image classification methods, the present invention proposes a high-resolution remote sensing image classification method based on a parallel hybrid convolution network, which combines the advantages of three-dimensional convolution and two-dimensional convolution to extract deep spatial Spectral features and fusion transformation to improve the classification accuracy and efficiency of high-resolution remote sensing images

Images