Semi-Bayesian deep learning remote sensing scene classification method based on Markov chain Monte Carlo and variational inference

Abstract

The invention discloses a semi-Bayesian deep learning remote sensing scene classification method based on Markov chain Monte Carlo and variational inference. Firstly, a semi-Bayesian deep convolutional neural network is built, for a convolutional layer, only a half of the convolutional layer is adopted to represent weight parameters by using a Gaussian distribution method, and weight parameters ofa full connection layer are represented by using a Gaussian distribution method. Secondly, approximate weight parameter distribution is preliminarily obtained by using a variational inference method;then, a Markov chain Monte Carlo method and a variational contrast divergence method are utilized to further iteratively improve approximate weight parameter distribution, and more accurate approximate weight parameter distribution is obtained. According to the method, a semi-Bayesian deep learning method is utilized, part of weight parameters in the deep convolutional neural network are used asrandom variables, uncertainty is introduced into the network model, the over-fitting phenomenon is avoided, the robustness of the network model is enhanced, and the method has good performance in remote sensing image scene classification application.

Description

technical field [0001] The invention belongs to the field of image processing, and in particular relates to a semi-Bayesian deep learning remote sensing scene classification method based on Markov chain Monte Carlo and variational inference. Background technique [0002] Remote sensing image classification is one of the important research directions in the field of remote sensing, and it is widely used in many application fields such as geological survey, disaster monitoring, traffic supervision and global temperature change. Therefore, in-depth research on remote sensing image classification has important application value for the development of future society. There are many kinds of classification algorithms for remote sensing images, such as ISODATA, K-means, minimum distance, maximum likelihood and other algorithms. According to the principles and emphases of the above classification algorithms, they can be divided into several categories, such as supervised and non-su...

AI Technical Summary

Problems solved by technology

[0008] (1) Feature extraction is complex and inefficient: Traditional classification methods need to use different image feature extraction algorithms to extract various types of features of images, and the process is cumbersome and complicated

At the same time, these classification algorithms generally only have a shallow structure, and the learned expression features cannot fully cover remote sensing image information, there is redundant information, and their classification performance and generalization ability are obviously insufficient.

[0009] (2) Lack of measurement of prediction uncertainty: the existing convolutional neural network model applied to remote sensing classification uses the method of point estimation to represent the weight, which can better fit the training image samples after a large number of training, but However, it will cause overfitting on the test image sample and cannot correctly predict the image label

[0010] (3) Variational inference shows that there is a deviation between the approximate distribution function and the real distribution function: for the Bayesian estimation method, the variational inference method is generally used to approximate the distribution function, but the approximate distribution function cannot be guaranteed to be similar to the real distribution function The effect of the algorithm is not robust enough, and the classification performance is poor

Images