A Calculation Method of Generic Probability of 3D Spatial Spectrum Spatial Dimension Pixels

Abstract

The invention discloses a densely connected three-dimensional space-spectrum separation convolution depth network and a construction method. The method includes: the input layer is three-dimensional space-spectrum data; the hidden layer unit consists of a spectral dimension one-dimensional convolution layer and a space dimension two-dimensional convolution Each hidden layer unit first performs spectral dimension convolution, and then performs spatial dimension convolution; the deep network is formed by stacking multi-layer hidden layer units; the input of each hidden layer unit is the input of all previous hidden layer units The output is concatenated; the output layer is the generic probability data of each spectral pixel; the network loss function is the mask cross entropy. The invention has the performance of rapid learning of spatial spectrum features and supervised classification of small sample labeling, better solves the imbalance of classification labels, and has excellent performance when applied to supervised classification of hyperspectral images.

Description

technical field [0001] The invention relates to a deep network technology, in particular to a densely connected three-dimensional spatial spectrum separation convolution deep network and a construction method. Background technique [0002] Deep learning has now become one of the important methods in the field of machine learning. In the field of deep learning, feedforward neural networks (FNNs) can approximate arbitrary functions with arbitrary precision and are widely used in regression, classification and other tasks. Recurrent Neural Networks (RNNs) have long-term dependencies and are often used in tasks such as speech recognition and machine translation. Convolutional Neural Networks (CNNs) have the ability to extract spatial features and are widely used in tasks such as object recognition and semantic segmentation. Different types of networks are suitable for different types of tasks and data formats. [0003] Spectral data is typically one-dimensional data. Each sp...

AI Technical Summary

Problems solved by technology

Using conventional convolutional neural networks to process spatial spectral data tends to weaken the contribution of spectral information to classification tasks

Images