Hyperspectral unknown category target detection method based on a probability model and deep learning

Abstract

The invention belongs to the field of hyper-spectral intelligent perception, and discloses a hyper-spectral unknown category target detection method based on a probability model and deep learning, which comprises the following steps: S1, inputting hyper-spectral training data into a trained CNN classification model, and outputting activation vectors of all samples of each category; S2, accumulating and averaging the activation vectors of all the samples which belong to the same category and are classified correctly to obtain an average activation vector, and representing the center of the category by using the average activation vector; S3, fitting a Weibull model belonging to each category based on the activation vectors of all the samples in each category and the mean activation vector of the category; S4, inputting hyperspectral test data into the CNN model and the probability model to form a network based on the Weibull fitting result of each category, and calculating the probability belonging to an unknown category; The method is clear in structure and easy to implement, the training requirement of a neural network learning model is lowered, and the effect of unknown categorytarget detection can be obviously improved.

Description

technical field [0001] The present invention mainly relates to the field of hyperspectral intelligent perception, in particular to a hyperspectral target detection method based on a probability model and deep learning. Background technique [0002] Hyperspectral data characterizes the two-dimensional information of the distribution of ground objects and the one-dimensional information describing the spectral characteristics of ground objects. And the continuous spectrum curve. Compared with visible light and multispectral, hyperspectral has more abundant spectral information of ground objects, which can reflect the subtle spectral attributes of ground objects to be measured in detail. According to the differences in spectral characteristics between different ground objects to be measured, people use computers or special equipment to run classification and recognition programs to assign labels for category marking to each pixel, and divide the ground objects to be measured i...

AI Technical Summary

Problems solved by technology

The deep neural network only uses the SoftMax layer to detect unknown category targets, and its detection and recognition accuracy is not high

Although the hyperspectral anomaly target detection algorithm can also be used for hyperspectral unknown category target detection, its calculation is more complicated and the running time is longer

Images