Hyperspectral remote sensing image target detection method based on multi-task learning

Abstract

The invention relates to a hyperspectral remote sensing image target detection method, in particular to a hyperspectral remote sensing image target detection method based on multi-task learning, and belongs to the technical field of image detection. The method comprises the following steps: firstly, determining the size of a neighborhood to be detected, calculating a neighborhood multi-feature stack by using a multi-feature learning algorithm, carrying out loop iteration for multiple times, traversing all pixels in a hyperspectral remote sensing image, and mapping a nonlinear hyperspectral remote sensing image to a multi-feature space; calculating a joint representation vector of the pixel feature stack in each window through a multi-task learning optimization function according to a givenover-complete dictionary; and finally, reconstructing a pixel space by using the over-complete dictionary and the joint representation vector. According to the method, the hyperspectral remote sensing image data is mapped into the feature space with high separability by introducing the multi-feature learning and multi-task learning optimization algorithm, so that the detection precision of the hyperspectral target detection method is improved.

Description

technical field [0001] The invention relates to a hyperspectral remote sensing image target detection method, in particular to a hyperspectral remote sensing image target detection method based on multi-task learning, which belongs to the technical field of image detection. Background technique [0002] Hyperspectral remote sensing data is a massive and heterogeneous data with high spectral resolution and rich spectral bands, so it has long attracted the attention of engineers and scholars. In hyperspectral data, the spectral reflectance curves of different ground objects are different. Using the above properties, different ground objects can be distinguished in the image. Target detection is an important application direction of hyperspectral remote sensing data, and its role is to identify pixels in remote sensing images as targets and backgrounds. Traditional hyperspectral image target detection methods include spectral matched filter (Spectral matched filter, SMF), supp...

AI Technical Summary

Problems solved by technology

However, the above method only utilizes the spectral information of the hyperspectral image, without considering its spatial information and multi-feature information, resulting in low detection accuracy

Images