An intelligent interpretation system for hyperspectral satellite on-board data and its realization method

Abstract

The present application discloses a hyperspectral satellite onboard data intelligent interpretation system and its implementation method, including a model generation module and an on-board inference module. The model generation module is used for training and deploying models, including sample data set production, model selection and Training and model quantization deployment, the onboard inference module is used to intelligently interpret the newly collected data using the trained neural network, including input data preprocessing, sliding window slicing, model inference, inference result splicing and post-processing, projection Geographic coordinates and send processing results. It has the following advantages: it can extract target information more objectively and efficiently, form thematic products, and reduce time and labor costs; by deploying the neural network on the satellite, users do not need to download the original image to the ground, which can solve the problem of timeliness, The cycle of information acquisition is greatly shortened.

Description

technical field [0001] The invention relates to a hyperspectral satellite onboard data intelligent interpretation system and an implementation method thereof, belonging to the technical field of image data recognition. Background technique [0002] In recent years, remote sensing technology has been widely used in ecological monitoring, economic development, military reconnaissance, resource exploration, disaster early warning, etc. Remote sensing satellites can be divided into high-resolution satellites, radar satellites, multi-spectral hyper / spectral satellites, etc., according to different sensors. Compared with high-resolution and radar satellites, hyperspectral satellites can capture a wider frequency range and higher resolution. Therefore, it is more advantageous in the analysis of ground objects. [0003] The traditional interpretation of hyperspectral image data requires the transmission of hyperspectral images from satellites to ground receiving stations. After a s...

AI Technical Summary

Problems solved by technology

On the one hand, due to the large number of bands in the hyperspectral image, the data volume of the image is extremely large, and it is difficult to transmit data to the ground station through relay satellite communication. In the case of limited distribution of ground stations, real-time and continuous observation become a problem

On the other hand, the downloaded images need to be analyzed by experts in the field, and the time and labor costs are huge

These two aspects have led to a long cycle from satellite acquisition to the generation of special products, and poor emergency response capabilities

Images