Automatic machine learning method and system for remote sensing semantic segmentation

Abstract

The invention provides an automatic machine learning method and system for remote sensing image semantic segmentation, belongs to the field of computer artificial intelligence, and aims to improve thesegmentation accuracy through background loss optimization and multi-remote sensing imaging index calculation based on a remote sensing image semantic segmentation algorithm of multi-remote sensing imaging indexes. Through a self-migration fine tuning training method and a hyper-parameter optimization method, hyper-parameter optimization is carried out on the deep neural network for a single-machine multi-GPU environment, and the segmentation accuracy and the hyper-parameter optimization efficiency can be improved at the same time. A parameter sharing search space is set, and multi-scale information of the deep neural network is extracted through a cavity space pyramid pooling module on the basis of a strategy gradient search strategy, so that the optimal internal network structure of thedeep neural network is searched, and the high efficiency of remote sensing image semantic segmentation and the accuracy of classification are improved.

Description

technical field [0001] The invention relates to an automatic machine learning method and system for remote sensing semantic segmentation, belonging to the field of computer artificial intelligence. Background technique [0002] Remote sensing, that is, remote sensing, is a science and technology that uses sensors to detect the reflection, radiation or scattering of electromagnetic wave signals by ground objects in a non-contact and long-distance manner. Remote sensing uses detection instruments on artificial earth satellites, aviation and other platforms to observe the earth's surface on a large scale far beyond the human visual space, comprehensively displaying the shape and distribution of various things including geology and hydrology. Remote sensing technology is widely used in agriculture, forestry, ocean, geology and other fields, and the ground object information extracted from it has played an important role in tasks such as natural disaster emergency response and ag...

AI Technical Summary

Problems solved by technology

Most of the currently applied network structures are manually designed by deep learning experts. The design process is time-consuming, labor-intensive and error-prone

Images