Comprehensive method for improving small target segmentation precision

Abstract

The invention relates to the technical field of remote sensing images, and discloses a comprehensive method for improving small target segmentation precision, and the method comprises the following steps: S1, constructing a Tiramisu coding-decoding deep convolutional network, and carrying out the parameter migration based on a typical natural image sample training network; S2, selecting any sample of a certain type of ground features as a reference style of data augmentation, and inputting other samples into a coding-decoding structure to generate a plurality of samples of the same type with the same style and different contents; S3, performing data augmentation on other types of samples, and guaranteeing that the number of samples of different types of ground features is balanced. According to the comprehensive method for improving the small target segmentation precision, the Tiramisu coding-decoding deep convolutional network can carry out parameter reduction, the number of model learning parameters is small, the training speed is high, and the style migration data augmentation technology ensures the balance of different types of samples, so that the sample representativeness is more sufficient, and the trained model has good segmentation precision for various ground features.

Description

technical field [0001] The invention relates to the technical field of remote sensing images, in particular to a comprehensive method for improving the segmentation accuracy of small targets. Background technique [0002] Remote sensing images refer to films or photos that record the size of electromagnetic waves of various ground objects. They are mainly divided into aerial photos and satellite photos. The background information usually accounts for a large proportion, and the number of some special target samples is usually small, which makes the training process of the deep learning model easy to be biased towards the background. boundary, for which we propose a comprehensive approach to improve the segmentation accuracy of small objects. Contents of the invention [0003] (1) Solved technical problems [0004] In view of the deficiencies in the prior art, the present invention provides a comprehensive method for improving the segmentation accuracy of small objects, w...

AI Technical Summary

Problems solved by technology

[0004] Aiming at the deficiencies of the prior art, the present invention provides a comprehensive method for improving the segmentation accuracy of small objects, which can improve the balance of various remote sensing feature samples, optimize the deep learning model training process to enhance the model's attention to small objects, Thereby improving the model performance and the segmentation accuracy of small targets, etc., and solving the problems raised in the above-mentioned background technology

Images