Per-pixel classification-based remote sensing image scene classification and extraction method

Abstract

The invention discloses a remote sensing image scene classification system. The system comprises an acquisition step, a grayscale processor, a fitting step, an edge detection step, a remote sensing image pixel classification step and a neural network trainer, wherein the acquisition step is used for acquiring original remote sensing images and transmitting the original remote sensing images to thegrayscale processor as samples; the grayscale processor is used for carrying out grayscale processing on the original remote sensing images transmitted in the acquisition step by adoption of a component method; the fitting step is used for fitting a grayscale histogram by adoption of a low-order spline function; the edge detection step is used for finding zero cross points, obtained by the images, of second derivatives by adoption of a zero cross-based method, so as to position edges; the remote sensing pixel classification step is used for judging surface feature category attributes expressed by pixels by adoption of pixel-based classification and carrying out classification to obtain a classified thematic map; and the neural network trainer is used for inputting the images into a convolutional neural network model to carry out training, so as to obtain a classification results, achieving requirement precision, of remote sensing image scenes. The system is high in classification correctness.

Description

technical field [0001] The invention relates to the field of remote sensing image classification, in particular to a remote sensing image scene classification extraction method based on pixel-by-pixel classification. Background technique [0002] At present, many famous scholars at home and abroad are devoted to the research of image classification algorithm. Image classification is an image processing method that distinguishes different types of objects according to the different characteristics reflected in the image information. [0003] However, the shortcomings of the existing scene image classification and extraction in the prior art mainly lie in (1) in the traditional classification method, the manual extraction of features is quite time-consuming and labor-intensive, and requires high accuracy, and generally the accuracy is not as good as that of neural network classification. method. (2) When using neural networks and related models to implicitly extract features ...

AI Technical Summary

Problems solved by technology

In general, the number of commonly used data sets or self-made data sets is not enough to allow the neural network to fully learn and avoid over-fitting, resulting in an accuracy that cannot meet expectations.

Images