Remote sensing image change detection method based on watershed and treelet algorithms

Abstract

The invention discloses a remote sensing image change detection method based on watershed and treelet algorithms, belonging to the technical field of image processing. The remote sensing image variation detection method comprises the steps of: imputing two remote sensing images with different phases, denoising each image to obtain two denoised images with different phases and constructing a difference-value difference image; applying the watershed algorithm to obtain an initial over-segmentation labeled graph of the different image and adopting the treelet algorithm to combine the textural features of an over-segmentation labeled region, obtaining variable and non-variable binary images by clustering, and carrying out spatial restraint on the binary images by virtue of a variable-weight Markov stochastic field model to obtain post-processed images with edge consistency, and combing the communicating regions of the clustered binary images and the images with edge consistency to obtain a final change detection result. The method disclosed by the invention can effectively improve the processing efficiency of change detection and keep the edge information of the images, and can be used for detection of image changes in disaster monitoring and land utilization.

Description

technical field [0001] The invention belongs to the technical field of digital image processing, and specifically relates to a remote sensing image change detection method based on watershed and treelet, which is suitable for image processing in agriculture, environment, and urban planning. Background technique [0002] The change detection of remote sensing images refers to the analysis of remote sensing images of the same geographical location in different periods to obtain the change information, which is the main development direction of the current remote sensing data processing technology. [0003] The change detection method for two remote sensing images after registration is generally to obtain the difference map first, and then classify the difference map as change or non-change. This method does not change the data itself because it directly compares the differences of the original data, and the information is more reliable. A common method to classify the differe...

AI Technical Summary

Problems solved by technology

[0004] In order to improve the accuracy of change detection, scholars such as Bruzzone proposed a method of adaptively selecting the image threshold in the article "Automatic Analysis of the Difference Image for Unsupervised Change Detection", but there will still be noise areas that are difficult to distinguish by the threshold algorithm, thus affecting The result of change detection

In the article "A new change detection method based on non-parametric density estimation and Markov random fields", scholars such as Guiting Wang proposed a change detection segmentation method based on the probability density of the Markov field. This method is based on random statistical samples Obtaining the probability density of the overall data requires a large number of statistical samples to obtain accurate results, and the correctness of the method is easily affected

Images