Remote Sensing Image Change Detection Method Based On IR-MAD Algorithm

Remote sensing image change detection is a surface change recognition technology based on bi-temporal images.By comparing and analyzing multi-temporal images,it can judge whether each pixel in the scene has changed,distinguish the changed area from the unchanged area,and determine the nature of the change.It can be used in the fields of environmental protection and geological survey.In recent years,the field of remote sensing has developed rapidly,and the spectral resolution and spatial resolution of images have become higher and higher.The obtained remote sensing data provide rich information for the accurate identification of ground objects.It plays a huge role in urban construction,agricultural forestry and environmental survey of large forests.In order to make the method of remote sensing image change detection more accurate,this thesis improves on the existing methods,mainly in the following aspects:(1)IR-MAD iterative weighted multivariate detection algorithm is an improved algorithm of MAD algorithm,which has advantages in noise processing and precision detection.This method can efficiently capture the changed pixels,judge the nature of the change and accurately identify the change information.In this paper,18 datasets are selected as the research area,and the IR-MAD algorithm is written in Python language to conduct experiments,and generate multi-band detection results,true color detection results and single-band detection image results change graphs.The results show that the IR-MAD algorithm can obtain high-precision change detection results,and can also achieve good results in suppressing noise.It is an excellent change detection method.(2)The superpixel segmentation method of CD-BS is proposed.The two-phase image and the difference image are superimposed.After the superposition,the common area of the superpixel block in the same geographical location is obtained,and the appropriate threshold is set by the OTSU threshold method.,divide the pixels in the image,and divide all the pixels into two categories: those that have changed and those that have not changed;and this paper proposes a new method of measuring distance,which is used to calculate the distance between two marked samples.The similarity is obtained to obtain the preliminary change detection results;finally,three scales of superpixel segmentation are performed on the preliminary results to remove the influence of noise on the experimental results.With the high accuracy obtained,the method can realize change detection in images without prior knowledge.