An Uncertainty Modeling and Measurement Method for Remote Sensing Image Features

Abstract

The invention discloses an uncertainty modeling and measurement method for remote sensing image features. This method mainly consists of two parts: uncertainty modeling of remote sensing image features in the geographic space domain and uncertainty modeling in the feature space domain, and then weighted combination of the geographic space uncertainty and the feature space uncertainty to obtain a The integrated feature uncertainty index FUI (Feature Uncertainty Index) is used to more accurately and comprehensively measure the feature uncertainty of remote sensing images. The present invention takes into account the different performance characteristics of image feature uncertainty under different viewing angles, can provide pixel-by-pixel feature uncertainty quantification results of the entire image, and can minimize manual intervention. It has high accuracy and adaptability, fast calculation efficiency, strong operability, easy implementation and strong scalability of the whole model. The uncertainty quantification result has a high indication ability to the classification error, therefore, the present invention has high practical value.

Description

technical field [0001] The invention belongs to the field of remote sensing image processing and statistical modeling, in particular to an uncertainty modeling and measurement method for remote sensing image features. Background technique [0002] Remote sensing image classification products have very important application value in many aspects such as natural disaster monitoring, environmental protection, urban planning and decision making. But current remote sensing image classification techniques still cannot achieve 100% accuracy or a level of accuracy that is reliable enough to be completely convincing. The root cause of errors in image classification results and low reliability of classification results is that there are uncertainties in every link of remote sensing image classification, and these uncertainties will continue to propagate and accumulate during the classification process, which will eventually affect the accuracy of classification results. precision and...

AI Technical Summary

Problems solved by technology

It only focuses on the uncertainty of the classification results, but ignores the uncertainty in the image classification process. The uncertainty in the classification process is the source of the uncertainty in the classification results. Therefore, the uncertainty of the existing methods The improvement effect of quantitative results on classification results is not obvious

Images