Heterogeneous image change detection method and device based on convolutional neural network fusion

Abstract

The invention discloses a heterogeneous image change detection method and device based on convolutional neural network fusion. The heterogeneous image change detection method comprises the following steps: extracting the convolutional neural network construction and homogeneous transformation features of a heterogeneous image; according to the homogeneous transformation feature loss function, fusing content features and style form features between the heterogeneous images so as to carry out homogeneous space transformation on the heterogeneous images, and obtaining a homogeneous space transformation result; and performing change detection according to a homogeneous space transformation result to obtain a final heterogeneous image change detection result. According to the heterogeneous image change detection method, the change area of the heterogeneous images can still be accurately detected under the complex terrain condition, and then the homogeneous transformation characteristics between the heterogeneous images are comprehensively and deeply extracted.

Description

technical field [0001] The invention relates to the technical field of remote sensing image fusion target detection, in particular to a method and device for detecting changes in heterogeneous images based on convolutional neural network fusion. Background technique [0002] Related technologies, (1) Earthquake damage assessment of buildings using VHRoptical and SAR imagery, the pioneering work of change detection for scenes of heterogeneous remote sensing images, here, the scene of heterogeneous remote sensing images refers to the images before and after the change are heterogeneous , there is no additional homogeneous remote sensing image for fusion change detection. This paper creatively proposes a method to construct a homogeneous feature space, that is, to map heterogeneous images into homogeneous images for change detection. Specifically, the terrain parameters of the optical image before the change are extracted, and the SAR image before the change is synthesized acc...

AI Technical Summary

Problems solved by technology

These complex image backgrounds and truly changing image regions are easily confused in the features extracted by traditional change detection algorithms, which in turn lead to detection errors

[0011] To sum up, in related technologies, there are limitations of heterogeneous image detection algorithms in homogeneous space construction in complex terrain environments, and it is difficult to fully and accurately perform homogeneous space transformation under complex terrain conditions, which needs to be solved urgently

Images