Research On Non-uniformity Correction Of Hyperspectral Remote Sensing Images Based On CMOS Sensor

As the remote sensing imaging technology that has developed rapidly and widely used in recent years,hyperspectral remote sensing image has been widely used in airto-ground information collection because of its wide field of view,multi-spectral acquisition and real-time monitoring.However,because hyperspectral images are susceptible to band noise,the information integrity of remotely sensed images collected by hyperspectral cameras is destroyed,which causes greater difficulties in subsequent image data processing such as information extraction and target recognition.For the hyperspectral camera with a high frame rate CMOS(Complementary Metal-OxideSemiconductor)as an image sensor,the hyperspectral remote sensing image collected by CMOS may be seriously affected by band noise and the noise distribution is irregular.Therefore,the main research content of this article is to propose a non-uniformity correction algorithm with superior performance and better denoising effect for hyperspectral remote sensing images based on CMOS sensors.This article first introduces the background and significance of the research on the correction of band noise non-uniformity in hyperspectral remote sensing images,and summarizes the current research development status at home and abroad,then analyzes the causes of band noise based on the imaging characteristics of hyperspectral remote sensing images and performance characteristics,followed by the analysis of the noise degradation model,and clarifies the subjective and objective evaluation criteria for image non-uniformity correction.Secondly,in the space domain,an improved moving window moment matching algorithm based on window threshold decision is proposed to further suppress band noise in view of the characteristics and deficiencies of the moving window moment matching algorithm.Finally,based on the idea of joint space domain and transform domain,combining wavelet transform and moving window moment matching algorithm,an improved strip noise removal algorithm is proposed,which also achieves a good non-uniform correction effect.By comparing with other related de-banding noise algorithms and comparing objective data with subjective visual effects,the algorithms proposed in this paper have certain advantages,effectively suppressing band noise,while maintaining the edge information of the image,Therefore,it can be applied to the preprocessing of remote sensing data of optical loads such as space and aviation to realize the non-uniformity correction of remote sensing images.