Research On Image Registration And Fusion Technology Of DMZ ? Camera Based On Adaptive Triangulation

With the development of photogrammetry and remote sensing technology,the area array digital aerial camera has become one of the main technical means for efficient acquisition of aerial mapping and remote sensing images.For image data that requires high spatial resolution and high spectral resolution in remote sensing applications,if an aerial camera is designed and produced to directly acquire the image data,the camera system will inevitably have many problems such as huge cost,large size,and difficulty in use.Therefore,in order to solve the above problems,the area array aerial camera basically adopts the current mainstream remote sensing observation technology,that is,the panchromatic camera and the multi-spectral camera are independently designed and separately imaged,and the panchromatic image and the multi-spectral image are acquired at a certain resolution.This has promoted the rapid development of image registration and fusion technology,making it a hot topic at home and abroad.This paper studies the image registration and fusion technology of DMZ ? aerial camera.According to the characteristics of DMZ ? aerial camera image,the research work of aviation multi-spectral image registration processing,panchromatic image and multi-spectral image registration processing,image information fusion processing and fusion image quality evaluation are completed.The specific research content of this paper has the following points:1.Briefly introduced the imaging principle and image characteristics of DMZ camera system,and expounded the current research status of image registration and image fusion technology.The basic theoretical knowledge that needs to be used in image registration and fusion processing is introduced,including the principle of feature point extraction algorithm,similarity measure,spatial transformation model,image differential correction principle and classical fusion algorithm principle.2.Aiming at the large field of view and large deformation of multi-spectral image of DMZ ? camera,an adaptive multi-spectral image registration method based on adaptive triangulation is proposed.The green band image is used as the reference image,and the other band images are registered by the method,and the final registration accuracy is less than 1 pixel,reaching the sub-pixel level,and the full registration of the multi-band imageoverlapping portion can be realized,and The registration accuracy is higher than the registration accuracy of the cubic polynomial transformation.The method has strong applicability in the registration process between multi-spectral images of DMZ ? camera,and can achieve high-precision registration of aerial multi-spectral images.3.Using SIFT algorithm and GPU fast processing technology to achieve different levels of image registration processing.The registration accuracy between panchromatic images and multi-spectral images is basically about 1 pixel.Except for the near-infrared band,the registration accuracy of the other three images reaches the sub-pixel level.The number of points with the same name is too small through analysis.Uneven distribution is the main reason for the low accuracy of registration.4.Image fusion experiments were carried out based on the results of image registration processing.The image fusion algorithms used include PCA transformation,ratio transformation,multiplicative transformation and HPF fusion algorithm,as well as wavelet fusion and IHS combined wavelet transform in the "area camera photography data processing system".Six evaluation indicators were used to objectively evaluate the quality of the fusion image: mean,standard deviation,average gradient,spatial frequency,information entropy and correlation coefficient.Finally,through the comprehensive analysis of the combination of objective evaluation and subjective evaluation,we can see that the multiplicative fusion algorithm can obtain the best fusion effect among the six fusion algorithms.At the same time,when the large-area aerial camera image is processed,the simpler fusion algorithm is used to obtain the fusion effect,and the fusion image with good quality and high spatial resolution and multi-spectral information can be obtained.