A View Stitching and Band Registration Method for Composite Cameras via Virtual Imaging
To overcome the challenges posed by the small intersection angle and unstable triangulation precision of large-format composite cameras,as well as the technical difficulties in band registration for multispectral composite cameras,the application of virtual imaging principles is explored in depth to expand the field of view and enhance spectral resolution.The research involves analyzing and optimizing the inherent geometric relationships between the individual physical cameras within a composite camera system,treating the position and attitude relationships of multiple physical cameras as rigid connections.By measuring homologous image points across different camera-captured images,the geometric transformation relationships between cameras are computed,and the images are reprojected onto a virtual imaging plane,ultimately achieving efficient and precise image fusion of independently captured images by each physical camera,in accordance with the principles of central projection.This optimization method not only extends the aerial photography range of large-format composite cameras but also improves the spectral resolution of multispectral composite cameras,particularly demonstrating exceptional adaptability and accuracy in processing textureless regions and terrain undulating scenes.Experimental results show that for large-format composite cameras,the image stitching accuracy of this method reaches 0.31 pixel;for multispectral composite cameras,the band registration accuracy achieves 0.25 pixel.This validates that virtual imaging technology can significantly enhance the processing performance and efficiency of composite camera data,and is fully compatible with existing photogrammetric processing software systems,thus broadening the scope of application of composite camera data in the field of aerial photogrammetry and offering new perspectives for the application of Geographic Information Systems.
国家科技期刊平台
NSTL
万方数据
