The Research Of Orthophoto Generation Based On GPU Parallel Acceleration

DOM(Digital Orthophoto Map) is one the most important achievements in the field of photogrammetry. As the rapid development of technology and economy, how to generate DOM quickly in order to meet the needs of economy and scientific research has become the key target to the photogrammetric researchists and computer workers. However, the period of generating DOM is so long that the deduction of generating time range is the conundrum in this field, due to the limitation of the computer software and hardware and the application of photogrammetry.In recent years, as computer technology develops much quicker, especially the GPU calculation capability, GPU is no longer only used in the image process, but in general-purpose computing. However, due to the absence of programming framework based on GPU hardware platform, the application of GPU in general-purpose computing is hard. The emergence of CUD A provides support for GPU in general-purpose computing. More and more institutions and corporations draw more attention to the advantages of GPU programming, and invests lots of manpower and material resources to welcome a new era of CPU +&GPU programming.Among the digital photogrammetry algorithms of image process, image data can be processed in parallel computing. The main research target of this paper focuses on GPU's parallel processing features, and has an in-depth research on the realization of the orthophoto generation GPU algorithm about double cubic convolution image interpolation and pixel coordinates. Through the GPU implementation of two types of algorithms, it analyzes the programming features and scopes of application, which gets many achievements.The main works are as follows:1.First introduced the concept of the digital photogrammetry and the DOM;2.Introduced the generative principle of the DOM and GPU's characteristics;3.Research on the realization of the orthophoto generation GPU algorithm about double cubic convolution image interpolation and pixel coordinates;4.Analysis the results of CUDA transformation.