| Compared with the traditional airborne SAR imaging algorithm, BP algorithm hasthe advantages of the simple principle and no deviation in imaging processing,especially can be used in any mode of SAR imaging. Therefore, relevant research aboutBP algorithm is of important practical application value. This thesis mainly studiesmotion compensation method and GPU parallel implementation method about airborneSAR BP algorithm for imaging. The main contents are summarized as follows.First, the sources of error of BP imaging algorithm is theoretically analyzed. Aconclusion that motion compensation is needed for BP algorithm is obtained. Themeasurement erro range of navigation equipment is analysised under radar parametersused in this thesis by simulation. And the parallel processing technology, CUDAplatform software mode and GPU architecture are researched.Second, according to the positional relationship between the navigation equipmentand antenna phase center needed for BP algorithm, methods on coordinatetransformation and fitting resampling are studied. The experimental data show thesemethods are available. At the same time, a technique estimating the radar squint angleand course speed from echo data is researched. This technique is derived from radontransformmation, average phase increment method and DCFT transformmation.Finally, aiming to solve the problem of inefficiency of BP algorithm for imaging,the method of GPU parallel to realize BP algorithm based on pixel is investigated.Because of big data size during actual imaging process and the capacity limitation ofGPU memory, one imaging algorithm named stream BP algorithm mode is designed.This method’s correctness and effectiveness are verified by simulation and measureddata. And also the conclusion that this method can effectively improve imagingefficiency is obtained. This method can basically meet the requirement of the mostpratical SAR imaging processing applications. |
PDF Full Text Request