Research On Fast Processing Technology Of Atmospheric Radiation Transmission Characteristics Based On GPU

In recent years, the development of infrared image processing technology has put forward the demand for fast and accurate infrared image processing. And GPU (graphics processing unit) high performance heterogeneous computing as a new high performance computing solution, with its high computing capacity and low cost is being used more and more widely in all walks of life. In the past few years, the high performance computing based on GPU achieved a rapid development, and a lot of researches have shown that GPU has great advantages compared with the traditional CPU computing.The thesis comes from an actual project The Key Technology and Demonstration of Promoting helicopter’s rescue ability on Severe Sea Condition’, the key technology and application system research of finding target quickly on Severe Sea Condition, which is a major science and technology project in the ministry of communications. In order to find the distress target accurately and quickly without the radiation interference, our group has established the atmospheric radiation transmission characteristic model and designed the calculation software. However due to the large amount of computation in the solution process, it is not able to use existing CPU to achieve the goal of real-time computing. So in order to make preparations for the image pretreatment as well as the accurate and timely detection of the target, this study uses GPU to quickly calculate the impact of atmospheric radiation transmission characteristics on the imaging path in real time.This paper start from the domestic and foreign related research about GPU acceleration and atmospheric radiation transmission characteristics parameters fast calculation. Firstly, it introduces the hardware structure and programming model of GPU heterogeneous computing platform. Then, by analyzing the features of the atmospheric radiation transmission characteristics parameter processing algorithms, the AMP C++ programming language with MVC program model was used to realize the fast calculation of the atmospheric transmission characteristic parameters on the GPU platform. Finally, by comparing the calculation results, the accuracy of the program was verified. And through data analysis, it could be found that AMD radeon HD 7870 graphics can reduce the average running time from 2304ms to 18ms in visible band and from 407ms to 16ms in mid infrared band, which satisfy the request of real-time image processing that should less than 40 ms.