Research And Implementation Of Parallel Processing Algorithm For Non-measurement Camera Image Based On OpenCL

With the development of non-measurement digital cameras and UAV platform technologies,the spatial resolution of remote sensing images acquired by UAV-equipped non-metric cameras has gradually increased,and the amount of information contained in images has been continuously enriched.A substantial increase in volume.Compared with the improvement of image acquisition technology,the traditional processing methods of image data appear to be lagging behind,resulting in low image processing efficiency.To solve this problem,this paper establishes a parallel processing algorithm based on OpenCL and GPU to greatly increase the computation speed of key image processing algorithms.This method adopts the heterogeneous parallel platform model of CPU+GPU,and allocates the pixel-intensive computing part of the traditional algorithm to the GPU for processing to accelerate the algorithm.In addition,the OpenCL development algorithm avoids the disadvantages of the CUDA model algorithm and the graphics card hardware.The main research work and conclusions of this paper are as follows:(1)Researching the software architecture of OpenCL heterogeneous parallel programming model and exploring the hardware architecture mechanism of GPU,forms a data parallel processing scheme based on OpenCL heterogeneous programming model;(2)Choose the aberration correction algorithm and the orthorectification algorithm in the whole process of image processing to parallel improve OpenCL,and form the acceleration processing software of the two algorithms.The experimental data are selected to test the improved algorithm on different hardware devices.The results show that under the condition that the image processing effect is not changed,the image processing speed is significantly improved compared to the traditional distortion correction and orthorectification algorithm,and the improved distortion is improved.The total acceleration ratio of the differential correction algorithm is up to 5.981,and the partial acceleration ratio is up to 63.297.The overall acceleration ratio of the improved orthorectification algorithm is up to 3.83,and the calculated partial acceleration ratio is up to 26.712.The algorithm has no limitation on hardware devices.(3)The whole flow of non-measurement camera image processing is analyzed.From the image preprocessing stage to the image’s air space encryption stage to the image output stage,the pixel-level algorithms can be improved in parallel through the OpenCL heterogeneous parallel platform.Improve the speed of the algorithm.Therefore,it is feasible to extend the heterogeneous Heterogeneous Algorithm based on OpenCL heterogeneous model to the whole process of remote sensing image processing.