Remote sensing image real-time processing method and system under GPU parallel and zero IO modes

Abstract

The invention discloses a remote sensing image real-time processing method and system under a GPU parallel and zero IO mode. The method comprises the following steps that real-time modeling is conducted on input data; image processing is carried out on the data subjected to real-time modeling by utilizing the GPU parallel and zero IO mode; and real-time modeling and image processing are operated by using a real-time processing framework. The method and system have the beneficial effects that traditional processing modes are subverted, the current processing node in the processing chain does not depend on the complete image processing result of the processing node in the last step; the real-time processing effect is achieved, intermediate results are not needed in the process, cross-level production is directly achieved, and therefore the production efficiency of conventional ortho-image production is improved, the production cost is saved, the mode is easy to use, and the DOM production capacity of software is greatly improved.

Description

technical field [0001] The invention relates to the field of high-performance processing of remote sensing images, in particular to a method and system for real-time processing of remote sensing images in GPU parallel and zero IO mode. Background technique [0002] With the continuous development of remote sensing technology, the amount of remote sensing image data is increasing day by day. In the process of remote sensing image processing, the production process is complicated, with many steps, a large number of input and output operations, and a large amount of intermediate data are generated, which is not conducive to the improvement of data management and storage utilization, and in the whole process, manual intervention and control are required. There are pressures and challenges in production technology, hardware storage devices, and production personnel; traditional remote sensing image production consumes a lot of resources, and the calculation cycle is long, and the...

AI Technical Summary

Problems solved by technology

In the process of remote sensing image processing, the production process is complicated, with many steps, a large number of input and output operations, and a large amount of intermediate data are generated, which is not conducive to the improvement of data management and storage utilization, and in the whole process, manual intervention and control are required. There are pressures and challenges in production technology, hardware storage devices, and production personnel; traditional remote sensing image production consumes a lot of resources, and the calculation cycle is long, and the subsequent processing process relies on the output of the previous step as input for current business parameter adjustment and operation processing, the whole production process takes a long time, the degree of automation is low, and the output of a large number of intermediate results occupies a huge amount of hard disk space, such as Figure 5 As shown, the following problems occur in the process of satellite image mosaic data processing: 1. The intermediate data brings a lot of IO overhead and storage space; 2. There is no good preview mechanism to cause repeated processing; 3. The complex processing process cannot satisfy Emergency scene

[0003] At present, in order to improve the efficiency of image processing, a graphics processing unit (GPU) is commonly used to accelerate the computing power. In the remote sensing image GPU parallel processing technology and implementation method, the non-graphics general-purpose calculation of the GPU requires the use of graphics APIs similar to DirectX or OpenGL. GPU to communicate, this programming method is too complicated, and can not effectively use the computing performance of the GPU; in the analysis and application of GPU parallel computing technology, in 2006, the release of CUDA made NVIDA's GPU a representative of GPU high-performance computing, however, CUDA can only support NVIDIA GPUs; research on the design of remote sensing imagery algorithms based on OpenCL shows that using OpenCL as a development platform can take advantage of GPU computing performance

Images