The Implementation Of Spaceborne Real-Time Processorfor Target Recognition In Remote Sensing Image

Among the various space remote sensing technologies, using remote sensing image to present the final observation results is a relative intuitive way. Remote sensing image, obtained by different sensors, is the spectral data of perception objects. Depending on the wave length difference of the electromagnetic spectrum, it can be classified to optical image(visible light), microwave image(SAR), infrared image and hyperspectral image etc. In this paper, we designs a high performance spaceborne remote sensing image real-time processor, which innovatively proposes three state-of-the-art technologies: common feature extraction technology of the remote sensing image processing algorithm, cross-mapping software and hardware method for algorithm implementation, hardware engine chains focused on the neighborhood block processing algorithm in image.Current commonly-used remote sensing image processing algorithm required a large scale computation, but the characteristics of computation are regular. The paper proposes common feature extraction technology for the remote sensing image processing algorithm. The creative point is ultimately classifying the remote sensing image processing operation into the point operation, neighborhood block operation and global operation categories. Setting up a solid foundation for the following hardware architecture design.On the basis of the analysis of algorithm, the paper deals with the architecture implementation. In order to cope with the fallacies in hardware performance, power, area in which traditional image processing algorithm implementation results. The thesis presents an algorithm architecture implementation method. To improve the efficiency of the hardware architecture, the paper proposes an algorithm hardware mapping method which is made up of software and hardware cross processing, and then the implementation of arithmetic logic operation, control architecture, memory architecture. The paper applied the hardware implementation method into the design of real-time remote sensing image processing architecture, from the theory of computer architecture, and proposed a typical image processing implementation method. Memory-centered remote sensing image real time processing architecture, the innovation is application of neighborhood image block processing algorithm engine chain. Finally, the architecture meets the real-time onboard processing(up to about 500 MIPS) and has a flexible algorithm adaptation ability.Finally, the paper applied memory-centered remote sensing image real time processing architecture into the a real-time processor. On the one hand, the paper did the image processing algorithm feature extraction. On the other hand, the algorithm optimization is carried out for the special node. Presented the specific logic implementation and scheduling tables of the neighborhood image block processing engine chain, and memory-centered remote sensing image real time processing architecture is verified. Eventually, after the test of the processor chip, the actual processing time, the processor power consumption, processor area index are given. It is proved that the image processing architecture can be used in various sizes, various resolutions, various modes and various sensors of the spaceborne image processing applications.