Preprocessing Method Of Multispectral Image With Zynq

The geostationary meteorological satellite is a kind of specialized remote sensing satellite.It plays a major role in all-weather meteorological observation.According to the satellite cloud imagery,the severe weather can be monitored and alerted to protect the people’s life and property safety.The multispectral remote sensing image data is huge.Therefore,compression and fusion pretreatment of multi-band infrared image before classification are of great practical significance.It can improve the resolution of cloud image,relieve transmission pressure and enhance the capability of night-time detection.By analyzing the characteristics of multi-spectral remote sensing images,it was found that the infrared cloud images of each band had certain relevance,low resolution and blurred edges.The K-L transform and the wavelet transform were applied to the preprocessing research based on data compression and fusion.Two kinds of transformation algorithms were realized by Matlab programming.Combined with the visual effect and quantitative indicators of the image,it was found that the K-L transform could effectively extract the cloud information,which was closest to the visible cloud image.The wavelet transform preserved the source image detailed information,and was suitable for the analysis of pretreated cloud structure.Therefore,K-L transform was more suitable for infrared cloud image preprocessing.The design was realized in Xilinx’s Zynq platform,using hardware and software co-design method to achieve the K-L transformation of multi-channel image data.In K-L IP core design,the C language programming was used to achieve K-L transform algorithm.The corresponding IP core was realized by the synthesis and optimization of the Vivado HLS tool.The RTL-level code converted by C was proved to be accurate by the C/RTL co-simulation verification.And finally the IP core packaging could be finished.The completed K-L IP core was verified in the Zedboard development board.Multi-spectral remote sensing image preprocessing system based on Zynq consisted of four parts: image acquisition,data caching,algorithm processing and pre-processed data transmission.Experimental results showed that perform a pre-processing operation which used K-L transform on three images with a resolution of 640 × 480,the PC software taked about 5000 milliseconds to do it,while Zynq hardware and software co-implementation only cost 91 milliseconds,in this means the speed was greatly improved.The multi-spectral remote sensing image pre-processing technology based on Zynq had many advantages such as real-time,high efficiency and convenience.So there were broad prospects for the research and application in the field of aerospace engineering.