Design And Implementation Of A Quick-View System For Remote Sensing Images With Different Quantization Bits

Remote sensing quick-view systems play an indispensable role in space exploration and Earth observation,enabling the rapid processing of raw data transmitted from spacecraft and the display of payload imaging quality and feature information.However,with the upgrade of optical remote sensing devices and the rapid growth of remote sensing data,the display efficiency of many remote sensing quick-view systems has become less efficient.Moreover,the post-processing of target detection and marking requires the processing of stored data,which further affects the timeliness of remote sensing target identification.To address these issues,this thesis focuses on the rapid display and target marking of remote sensing images with different quantization bits to improve the extraction efficiency of multi-source remote sensing data and the real-time display efficiency of remote sensing images,as well as the rapid target marking of specific areas in remote sensing images.The specific research contents are as follows:(1)To quickly parse remote sensing data with different quantization bits,a parallel data extraction algorithm based on CUDA acceleration was studied,and a hierarchical block organization structure for remote sensing images was designed.Using CUDA programming,the parallel extraction of remote sensing data with different quantization bits was carried out,effectively improving data parsing efficiency.At the same time,the image pyramid technique was used to sample and compress the original images,generating multi-level remote sensing images with different resolutions.The image block technique was then used to read and process remote sensing data in blocks,enhancing the organization and processing efficiency of the system for remote sensing data.(2)A GPU-accelerated remote sensing image rapid display scheme was investigated.A multi-threaded asynchronous remote sensing data scheduling strategy was adopted to reduce data loading blockage and improve loading efficiency.Simultaneously,using the interoperation technology of CUDA and OpenGL,the rapid rendering of remote sensing images was realized,significantly enhancing the display quality and speed of remote sensing images.Experimental results in different display modes show that the display efficiency of this technical scheme is more than twice that of traditional CPU processing schemes,while reducing the CPU resource occupancy rate.(3)An improved connected-domain target marking algorithm based on a single scan was developed.This algorithm utilizes the imaging characteristics of remote sensing images,allowing connected-domain marking with just one scan while also utilizing GPU parallel acceleration processing,enhancing the display efficiency of remote sensing images after target marking.Compared to traditional processing schemes,the display efficiency is increased by about three times.Using the research scheme in this thesis,a quick-view and target marking system for remote sensing images with different quantization bits was designed and implemented.This system significantly improves the rendering efficiency of remote sensing images and the preliminary target marking display effect,enhancing the user’s interactive experience.