CAJ | 학술논문

随着卫星遥感数据的增加，单机环境下完成海量影像的Sobel边缘检测将比较耗时，如何提高效率已成为时下遥感领域所面临的一个新的挑战。传统方法通过获得遥感影像的数学形态变化特征（即二值化图像），进而提取其边缘二值化影像。鉴于MapReduce是运行在Hadoop集群环境下的并行分布式计算模型，在处理海量数据时的效率之高，故本文将其应用到遥感领域，并行实现海量影像的Sobel边缘检测。实验结果表明较传统的方法而言，集群环境下实现海量影像边缘检测的用时显著减少，并且该用时随着Hadoop集群中节点数目的增加而线性递减。
수착위성요감수거적증가，단궤배경하완성해량영상적Sobel변연검측장비교모시，여하제고효솔이성위시하요감영역소면림적일개신적도전。전통방법통과획득요감영상적수학형태변화특정（즉이치화도상），진이제취기변연이치화영상。감우MapReduce시운행재Hadoop집군배경하적병행분포식계산모형，재처리해량수거시적효솔지고，고본문장기응용도요감영역，병행실현해량영상적Sobel변연검측。실험결과표명교전통적방법이언，집군배경하실현해량영상변연검측적용시현저감소，병차해용시수착Hadoop집군중절점수목적증가이선성체감。
Along with the rapid increase of Remote Sensing satellite image data, the efficient sobel edge detection is becoming a new challenge for large volumes of image data in this field. The traditional method for remote sensing image edge detecting is done after getting its mathematical morphological variation characteristics (binary image). According to the characteristics of parallel distributed computing efficiently of MapReduce in Hadoop cluster environment and the advantages of conducting mass data in a relatively short time, the authors of the paper applied the MapReduce pattern to the field of Remote Sensing so that more images could be conducted quickly for Sobel image edge detecting. As a result, it turns out that compared with the traditional methods, we spent less time on sobel image edge detecting for massive Remoting Sensing images with the help of MapReduce running in Hadoop cluster environment, and that the more we increased the quantities of computer, the less time this operation took proportionally.