A Fast Cloud Detection based on Improved 2D OTSU with Zynq SoC

Abstract

Cloud detection plays an important role in pre-processing of the remote sensing image to locate the potential area of the target fast and improve the detection efficiency. Although the cloud detection performance is not ideal in the complex ground, such as bright buildings on the ground, the OTSU based method is used to achieve fast cloud detection onboard for the single scenarios due to the simple calculation process. However, due to the sequential processing capabilities provided by the onboard processors, the processing complexity of OTSU increases significantly as the image size increasing. In this article, a fast cloud detection method based on the improved 2D OTSU with Zynq system on chip (SoC) is proposed. The improved 2D OTSU method is utilized for cloud detection to provide the capability to distinguish highlight buildings on the ground. And the parallel processing for key process in the 2D OTSU provided by the field programmable gate array in Zynq SoC to improve computing efficiency. Compared with the 1D OTSU, experimental results show that the proposed method could achieve high detection precision. Meanwhile, the processing time is about 0.18s for a 512×512 size image, which is a relative fast speed for the practical application.