Research Of Cloud Detection Algorithm For Remote Sensing Image Based On FPGA

During the impact of climate, it is hard to obtain completely clear images without cloud covering. The presence of clouds may affect the analysis and research of the images. If we discard the whole images that containing clouds, image resources will be wasted. Furthermore, in some southern areas, completely cloudless images cannot be obtained due to the rainy weather. Therefore, the cloud detection process plays an important role in the application of satellite resources. Recently, a lot of cloud detection algorithms that based on general computer software have been proposed. However, these methods cannot satisfy the requirement of real-time satellite system due to their low-speed. To meet the needs of both high efficiency and high accuracy, more and more image processing algorithms based on FPGA have been put forward.First, we analyzed existing cloud classification methods and then proposed a novel classification method which classifies clouds as thin clouds and thick clouds due to the needs of our research. Meanwhile, we verified the results of our classification method on Matlab. After that, we proposed a feature extraction algorithm to distinguish thick clouds and ground objects. These features are defined as gray, texture, edge and frequency domains. Moreover, we compared our feature extraction algorithm with existing similar algorithms and verified their performance.Second, we designed a cloud detection algorithm based on optical remote sensing images, which including select features from clouds and ground objects, calculate the adaptive threshold, mark connected component and minimum bounding rectangle. Meanwhile, we enumerated some examples to test the results of this algorithm.Then, we designed an FPGA-base system to separate land and waters. The algorithm fetches information from the database of land areas, calculates minimum bounding rectangle of polygon, judges the relationship between points and polygon, and determines the position of the images. Furthermore, we evaluated the simulation of each module.Finally, we designed a FPGA-base cloud detection algorithm in water images. It includes calculating the value of gray features and the adaptive threshold, binarizing images, and outputting rectangle coordinates. Then, we adopted Xilinx Vertex-6 ML605 board to implement our algorithm. According to the simulation results and on-board debugging, it shows that the function of cloud detection algorithm can reach the system requirements.