Terrain information is difficult to obtain due to cloud cover, these regions will become blind in remote sensing images, which led to the remote sensing images can not play its due value. So we need to use some detection technology to detect the cloud in remote sensing image to avoid a lot of useless images captured by the satellite wastes a large amount of storage space and transmission bandwidth. This paper relay on a delicate high-resolution earth observation satellite in synchronous orbit, combined the CNN with SVM for cloud detection, and finally designed a new algorithm which can be realized in remote sensing satellite video processing system to statistic the cloud coverage ratio in orbit. Firstly, the large image is cut into small pieces, which will be send into the algorithm module. The new algorithm uses the deep learning algorithm to automatically abstract the effective features, and after that SVM classifier is used to recognize the thick cloud. The CNN net is designed with a matlab toolbox Matconvnet, which is easy to use and provides many pre-trained model. The most important is that the CNN network is simple enough so that it can be realized in embedded system like Zynq platform. Final 1000 cloud subimages and 1000 no-cloud subimages are used as training dataset, and another 1000 cloud subimages and 1000 no-cloud subimages are used as testing dataset, and the illation labels are compared to the true labels, the result shows that this algorithm is much more accurate than traditional cloud detection such as threshold method or SVM. In the future, this new algorithm can be easily combined with dynamic partial reconfiguration of FPGA, then the algorithm codes can be upload from earth to the satellite and reconfigure the algorithm module in FPGA. So the satellite can advance with the times.
Cloud detection Deep learning SVM CNN
This is a preview of subscription content, log in to check access.
Sarkar, N., Chaudhuri, B.B.: An efficient differential box-counting approach to compute fractal dimension of image. IEEE Trans. Syst. Man Cybern. 24(1), 115–120 (1994)CrossRefGoogle Scholar
Zhirong, L., Mcodonnell, M.J.: Using dual thresholds for cloud and mosaicking. Int. J. Remote Sens. 7(10), 1349–1358 (1986)ADSCrossRefGoogle Scholar
Shin, D., Pollard, J.K.: Cloud detection from ATSR images using a segmentation techniques. SPIE 2578, 46–52 (1995)ADSGoogle Scholar
Saunders, R.W.: An automated scheme for the removal of cloud contamination from AVHRR radiances over western Europe. Int J. Remote Sens. 7(7), 867–886 (1986)ADSCrossRefGoogle Scholar
Zavody, A.M., Mutlow, C.T., Llewelyn-Jones, D.T.: Cloud clearing over the ocean in the processing of data from the along-track scanning radiometer (ATSR). J. Atmos. Ocean. Technol. 17(5), 595–615 (2000)ADSCrossRefGoogle Scholar
Liu, R., Liu, Y.: Generation of new cloud masks from MODIS land surface reflectance products. Remote Sens. Environ. 133(15), 21–37 (2013)ADSCrossRefGoogle Scholar
Sedano, F., Kempeneers, P., Strobl, P., et al.: A cloud mask methodology for high resolution remote sensing data combining information from high and medium resolution optical sensors. ISPRS J. Photogramm. Remote Sens. 66(5), 588–596 (2011)ADSCrossRefGoogle Scholar
Jedlovec, G.J., Haines, S.L., Lafontainef, J.: Spatial and temporal varying thresholds for cloud detection in GOES imagery. IEEE Trans. Geosci. Remote Sens. 46(6), 1705–1717 (2008)ADSCrossRefGoogle Scholar
Luo, Y., Trishchenko, A.P., Khlopenkov, K.V.: Developing clearsky, cloud and cloud shadow mask for producing clear-sky composites at 250-meter spatial resolution for the seven MODIS land bands over Canada and North America. Remote Sens. Environ. 112(12), 4167–4185 (2008)ADSCrossRefGoogle Scholar
Wilson, M.J., Oreopoulos, L.: Enhancing a simple MODIS cloud mask algorithm for the Landsat data continuity mission. IEEE Trans. Geosci. Remote Sens. 51(2), 723–731 (2013)ADSCrossRefGoogle Scholar
Oreopoulos, L., Wilson, M.J., Vhrnai, T.: Implementation on Landsat data of a simple cloud-mask algorithm developed for MODIS land bands. IEEE Geosci. Remote Sens. Lett. 8(4), 597–601 (2011)ADSCrossRefGoogle Scholar