Font Size: a A A

Research On Thick Cloud Removal In The Remote Sensing Image

Posted on:2015-02-17Degree:MasterType:Thesis
Country:ChinaCandidate:Z WangFull Text:PDF
GTID:2268330431958413Subject:Pattern Recognition and Intelligent Systems
Abstract/Summary:PDF Full Text Request
The key technology of remote sensing image to the cloud is cloud and its shadow detection technique.Normal cloud processing algorithm varies with the pad under the surface of the cloud. For the image with thick clouds, using the resolution remote sensing images of the same area at different phases in Landsat ETM and ETM remote sensing image data of the same cycle phase or nearly the same season in different years, according to the relative change in spectral characteristics of each band, the authors designed a thick cloud and its shadow geographical segments enhanced model,and combining with the model and the conventional classification model,using linear regression analysis of image matching and replacing pixel operations, proposed the approach to eliminate or reduce the cloud area impact in Landsat ETM remote sensing image data.The experimental data Institute used are from the High Resolution Satellite Imagery of Zhuhai shoot by Landsat ETM, the cloud images and without cloud images acquisition time was September14,2000and November1,2000.Among the object in the experiment, the DN value of cloud, the buildings underlying surface of the cloud, sand, etc. is coincident, but cloud, cloud shadows and water bodies and other shadow DN values are substantially coincident, so we can not use traditional threshold segmentation method to extract cloud and its shadow. After the analysis of more traditional thresholding methods, the paper proposed image fusion method based on wavelet transform, use the integrity of the ETM1blue channel image cloud region and the significantly difference ETM4near infrared channel image cloud and the underlying surface, its wavelet fusion, enhanced image contrast cloud region and the underlying surface. Then use ETM2, ETM5, ETM7three-band Synthetic grayscale image threshold segmentation to extract the cloud shaded area. Finally, using the radius of3,5,7planar disc structure of the expansion image accurately to invert cloud, cloud shaded area, and the optimum coefficient of expansion. The original algorithm is capable of making thick cloud area in remote sensing image completely replaced by a cloudless regions, basically eliminating interference of the building region and the shaded area, increasing the availability of the original image information, but also maximizing the retention of the original image data, takes good use of the advantages of each channel local area features. Compared to the previous algorithm, this algorithm solves the problem of the underlying surface coincides with DN value of buildings, and the misjudgment situation of the, sand, water, etc, greatly reduces the consumption of system operation runtime, speeds up processing cloudy detection is a simple, efficient algorithm.
Keywords/Search Tags:Remote sensing image, Cloud Removal, Pixel replacement, Wavelet transform, Weighted fusion, The disc structure
PDF Full Text Request
Related items