Quantile Regression-based Change Detection Using Landsat Analysis Ready Data
Author | : Xiaoyu Liang |
Publisher | : |
Total Pages | : |
Release | : 2020 |
ISBN-10 | : OCLC:1280067552 |
ISBN-13 | : |
Rating | : 4/5 ( Downloads) |
Download or read book Quantile Regression-based Change Detection Using Landsat Analysis Ready Data written by Xiaoyu Liang and published by . This book was released on 2020 with total page pages. Available in PDF, EPUB and Kindle. Book excerpt: Land surface is experiencing human interference of unprecedented frequency and intensity, which impairs sustainability. Continuously monitoring the land surface and timely detecting the land cover changes become the key to understand the human-nature interaction and balance the relationship between social development and natural resources. Many Landsat time series-based change detection methods have been developed to capture fine-scale human-induced changes with the temporal accuracy as high as sub-annual level. However, time series dominated the field of continuous change detection while spatial information and spatial-temporal modelling for change detection have not been fully exploited. This research developed a quantile regression-based change detection method that incorporated both spatial information and time series to enhance the change detection performance. The algorithm was verified in central Worcester which experienced intensive human activities at the beginning of 2000s. Comparing with the pixel-based state-of-art, the proposed method reached an acceptable accuracy with 88% F1 score and 16 correctly detected changes out of 25 in total. It had the lowest temporal Root Mean Square Error (RMSE) with 10.9 days, proving its ability of early detection. Besides, it showed the best result compared to the competitors without extra cloud filter, which indicated greater robustness to outliers. The proposed method is one of the first methods that have used spatial-temporal modelling in the continuous change detection framework. Future work will be focused on improving the model development, enhancing the computational efficiency and proving the generalization.