東華大學圖書館 |

語系: 繁體中文

說明(常見問題)

回圖書館首頁

手機版館藏查詢

登入

回首頁

切換: 標籤 | MARC模式 | ISBD

Tackling uncertainties and errors in...

Song, Kuan.

FindBook

Google Book

Amazon

博客來

Tackling uncertainties and errors in the satellite monitoring of forest cover change.

紀錄類型:	書目-語言資料,印刷品 : Monograph/item
正題名/作者:	Tackling uncertainties and errors in the satellite monitoring of forest cover change./
作者:	Song, Kuan.
面頁冊數:	187 p.
附註:	Source: Dissertation Abstracts International, Volume: 71-08, Section: A, page: 2994.
Contained By:	Dissertation Abstracts International71-08A.
標題:	Geography. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3409971
ISBN:	9781124078786

Tackling uncertainties and errors in the satellite monitoring of forest cover change.
Song, Kuan.

Tackling uncertainties and errors in the satellite monitoring of forest cover change. - 187 p.

Source: Dissertation Abstracts International, Volume: 71-08, Section: A, page: 2994.

Thesis (Ph.D.)--University of Maryland, College Park, 2010.

This study aims at improving the reliability of automatic forest change detection. Forest change detection is of vital importance for understanding global land cover as well as the carbon cycle. Remote sensing and machine learning have been widely adopted for such studies with increasing degrees of success. However, contemporary global studies still suffer from lower-than-satisfactory accuracies and robustness problems whose causes were largely unknown.

ISBN: 9781124078786Subjects--Topical Terms:

524010
Geography.

Tackling uncertainties and errors in the satellite monitoring of forest cover change.
LDR:03750nam 2200373 4500 001 1393869
005 20110415112005.5
008 130515s2010 ||||||||||||||||| ||eng d
020 $a 9781124078786
035 $a (UMI)AAI3409971
035 $a AAI3409971
040 $a UMI $c UMI
100 1 $a Song, Kuan. $3 1672448
245 1 0 $a Tackling uncertainties and errors in the satellite monitoring of forest cover change.
300 $a 187 p.
500 $a Source: Dissertation Abstracts International, Volume: 71-08, Section: A, page: 2994.
500 $a Adviser: John R. G. Townshend.
502 $a Thesis (Ph.D.)--University of Maryland, College Park, 2010.
520 $a This study aims at improving the reliability of automatic forest change detection. Forest change detection is of vital importance for understanding global land cover as well as the carbon cycle. Remote sensing and machine learning have been widely adopted for such studies with increasing degrees of success. However, contemporary global studies still suffer from lower-than-satisfactory accuracies and robustness problems whose causes were largely unknown.
520 $a Global geographical observations are complex, as a result of the hidden interweaving geographical processes. Is it possible that some geographical complexities were not expected in contemporary machine learning? Could they cause uncertainties and errors when contemporary machine learning theories are applied for remote sensing?
520 $a This dissertation adopts the philosophy of error elimination. We start by explaining the mathematical origins of possible geographic uncertainties and errors in chapter two. Uncertainties are unavoidable but might be mitigated. Errors are hidden but might be found and corrected. Then in chapter three, experiments are specifically designed to assess whether or not the contemporary machine learning theories can handle these geographic uncertainties and errors. In chapter four, we identify an unreported systemic error source: the proportion distribution of classes in the training set. A subsequent Bayesian Optimal solution is designed to combine Support Vector Machine and Maximum Likelihood. Finally, in chapter five, we demonstrate how this type of error is widespread not just in classification algorithms, but also embedded in the conceptual definition of geographic classes before classification. In chapter six, the sources of errors and uncertainties and their solutions are summarized, with theoretical implications for future studies.
520 $a The most important finding is, how we design a classification largely pre-determines the "scientific conclusions" we eventually get from the classification of geographical observations. This happened to many contemporary popular classifiers including various neural nets, decision tree, and support vector machine. This is a cause of the so-called overfitting problem in contemporary machine learning. Therefore, we propose that the emphasis of classification work be shifted to the planning stage before the actual classification. Geography should not just be the analysis of collected observations, but also about the planning of observation collection. This is where geography, machine learning, and survey statistics meet.
590 $a School code: 0117.
650 4 $a Geography. $3 524010
650 4 $a Statistics. $3 517247
650 4 $a Remote Sensing. $3 1018559
690 $a 0366
690 $a 0463
690 $a 0799
710 2 $a University of Maryland, College Park. $b Geography. $3 1669229
773 0 $t Dissertation Abstracts International $g 71-08A.
790 1 0 $a Townshend, John R. G., $e advisor
790 1 0 $a Goward, Samuel $e committee member
790 1 0 $a Liang, Shunlin $e committee member
790 1 0 $a Huang, Chengquan $e committee member
790 1 0 $a Jaja, Joseph $e committee member
790 $a 0117
791 $a Ph.D.
792 $a 2010
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3409971