東華大學圖書館 |

Language: English

Help

回圖書館首頁

手機版館藏查詢

Back

Switch To: Labeled | MARC Mode | ISBD

Neural network-based crop growth mod...

Koller, Michal.

Linked to FindBook

Google Book

Amazon

博客來

Neural network-based crop growth model to predict processing tomato yield on a site-specific basis using remotely sensed data.

Record Type:	Language materials, printed : Monograph/item
Title/Author:	Neural network-based crop growth model to predict processing tomato yield on a site-specific basis using remotely sensed data./
Author:	Koller, Michal.
Description:	144 p.
Notes:	Source: Dissertation Abstracts International, Volume: 64-03, Section: B, page: 1351.
Contained By:	Dissertation Abstracts International64-03B.
Subject:	Agriculture, Food Science and Technology. -
Online resource:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3082553

Neural network-based crop growth model to predict processing tomato yield on a site-specific basis using remotely sensed data.
Koller, Michal.

Neural network-based crop growth model to predict processing tomato yield on a site-specific basis using remotely sensed data. - 144 p.

Source: Dissertation Abstracts International, Volume: 64-03, Section: B, page: 1351.

Thesis (Ph.D.)--University of California, Davis, 2003.

Remote sensing is one of the major data acquisition tools available to rapidly acquire soil and plant related information over a wide area for use in precision agriculture. Green canopy has very specific reflectance characteristics distinguishing it from other materials such as soil and dry vegetative matter. Reflectance values in red (R) and near infra-red (NIR) spectral bands have been widely used for calculating normalized difference vegetation index (NDVI). Many researchers have related NDVI values to plant vigor, water stress, leaf area index (LAI) and/or yield. However, vegetative indices such as NDVI are usually sensitive to background reflectance characteristics. Often soil adjusted vegetation indices (SAVI) are used to minimize the background effect. In this study we have developed a relationship between the processing tomato yield and SAVI based on the R and NIR reflectance. Eight three band (R, NIR and green) aerial images were obtained at approximately two-week intervals during the 2000 processing tomato growing season. These images were analyzed to obtain SAVI values which were in turn related to LAI using regression techniques. A tuned neural network was developed to predict daily LAI values based on the biweekly experimental LAI values derived from aerial images. The coefficients of multiple determination between the actual LAI and neural network predicted LAI values were greater than 0.96 for all 56 grid points. The LAI values were numerically integrated over the whole growing season to obtain cumulative leaf area index days (CLAID). The CLAID values predicted from the neural network correlated very well with experimentally derived CLAID values (coefficient of determination, r<super>2</super> = 0.83) indicating that the neural network model simulated processing tomato growth well. A crop growth model that was capable of predicting crop yield based on neural network predicted LAI values and CIMIS weather data was developed. Although predicted yield tended to be low where the true yield was low, the coefficient of determination between predicted and experimental yield was poor on a grid point by grid point basis. However, the correlation coefficient improved when a classification technique was used to classify the yield into five (r<super>2</super> = 0.53) or nine zones (r<super>2</super> = 0.58). This research clearly shows that aerial images have a great potential in predicting crop yield, if they are used with sound analytical models and properly took into account relevant weather data.Subjects--Topical Terms:

1017813
Agriculture, Food Science and Technology.

Neural network-based crop growth model to predict processing tomato yield on a site-specific basis using remotely sensed data.
LDR:03393nam 2200253 a 45 001 937609
005 20110511
008 110511s2003 eng d
035 $a (UnM)AAI3082553
035 $a AAI3082553
040 $a UnM $c UnM
100 1 $a Koller, Michal. $3 1261471
245 1 0 $a Neural network-based crop growth model to predict processing tomato yield on a site-specific basis using remotely sensed data.
300 $a 144 p.
500 $a Source: Dissertation Abstracts International, Volume: 64-03, Section: B, page: 1351.
502 $a Thesis (Ph.D.)--University of California, Davis, 2003.
520 $a Remote sensing is one of the major data acquisition tools available to rapidly acquire soil and plant related information over a wide area for use in precision agriculture. Green canopy has very specific reflectance characteristics distinguishing it from other materials such as soil and dry vegetative matter. Reflectance values in red (R) and near infra-red (NIR) spectral bands have been widely used for calculating normalized difference vegetation index (NDVI). Many researchers have related NDVI values to plant vigor, water stress, leaf area index (LAI) and/or yield. However, vegetative indices such as NDVI are usually sensitive to background reflectance characteristics. Often soil adjusted vegetation indices (SAVI) are used to minimize the background effect. In this study we have developed a relationship between the processing tomato yield and SAVI based on the R and NIR reflectance. Eight three band (R, NIR and green) aerial images were obtained at approximately two-week intervals during the 2000 processing tomato growing season. These images were analyzed to obtain SAVI values which were in turn related to LAI using regression techniques. A tuned neural network was developed to predict daily LAI values based on the biweekly experimental LAI values derived from aerial images. The coefficients of multiple determination between the actual LAI and neural network predicted LAI values were greater than 0.96 for all 56 grid points. The LAI values were numerically integrated over the whole growing season to obtain cumulative leaf area index days (CLAID). The CLAID values predicted from the neural network correlated very well with experimentally derived CLAID values (coefficient of determination, r<super>2</super> = 0.83) indicating that the neural network model simulated processing tomato growth well. A crop growth model that was capable of predicting crop yield based on neural network predicted LAI values and CIMIS weather data was developed. Although predicted yield tended to be low where the true yield was low, the coefficient of determination between predicted and experimental yield was poor on a grid point by grid point basis. However, the correlation coefficient improved when a classification technique was used to classify the yield into five (r<super>2</super> = 0.53) or nine zones (r<super>2</super> = 0.58). This research clearly shows that aerial images have a great potential in predicting crop yield, if they are used with sound analytical models and properly took into account relevant weather data.
590 $a School code: 0029.
650 4 $a Agriculture, Food Science and Technology. $3 1017813
650 4 $a Engineering, Agricultural. $3 1019504
650 4 $a Remote Sensing. $3 1018559
690 $a 0359
690 $a 0539
690 $a 0799
710 2 0 $a University of California, Davis. $3 1018682
773 0 $t Dissertation Abstracts International $g 64-03B.
790 $a 0029
791 $a Ph.D.
792 $a 2003
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3082553