Application of BP neural network in predicting winter wheat yield based on thermography technology

Publication Type

Journal contribution (peer reviewed)

Authors

Hu, Z; Zhang, L; Wang, Y; Zia, S; Zeng, A; Song, J; Liu, Y; Spreeer, W; Müller, J; He, X.

Year of publication

2013

Published in

Spectroscopy and Spectral Analysis

Band/Volume

33/6

DOI

１０．３９６&#

Page (from - to)

1587-1592

Keywords

BP neural networks, ICWSI, Thermal camera, Winterweizen, Winter wheat yield

Ｕｓｉｎｇ　ａ　ｔｈｅｒｍａｌ　ｃａｍｅｒａ　ｔｏ　ｏｂｔａｉｎ　ｃａｎｏｐｙ　ｔｅｍｐｅｒａｔｕｒｅｓ　ｆｏｒ　ｗｉｎｔｅｒ　ｗｈｅａｔ，ａｎ　ｉｎｆｒａｒｅｄ　ｃｒｏｐ　ｗａｔｅｒ
ｓｔｒｅｓｓ　ｉｎｄｅｘ（ＩＣＷＳＩ）ｗａｓ　ｃａｌｃｕｌａｔｅｄ　ｉｎ　ｔｈｅ　ｍａｉｎ　ｗａｔｅｒ－ｒｅｑｕｉｒｅｍｅｎｔ　ｓｔａｇｅ．Ｔｈｅ　ｐｅｒｆｏｒｍａｎｃｅ　ｏｆ　ａ　ＢＰ ｎｅｕｒａｌ　ｎｅｔｗｏｒｋ ｗａｓ　ｔｅｓｔｅｄ　ｗｉｔｈ　ＩＣＷＳＩ　ｖａｌｕｅｓ　ｆｏｒ　ｔｈｒｅｅ　ｄｉｆｆｅｒｅｎｔ　ｐｅｒｉｏｄｓ　ｉｎ　ｏｎｅ　ｉｒｒｉｇａｔｉｏｎ　ｃｉｒｃｌｅ　ａｓ　ｉｎｄｅｐｅｎｄｅｎｔ　ｉｎｐｕｔ　ｆａｃｔｏｒｓ　ａｎｄ　ｏｂｓｅｒｖｅｄ　ｗｉｎｔｅｒ　ｗｈｅａｔ　ｙｉｅｌｄ　ａｆｔｅｒ　ｈａｒｖｅｓｔ　ａｓ　ｔｈｅ　ｏｕｔｐｕｔ．Ｔｈｅ　ｔｏｐｏｌｏｇｙ　ｏｆ　ｔｈｅ　ｎｅｕｒａｌ　ｎｅｔｗｏｒｋ　ｗａｓ　３－５－１，ａｎｄ　ａｆｔｅｒ　ｄａｔａ　ｎｏｒｍａｌｉｚａｔｉｏｎ，ｃｏｎｖｅｒｇｅｎｃｅ　ｐｅｒｆｏｒｍａｎｃｅ　ｗａｓ　ｅｎｈａｎｃｅｄ．Ｒｅｓｕｌｔｓ　ｓｈｏｗｅｄ　ｔｈａｔ　ｔｈｅ　ｍａｘｉｍｕｍ ｒｅｌａｔｉｖｅ　ｅｒｒｏｒ　ｗａｓ　ｏｎｌｙ　３．４２％．Ｔｏ　ｃｏｎｆｉｒｍ　ｔｈｅ　ｓｕｐｅｒｉｏｒｉｔｙ　ｏｆ　ｔｈｉｓ　ｍｅｔｈｏｄ，ａ　ｃｏｍｍｏｎ　ｎｏｎｌｉｎｅａｒ　ｒｅｇｒｅｓｓｉｏｎｍｏｄｅｌ　ｗａｓ　ａｌｓｏ　ｂｕｉｌｔ　ｔｏ　ｃｏｍｐａｒｅ　ｔｈｅ　ｐｒｅｄｉｃｔｉｏｎｓ　ｗｉｔｈ　ＩＣＷＳＩ　ｖａｌｕｅｓ　ａｎｄ　ｔｈｅ　ｏｂｓｅｒｖｅｄ　ｙｉｅｌｄ　ｏｆ　ｗｉｎｔｅｒ　ｗｈｅａｔ，ｂｕｔ ｔｈｅ　ｍａｘｉｍｕｍ　ｒｅｌａｔｉｖｅ　ｅｒｒｏｒ　ｏｆ　ｔｈｉｓ　ｍｏｄｅｌ　ｗａｓ　ｈｉｇｈｅｒ（１８．８７％）．Ｃｏｍｐａｒｉｓｏｎ　ｂｅｔｗｅｅｎ　ｔｈｅｓｅ　ｔｗｏ　ｍａｔｈｅｍａｔｉｃａｌ ｍｅｔｈｏｄｓ　ｓｈｏｗｓ　ｔｈａｔ　ｔｈｅ　ａｐｐｒｏａｃｈ　ｏｆ　ｃｏｍｂｉｎｉｎｇ　ｔｈｅｒｍａｌ　ｃａｍｅｒａ　ｔｅｃｈｎｏｌｏｇｙ　ｗｉｔｈ　ａ　ＢＰ　ｎｅｕｒａｌ　ｎｅｔｗｏｒｋ　ｐｒｅｄｉｃｔｉｏｎ
ｍｏｄｅｌ，ｗｈｉｃｈ　ｉｓ　ｍｏｒｅ　ｐｒｅｃｉｓｅ　ｆｏｒ　ｎｏｎｌｉｎｅａｒ　ｐｒｅｄｉｃｔｉｏｎ，ｗａｓ　ｓｕｆｆｉｃｉｅｎｔｌｙ　ｂｅｔｔｅｒ　ｔｈａｎ　ｏｔｈｅｒ　ｍｏｄｅｌｓ　ｔｏ　ｐｒｅｄｉｃｔ　ｔｈｅ ｗｉｎｔｅｒ　ｗｈｅａｔ　ｙｉｅｌｄ　ｓｕｃｃｅｓｓｆｕｌｌｙ　ａｎｄ　ａｃｃｕｒａｔｅ　ｅｎｏｕｇｈ　ｔｏ　ｍｅｅｔ　ｐｒｏｄｕｃｔｉｏｎ　ｒｅｑｕｉｒｅｍｅｎｔｓ．