Prediction of Soil Total Nitrogen Content based on Machine Vision

10.6919/ICJE.202205_8(5).0015

Yi Chen；Sichao Lin；Qingying Gao

Soil Total Nitrogen ； Machine Vision ； Soil Color ； Quantification

International Core Journal of Engineering

8卷5期（2022 / 05 / 01）

115 - 122

英文

The traditional analysis of soil total nitrogen content - Kjeldahl nitrogen analyzer is time-consuming and laborious. How to find a convenient and fast method to predict soil total nitrogen content is becoming more and more crucial to precision agriculture. This study proposed a new, convenient and affordable method to predict soil total nitrogen content. In this study, smart phones and customized darkrooms were used to obtain soil sample images, Python and OpenCV were used to process the images and obtain soil RGB three-channel values, and kjeldahl nitrogen analyzer was used to obtain soil total nitrogen content data. After quantitative analysis, a rapid prediction model of soil total nitrogen content based on machine vision was finally established. The results showed that there was a significant negative correlation between soil total nitrogen content and RGB three-channel values, and the highest correlation between total nitrogen content and RGB value was R (Red channel) (correlation coefficient -0.82). Although the model built in this study has not yet realized the prediction of total nitrogen content for all types of soil, the research results of soil samples in this experiment show that the model established by using smart phones to obtain soil RGB three-channel values and processing graphics by machine vision has the potential to achieve rapid prediction of total nitrogen content in soil.

1. Yassine Bouslihim et al. "Machine Learning Approaches for the Prediction of Soil aggregate Stability", 2011.
   連結：
2. Lei Zhang et al. A Self-training semi-supervised machine learning Method for Predictive Mapping of soil classes Limited Sample Data, 2020.
   連結：
3. Du Shangjia, Wu Haixia, Chen Guode, Fu Rong & Chen Yu.(2021). Soil nutrient characteristics and fertility quality evaluation of secondary forest community in Wuzhishan. Chinese Journal of Tropical Agriculture (12),18-24. Doi.
4. Wang Xin, Xiang Zhao, Li Ce, Zhu Jing, Wu Jing, Zhang Mingjie... Optimization of automatic Kjeldahl nitrogen analyzer for determination of soil total nitrogen content. Journal of Shandong Agricultural University (Natural Science)(03),438-440+446. Doi.
5. Wu CAIwu et al. Prediction of soil organic matter based on color extraction from digital image. 2019.
6. Gelder BK,ANex RP,Kaspar TC,et al.Soil Science Society of America Journal,2011,75(5):1821.
7. Liu Chao, Yuan Man, Zhuang Wenhua, Deng Dongzhou, & Zhang Tao. (2015). Quantitative relationship between soil color and organic matter in Beichuan Mountain based on photographic RGB value extraction. Chinese Science and Technology, 10(9), 5.
8. Kilrillova NP,Kemp DB,Artemyeva ZS.European Journal of Soil Science,2017,68(4):420.
9. soriano-disla JM,janik LJ,Viscarra Rossel RA, et al. Applied Spectroscopy,2014,49(2):139.
10. Li YUqing, Chen Ling, & Zhao Jianfu. (2006). Comparison of soil total nitrogen determination methods. Guangzhou Environmental Science, 21(3), 2.
11. Sun Ning,Chang qing-rui,Liu meng-yun, et al. Journal of Northwest Forestry University, 2011,26(1):56.
12. Weichao Wang, Wei Yang, Yulu Cui, Peng Zhou, Dong Wang & Minzan Li.(2021). Prediction of Soil Total Nitrogen Content based on CatBoost algorithm and Map Feature Fusion. Transactions of the Chinese Society for Agricultural Machinery (S1),316-322. Doi.
13. Rossel, R. A. V. , Walter, C. , & Fouad, Y. (2003). Assessment of two reflectance techniques for the quantification of field soil organic carbon.