Knowledge-based expert system is widely used in remote sensing research and
digital image processing. Also, based on the its advantages, knowledge-based expert
system performs very well in many agricultural image processing tasks such as
image segmentation, image thresholding, and image classification. In Romeo et al.
(2013), an automatic and robust expert system for greenness identification is proposed which consists of histogram analysis-based decision-making module and
greenness identification module based on classical methods and fuzzy clustering
approach. In Montalvo et al. (2013), an automatic expert system for weeds/crops
identification in images from maize fields is proposed which is able to identify
weeds/crops when they have been contaminated with materials coming from the soil,
due to artificial irrigation or natural rainfall.
6.5.2 Machine Learning–Based Decision Tree
Machine learning has many definitions. Copeland (2016) defines machine learning
as “the practice of using algorithms to parse data, learn from it, and then make a
determination or prediction about something in the world.” Ng (2013) gives a short
definition that machine learning is “the science of getting computers to act without
being explicitly programmed.” Various machine learning programs, libraries, and
frameworks are developed and used in digital image processing and in the computer
vision field (Pedregosa et al. 2011; Vedaldi and Fulkerson 2010; Collobert et al.
2011). As the result of the quick development of computer hardware and software in
the past few years, many computer vision problems that cannot be handled by
traditional image processing methods have been solved by machine learning
approaches (Jordan and Mitchell 2015; Sonka et al. 2014; Rosten and Drummond
2006). Meanwhile, machine learning is efficient and effective to automatically
discover intricate patterns and structures in agro-geoinformation data. A variety of
machine learning-based approach has been developed and applied to support agricultural applications and researches, such as cropland extent mapping (Teluguntla
et al. 2018), crop type classification (Hao et al. 2020), drought monitoring (Park et al.
2016), and agricultural sustainability (Sharma et al. 2020).
Based on the task type, machine learning could be divided into supervised
learning and unsupervised learning. The major difference between supervised and
unsupervised learning is the dataset adopted during the learning process where
supervised paradigm (e.g., classification, regression, recommendation system)
deals with the labeled data and unsupervised paradigm (e.g., clustering, outlier
detection, association rule mining) deals with the unlabeled data. Further, as
described in Bishop (2006), supervised learning solves the problems where “training
data comprises examples of the input vectors along with their corresponding target
vectors”; unsupervised learning aims to “discover groups of similar examples within
the data, where it is called clustering, or to determine the distribution of data within
the input space, known as density estimation, or to project the data from a highdimensional space down to two or three dimensions for the purpose of
92
C. Zhang and L. Lin
digital image processing. Also, based on the its advantages, knowledge-based expert
system performs very well in many agricultural image processing tasks such as
image segmentation, image thresholding, and image classification. In Romeo et al.
(2013), an automatic and robust expert system for greenness identification is proposed which consists of histogram analysis-based decision-making module and
greenness identification module based on classical methods and fuzzy clustering
approach. In Montalvo et al. (2013), an automatic expert system for weeds/crops
identification in images from maize fields is proposed which is able to identify
weeds/crops when they have been contaminated with materials coming from the soil,
due to artificial irrigation or natural rainfall.
6.5.2 Machine Learning–Based Decision Tree
Machine learning has many definitions. Copeland (2016) defines machine learning
as “the practice of using algorithms to parse data, learn from it, and then make a
determination or prediction about something in the world.” Ng (2013) gives a short
definition that machine learning is “the science of getting computers to act without
being explicitly programmed.” Various machine learning programs, libraries, and
frameworks are developed and used in digital image processing and in the computer
vision field (Pedregosa et al. 2011; Vedaldi and Fulkerson 2010; Collobert et al.
2011). As the result of the quick development of computer hardware and software in
the past few years, many computer vision problems that cannot be handled by
traditional image processing methods have been solved by machine learning
approaches (Jordan and Mitchell 2015; Sonka et al. 2014; Rosten and Drummond
2006). Meanwhile, machine learning is efficient and effective to automatically
discover intricate patterns and structures in agro-geoinformation data. A variety of
machine learning-based approach has been developed and applied to support agricultural applications and researches, such as cropland extent mapping (Teluguntla
et al. 2018), crop type classification (Hao et al. 2020), drought monitoring (Park et al.
2016), and agricultural sustainability (Sharma et al. 2020).
Based on the task type, machine learning could be divided into supervised
learning and unsupervised learning. The major difference between supervised and
unsupervised learning is the dataset adopted during the learning process where
supervised paradigm (e.g., classification, regression, recommendation system)
deals with the labeled data and unsupervised paradigm (e.g., clustering, outlier
detection, association rule mining) deals with the unlabeled data. Further, as
described in Bishop (2006), supervised learning solves the problems where “training
data comprises examples of the input vectors along with their corresponding target
vectors”; unsupervised learning aims to “discover groups of similar examples within
the data, where it is called clustering, or to determine the distribution of data within
the input space, known as density estimation, or to project the data from a highdimensional space down to two or three dimensions for the purpose of
92
C. Zhang and L. Lin