Article Preview
Top1. Introduction
Citrus is the main cash crop in Chongqing, China. According to the official statistics, the total citrus planting area in Chongqing reached 198, 200 (hm2) and the yield was 2680, 000 (t) in 2018 (Zhao, 2019). As Chongqing is a mountainous region and most of citrus orchards located in hills, citrus production management is crucial for the quantity and quality of citrus yield. Domain knowledge plays a critical role in guiding the production management for citrus orchards that have complex soil and terrain conditions (Wang et al., 2016). In recent years, information technology (IT)-based systems have been widely studied in agricultural domain, with the aim to improve the efficiency of filed management (Wang et al., 2015; Santos et al., 2019). Then, a number of automation systems have been introduced including automatic irrigation and fertilization machines, and automatic growing and harvest machines for saving labor and boosting agricultural production efficiency (Partel et al., 2019; Sulistyo et al., 2017; Berenstein & Edan, 2017).
The main shortcoming of these agricultural automation systems is lack of sufficient domain knowledge, which is critical essential in decision support systems for realizing high-quality crop production management. In other words, agricultural systems are knowledge-intensive IT systems that need complex and even cross-area domain knowledge to help local farmers, who usually have limited domain knowledge, to make reliable decisions. Due to their knowledge modeling and reasoning capabilities, semantic technologies, including a set of semantic standards and ontology models, have achieved successes in many agricultural fields (Santos et al., 2019; Haverkort &Top, 2011; Beck et al., 2009). For example, Haverkort and Top (2011) created a potato ontology, a controlled vocabulary of the potato domain, to support automated decision making and data exchange. Beck et al. (2009) modeled soil, water, and nutrients for citrus and sugarcane based on ontology and implemented an ontology-based simulation environment. Wang et al. (2015) have developed an ontology-based application that realized fertilization, nutrition-related disease diagnosis, and water monitoring for citrus orchards. However, ontologies used in these systems were simple and at the vocabulary level (Vrandečić, 2004), which were insufficient for making a complicated decision.
With respect to the issue of automatic fertilization in citrus planting, however, the relevant semantic-based decision systems do not exist. That is to say, applying the technique of ontology to transparently and efficiently generate helpful decisions to carry out corresponding fertilization activities in citrus cultivation has not been found in the published literature. Then, we present an ontology-based fertilization system for citrus orchards, and it currently focus on the orchards located in mountainous regions of Chongqing. To the best of our knowledge, current fertilization systems can only suggest quantities of pure nitrogen (N), phosphorus (P), and potassium (K) for citrus trees, but fail to provide specific fertilizers for citrus trees planted in different orchards with various terrain and soil conditions.
In this paper, we create a citrus fertilization ontology (CFO) for citrus planting, which contains 103 classes, 34 properties, 800 instances, and 3056 resource description framework (RDF) triples. After that, we extend the CFO with Bayesian network (BN) to classify citrus trees into specific type of nutrition-related diseases (NRDs) according to their symptoms, to benefit making proper fertilization decisions in different scenarios. Finally, we develop a fertilization system to calculate the quantities of specific types of fertilizers at different growth stages of citrus trees in citrus planting.
The rest of the paper is organized as follows: in Section 2, the related work is discussed; Section 3 describes the citrus fertilization system including the design of the ontology and the fertilization management based on the CFO; Section 4 evaluates our system and presents a prototype fertilization machine. At last, we conclude the paper and outline the future work in Section 5.