Article Preview
Top1. Introduction
Most agricultural technologies in Taiwan are more traditional than other developed countries, and the areas for agriculture in Taiwan are considerably smaller than those countries. Therefore, Taiwan’s development of mechanization and automation technologies for agriculture has evolved slowly. Nonetheless, Taiwan is slowly entering into smart agriculture (Li, 2012). Agricultural mechanization and automation technologies are mostly applied to agricultural crops such as sweet potato, groundnut, and corn, or to automatic fertilization of the farm. Due to the limited agricultural technologies, some refined agricultural products such as fruits can only be harvested manually, and the limited agricultural technologies also lead to a lack of workforce in the harvest season.
The wide scope of smart agriculture has led to an increase in research on recognizing object mechanisms. While object recognition fulfills most smart agriculture requirements, the design of accurate real-time recognition is still an open issue due to the challenges of characterizing an accurate and imaging sensitivity effect commonly encountered in the image recognition of cloud-based services.
With the rapid development of information technologies, many developed countries have gradually implemented automotive robots for agricultural development, and they have even implemented some aspects of artificial intelligence (AI) for farming and harvesting.
In recent years, because of the impact of industrial structure transformation, the Council of Agriculture, Executive Yuan indicated agricultural population in Taiwan fell from 2,852,588 in 2012 to 2,690,000 in 2019 (Council of Agriculture, 2021, https://www.ey.gov.tw/state/CD050F4E4007084B/0ededcaf-8d80-428e-96b7-7c24feb4ea0d). According to the above information, the agricultural population has declined in Taiwan because the young generation is not willing to engage in agricultural work; they prefer to work in the high-tech industry and the service industry. Therefore, because of the limited workforce resources, farmers cannot efficiently balance agricultural product volume and workforce resources. Due to the agricultural problems, we proposed the problems can be solved by AI technologies, and to design a more effective solution, we adopted the cloud-based service method. Therefore, this study implements recognition service methods to solve agricultural problems by using artificial intelligence technologies to calculate fruit quantity and help farmers achieve their harvest goals more easily. Farmers can also avoid excessive labor costs and recruit suitable assistants to harvest fruit during the COVID-19 epidemic. According to Taiwan CDC regulations, if people work outside, they should wear a mask and socially distance themselves from others. Hence, the number of workers to be recruited for harvesting is a critical issue during the COVID-19 epidemic. In Taiwan, the most important agricultural problem needing to be solved in time is the seasonal lack of agricultural workforce. The same agricultural products need to be harvested in the same season; therefore, the demand for an agricultural workforce is concentrated in the same harvest season, and the agricultural products may be overproduced. Therefore, the agricultural workforce is often in short supply.
Many prior studies proposed a large number of algorithms and technologies to calculate the quantity of the target object in the images. When we developed a solution using a cloud-based service, we considered the feasibility methods, including watershed segmentation, histogram of oriented gradient (HOG), convolutional neural network (CNN), K-means, and support vector machine (SVM) (Bartell et al., 2017; Esteva et al., 2017; Leachman and Merlino, 2017; Lv, 2019; Maldonado Jr et al., 2016; Malik et al., 2018). After considering the successful methods for calculating the fruit quantity or something else in prior studies, we adopted the most suitable method, Hybrid-based CNN, for solving the problem of the seasonal lack of agricultural workforce because of its ability to detect different objects within a short period.