Using AI methods for solving agricultural problems in Taiwan via service design method － A case study of the lack of agricultural workforce
| Autor: | GUAN-YI WU, 吳冠億 |
|---|---|
| Rok vydání: | 2019 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 107 Because of the transformation of the industrial structure in Taiwan, fewer Taiwanese are willing to engage in agricultural work. Several issues related to agricultural difficulties, such as a lack of workforce and cost management, need to be resolved soon. Most of the farmers face these issues in Taiwan as well as in other countries around the world. Currently, most of the farmers face the problem of a lack of workforce during harvest season. To solve the agricultural problems, we studied the situation of agriculture in Taiwan by using a modified double diamond model of service design and implementing innovative information technologies with AI to support farmers in solving the abovementioned problem of a lack of agricultural workforce during harvest season. We implemented a convolutional neural network (CNN) method called RetinaNet to analyze an image taken in a watermelon field because of its balance of precision and detection speed. To train the CNN, we used a drone to take aerial pictures for our dataset. We used the confusion matrix to validate our service design solution, and the results of implementing RetinaNet for calculating the quantity of watermelons ready for harvest instead of having the farmers personally count and obtain this value, showed an average precision of up to 98.8% for watermelon recognition. Moreover, we used IV the F1-score to evaluate the effectiveness between precision and recall. The average precision and F1-score were used to determine whether the proposed model was effective. In this research, we purposed a process to solve an agricultural problem in Taiwan with artificial intelligence technologies based on a service design method, and helped farmers to calculate the volume of agricultural products in order to enable them to control the workforce cost and find more sales channels to sell their excessive agricultural products. In addition, in this research, we illustrated that RetinaNet obtained a good result with respect to the calculation of the quantity of watermelons ready for harvest. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|