2023-07-202024-05-15https://scholars.lib.ntu.edu.tw/handle/123456789/663978水果為台灣重要的農業產品，2020年時水果年度產值已到兩百億美金，其中包含四十九億美金的出口金額。介殼蟲為台灣水果產業中重大害蟲之一。近年來中國、日本、澳洲等台灣水果進口國的海關，採取更嚴格的介殼蟲檢疫標準。因此我國產值較高的鳳梨、釋迦、檸檬等水果，也因介殼蟲害而減少外銷。而台灣在介殼蟲防疫上仍舊採取肉眼檢視，因此需要更自動的介殼蟲檢測方法。本研究持續上一期（2022年）「介殼蟲判釋及計數系統」研究計劃，持續提高系統的精確度。本期(2023年)將透過增將影像與標註，與利用更先進的卷積神經網路(YOLOv7)來提高介殼蟲判釋與計數系統的效能,以維持台灣水果出產的品質。 Fruits are essential agricultural products in Taiwan. In 2020, the annual revenue of fruit production reached $20.2 billion USD, with the export value accounted for $4.9 billion USD. Scale insects are major pests in Taiwan. In recent years, the customs of countries that import Taiwan fruits, such as China, Japan, and Australia, implement strict quarantine measures for fruit with scale insects. Thus, the export amount of high value fruits, such as pineapple, sugar apple, and lemon, are significantly reduced due to exceeded amounts of scale pests were measured. Nowadays, the screening of scale insects on fruits in Taiwan are still performed manually using naked-eye inspection. Manual inspection is time-consuming and can be inaccurate due to fatigue. Therefore, it is crucial to develop a more precise and efficient method for scale insect inspection. This project continues the project “the identification and counting system of scale insects” of last year (i.e., 2022). The major objective of this project is to improve the performance of the model trained last year by including more training images and labels and by using the-state-of-the-art convolutional neural network, YOLOv7. The trained model is expected to initialize the automation of scale insect detection and, thus, to increase the quantity of fruit export for Taiwan.介殼蟲害;機器視覺;卷積神經網路;Scale insect; Machine vision; Convolutional neural networks