智慧化甜菜夜蛾蟲蛹性別辨識系統之開發

Abstract

為了防範害蟲，化學農藥的過度使用已經導致環境污染、食品安全問題以及害蟲對農藥的抗藥性增加。因此，生物防治方法逐漸備受關注，生物防治利用天然生態平衡中的生物來對抗害蟲，符合永續農業原則。在台灣，甜菜夜蛾是一個重要的害蟲，對多種作物造成了嚴重損害，並且在氣候變遷和農藥濫用的情況下，其繁殖能力以及抗藥性提升，使得防治作業變得更加困難，過去曾研究發現SeNPV生物製劑對防治甜菜夜蛾極其有效，但是SeNPV的生產需要依靠大量甜菜夜蛾幼蟲作為載體，因此若要推廣SeNPV的使用並使其商品化，就必須建立一套甜菜夜蛾養殖系統提供大量且穩定的蟲源。本研究為了減少飼養成本並協助擴大飼養規模，針對飼養流程中人力與時間成本相對較高的流程進行優化和改良，建立一套自動化甜菜夜蛾蟲蛹性別辨識系統。此系統利用影像偵測系統，檢測和辨識蟲蛹的位置，配合三軸機台裝置收集蟲蛹，接著再次利用影像系統進行蟲蛹的性別分類，並根據結果使用負壓系統對蟲蛹進行分類。結果顯示上，在蟲蛹偵測的準確度為100%，而在吸取點的定位平均誤差為0.16 mm，蟲蛹性別辨識的準確度達95.33%，且各流程的成功率達95%以上，此外實驗測試蟲蛹經過各流程後的羽化率為100%，表示此系統不會傷害到蟲蛹。執行時間上相較於傳統的方法可節省1/3的時間，後續可利用此系統進行飼養規模的擴增，也可再整合其他養殖流程，使整套繁雜的甜菜夜蛾養殖流程成本降低，進而降低SeNPV病毒的生產成本。

Excessive use of chemical pesticides to eliminate pests has led to environmental pollution, food safety concerns, and pesticide resistance increases among pests. Biological control methods, which utilize natural ecological balances to deal with pest issues, are gaining attention when considering sustainable agriculture principles. In Taiwan, the beet armyworm is a significant pest causing severe damage to various crops. Its reproductive capacity and pesticide resistance have increased due to climate change and pesticide overuse, making control efforts more challenging. Previous studies have found that seNPV, a biopesticides, is highly effective in controlling the beet armyworm. However, the production of seNPV relies on a large number of beet armyworm larvae as carriers. To promote the use and commercialization of seNPV, a system must be established to provide a large and stable source of beet armyworm larvae. This study therefore aims to reduce the rearing costs of beet armyworm larvae and assist in expanding rearing scales by optimizing the rearing process, which was often labor and time-intensive in the past. An automated beet armyworm pupa gender identification system is developed, which uses an image detection system to detect and identify the location of a pupa, and a three-axis machine device to collect the pupa. Subsequently, the gender of the pupa is classified by the image system, and a vacuum system is employed to sort the pupae based on the gender classification results. The research results show a 100% accuracy in pupa detection, an average location error of 0.16 mm at the extraction points, a 95.33% accuracy in pupae gender identification, and a success rate of over 95% for each process. Additionally, a 100% emergence rate of pupae is found after each process is complete, indicating that the system does not harm the pupae. Compared to traditional methods, the execution time is reduced by one-third. This system can be used to expand rearing scales in the future and integrate other rearing processes, thereby lowering the cost of the complex beet armyworm rearing workflow and subsequently reducing the production cost of the seNPV virus.