隨著航空攝影技術的發展，農業影像辨識已成為精準農業的核心。然而，在連續 觀測作物生長週期的實務應用中，模型常面臨「領域增量學習 (Domain Incremental Learning， DIL)」的情境。農作物具備高度的「細粒度 (Fine-grained)」特性，例如不同 作物在特定生長期 (如玉米與花生的收割期) 具有極高的類間相似度，且同一作物在 不同階段亦存在巨大的類內變異度。傳統的微調 (Fine-tuning) 方法在學習新階段特 徵時，極易產生災難性遺忘 (Catastrophic Forgetting)；而基於歷史資料回放 (Replay) 的技術則受限於龐大影像的儲存成本與實務限制；完整重新訓練 (Retrain) 則會導致 時間成本呈非線性增長，無法滿足農業系統即時部署的需求。 為解決上述挑戰，本研究建立一套基於側掛式適應器 (ν-Adapter) 的「無回放 (Exemplar-Free)」DIL 框架，應用於細粒度農業航空影像辨識。本框架透過參數隔離 機制，僅需更新極少量的適應器參數 (約 0.13M 至 0.44M)，即可完整保留舊有任務 知識，從根本上避免災難性遺忘。此外，為克服農業影像的細粒度混淆問題，本研 究結合了監督式對比損失 (SupCon Loss) 與焦點損失 (Focal Loss) 之雙重損失函數， 有效在潛在特徵空間中解耦高度相似的作物特徵，並強化模型對困難樣本的鑑別力。 實驗採用農業部提供之高解析度 (15 cm/pixel) DMC III 多光譜 (Red-Green-BlueNear-Infrared， RGB-NIR) 航空影像資料集進行驗證。結果顯示，本研究所提之 ν-Adapter 在經歷連續增量任務後，能維持高達 0.9505 之平均 F1-Score，且增量 效能落差與相對效能落差僅為 1.40%，顯著優於現有的尖端技術 (State-of-the-Art， SOTA) 基準模型 (如 Model Surgery (MOS)、Selecting the Optimal Parameters is All You Need (SOYO) 等)。同時，相較於完整重新訓練，本方法依適應器維度不同可縮短約 61.99% 至 65.87% 的訓練耗時。本研究成功兼顧了高精準度、抗遺忘能力與實務部 署之高時效性，為智慧農業之持續性學習提供了一套極具潛力的解決方案。;Aerial image recognition is essential for precision agriculture. However， continuous crop monitoring requires models to handle Domain Incremental Learning (DIL) scenarios. Agricultural images are highly fine-grained， exhibiting severe inter-class similarity (e.g.， corn and peanuts at harvest) and massive intra-class variance across growth stages. Traditional finetuning suffers from catastrophic forgetting， replay-based methods face storage constraints， and full retraining incurs unacceptable time costs for rapid deployment. To address these challenges， this study develops an exemplar-free DIL framework based on a lightweight side adapter (ν-Adapter). Through parameter isolation， it updates only 0.13M to 0.44M parameters， retaining previous knowledge and mitigating catastrophic forgetting from the root. Furthermore， a dual loss function combining Supervised Contrastive Loss (SupCon Loss) and Focal Loss is integrated to decouple highly similar crop features in the latent space and enhance discrimination on hard examples. Experiments on the DMC III high-resolution multispectral dataset (RGB-NIR) demonstrate that the ν-Adapter achieves an outstanding average F1-Score of 0.9505 across sequential tasks. With a minimal incremental learning gap measure rate of 1.40%， it significantly outperforms state-of-the-art baselines (e.g.， MOS， SOYO). Additionally， depending on the adapter dimension， it reduces training time by 61.99% to 65.87% compared to full retraining. This framework successfully balances high accuracy， anti-forgetting capability， and time efficiency， offering a robust solution for continual learning in smart agriculture.
