無人機RGB影像衍生植生指數預測Humidicola (Brachiaria humidicola)產量和品質

Abstract

飼草是反芻動物營養的重要組成部分，依賴適當的牧場管理。然而，不當的管理會導致過度放牧和牧場生產力的總體下降，從而影響動物的表現。為了防止這種情況發生，畜牧農民需要及時和相關的飼草產量和質量信息。傳統的飼草產量和質量測量方法費時、勞動密集，並且對更廣泛的區域精度較低。利用無人機進行這種應用越來越受歡迎，因為它提供了更多的操作靈活性。然而，先前的研究由於使用覆雜的工作流程和非常昂貴的設備，其應用仍然有限。因此，有必要開展研究，開發一種準確而簡單、使用更便宜的工具、計算效率更高的工作流程。在這項研究中，使用了一架配備RGB相機的DJI Mavic 3 Classic 無人機獲取圖像。從這些圖像中提取了植被指數作為預測變量，開發了一個用於預測飼草產量和質量的多項式多元回歸模型。結果顯示，通過結合線性、二次和三次植被指數，可以從20米的高度高度預測鮮重和幹物質，調整後的R2分別為0.71和0.66。然而，預測粗蛋白和中性洗滌纖維含量的準確性較低，調整後的R2分別為0.48和0.53。這表明了一種替代簡單工作流程、成本效益工具和統計方法在Brachiaria humidicola牧場產量預測中的潛力。

Forages are a significant component of ruminant nutrition, which relies on proper pasture management. However, improper management results in overgrazing and an overall decline in pasture productivity, causing poor animal performance. To prevent this, timely and relevant information on forage yield and quality is necessary for the livestock farmer. Traditional ways of measuring forage yield and quality are laborious, time-consuming, and less accurate for wider areas. Using drones for this application is gaining popularity because it offers more operational flexibility. Previous research still has limited application because, despite promising results, they used complicated workflows and very expensive equipment. Thus, there is a need for research to develop a workflow that is accurate yet simple, uses cheaper tools, and is computationally efficient. In this research, a DJI Mavic 3 Classic drone with an RGB camera was used to obtain images. Vegetation indices were extracted from these images as predictors to develop a polynomial multiple regression model for the prediction of forage yield and quality. Results showed that fresh and dry matter can be highly predicted from 20-meter altitude with an adjusted R2 of 0.71 and 0.66, respectively by combining linear, quadratic, and cubic vegetation indices. However, the predictors have lower accuracy in predicting content of crude protein and neutral detergent fiber, with an adjusted R2 of 0.48 and 0.53, respectively. This demonstrates the potential of an alternative simple workflow, cost-effective tools, and statistical method for yield prediction in Brachiaria humidicola pastures.