開發天車與無人地面載具表型平台於溫室葉菜作物株高及葉面積估測

Abstract

目前現有的表型平台大多應用於糧食及蔬果作物為主，然而在蔬菜類作物中，較少應用於葉菜類作物，葉菜類作物生長通常相對容易且生長週期短，提供快速的農作物供應。若能夠建立針對葉菜類作物之表型平台，針對作物生長和表現的全面監測與評估，便能提高葉菜類作物的生產效率和品質。本研究開發天車與無人地面載具作物表型平台用於量測葉菜類作物之株高及葉面積，其架構主要分為三個部分，天車與無人地面載具表型平台、深度校正與株高及葉面積估測演算法。天車與無人地面載具表型平台使用RGB-D相機收集RGB及深度影像，然而所收集之深度影像，會因為環境亮度等原因導致誤差，因此本研究預先建立深度校正查找表，對相機量測到之深度進行校正。株高及葉面積估測演算法，以AR marker作為參考基準，首先，透過影像處理技術將AR marker及植株進行分割，再透過演算法將株高及葉面積估算出來。本研究分別使用天車及無人地面載具對小白菜及青江菜之株高及葉面積進行調查，天車之株高RMSE為0.99 cm，葉面積RMSE為5.532cm^2 ; 無人地面載具之株高RMSE為1.365 cm，葉面積RMSE分別為6.969cm^2。結果顯示，本研究天車及無人地面載具表型平台可以針對株高及葉面積進行估測。

The current existing phenotyping platforms are mostly applied on food, vegetable and fruit crops, however, in vegetable crops, it is less used in leafy vegetable crops. Leafy vegetable crops are usually relatively easy to grow and have a short growth cycle, which provide rapid crop supply. If phenotyping platform for leafy vegetable crops can be developed to monitor and evaluate crop growth and performance, the production efficiency and quality of crops can be improved. This study develops crop phenotype platform for crane and unmanned ground vehicle (UGV) to measure plant height and leaf area of leafy vegetable crops. The architecture is divided into three parts: the crane and UGV phenotype platform; the depth calibration; plant height and leaf area estimation algorithm. Crane and UGV phenotyping platforms are used to move and capture RGB and depth images. The depth images will have errors due to environmental brightness and some other elements, therefore, this study intend to reduce the errors by utilizing depth calibration lookup table. The plant height and leaf area estimation algorithm use the AR marker as reference. The AR marker and the plant are segmented with image processing approach, while the plant height and leaf area are estimated through the algorithm. In this study, the crane and UGV were used to investigate the plant height and leaf area of Bokchoy and Spoon Cabbage. The measurement of plant height RMSE of crane is 0.99 cm, and the leaf area RMSE is 5.532cm^2, meanwhile, RMSE of UGV is 1.365 cm, and the leaf area RMSE is 6.969cm^2. The results show that the phenotype platform of the crane and UGV in this study can estimate the plant height and leaf area.