基于YOLO v5的植物叶绿素含量估测与可视化技术

Chlorophyll content plays an important role in plant photosynthesis, and is indicative of the growth and health of plants. There has been strong interest to measure chlorophyll content quickly and nondestructively, and visualize its spatial distribution in plants. A custom-made imaging platform was used to acquire multi-view RGB images of the seedlings of Salix suchowensis Cheng, a close sister species of poplar. An experiment in growth chamber was conducted involving 32 seedlings. These seedlings was subjected to four levels of nitrogen rates. A series of image processing algorithms was developed, which allowed us to detect the main branch of the plants (using YOLO v5), extract color indices from the main branch to estimate SPAD values (using ridge regression), and obtain the best color factor combination regression model by comparing various model regression methods. The results showed that the best performing regression model to estimate SPAD values employed six color indices derived from the RGB images as predictor variables, with R² of 0.73 and RMSE of 2.16. Finally, the spatial distribution of chlorophyll content of the whole seedling was developed and visualized. In conclusion, the rapid and non-destructive approach to estimate chlorophyll content of poplar seedlings using high-throughput, multi-view RGB imaging was investigated. The imaging platform, the algorithms for plant image analysis and color indices extraction, as well as the models to estimate SPAD readings, provided technical feasibilities to continually assess growth and health related parameters for tree seedlings, and could guide the early diagnosis of nitrogen stress in plants and suitable application of nitrogen fertilizers.