ABSTRACT
The objective of this research was to classify plant water-stress levels using PCA (Principle Component Analysis) and a Back-Propagation (BP) neural network via chlorophyll fluorescence induction profiles. This was achieved using chlorophyll fluorescence images combined with the chlorophyll fluorescence kinetics parameters. The images were recorded at 690 nm with a high-resolution imaging device consisting of Light Emitting Diodes (LEDs) for an excitation at 460 nm, and an EMCCD camera. Then, the chlorophyll fluorescence kinetics curve was obtained and chlorophyll fluorescence kinetics parameters could be calculated. Furthermore, the characteristics of chlorophyll fluorescence kinetics curve were extracted and their dimensions were reduced by PCA. According to the accumulative variance contribution rate, three principle components were selected to replace the character parameters. We clustered and classified different water-stress levels with these principle components, using a BP neural network. The result showed that the accuracy rate of clustering water-stress on different levels could reach up to 92%. This technique has the potential to monitor the water-stress condition of plants non-destructively and simply.
