Comparitive Analysis of Segmentation Methods for Wheat Canopy Extraction
Global food demand is expected to be doubled by 2050, while natural resources are continuously under threat due to unpredictable climatic changes. This challenge can be tackled by increasing the yield of the crops and by reducing abiotic stresses such as water stress. Research shows that due to water stress the morphology and the structure of plant’s canopy changes. The first step in building early water stress detection system is to extract accurate area where photosynthetic activities of the plant are occurring. In this research work, comparative analysis of seven different segmentation algorithms viz., convolution gradient-based, watershed, mean-shift, k-means, Global static thresholding, Otsu thresholding and hybrid approach (combination of Global Static thresholding with k-means) has been analyzed in order to identify the most probable area of canopy where maximum photosynthetic signals can be captured. The comparison is done in terms of IoU metric. The comparative results indicate that the most appropriate method for wheat canopy segmentation is a hybrid approach, which achieves IoU score of 59.8 and its runner up algorithm is Global Static Thresholding with an IoU score 53.8.
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.