Microscopic image recognition method of stomata in living leaves based on improved YOLO-X

Abstract Stomata are the main medium of water exchange in plants, regulating gas exchange and responsible for the processes of photosynthesis and transpiration. Stomata are surrounded by guard cells and the transpiration rate is controlled by opening and closing stomata. Stomatal state (open and close) plays an important role in describing the health of plants. In addition, counting the number of stomata is of great significance for scientists to study the number of opening and closeing stomata and to measure their density and distribution on the leaf surface through different sampling techniques. Although some techniques for calculating the number of stomata have been proposed, these methods are used to produce samples in isolation and then to identify and classify the states in the sample leaves. We improved YOLO-X and then implemented a transfer learning method to count the number of stomata and identify the stomatal opening and closing status of live black poplar leaves. In the end, the average accuracy and recall of the method were 98.3% and 95.9%, which helped researchers to obtain accurate information on leaf stomatal opening and closing status in an efficient and simple way.