Automatic counting for trapped urediospores of Puccinia striiformis f. sp. tritici based on image processing

Abstract: Using spore traps to capture the airborne plant pathogen spores in the fields is a key means to monitor the pathogen amount. It is of great significance for forecasting and management decision-making of airborne plant diseases. Currently, the traditional microscopic spore counting method is usually used to count the trapped spores. Due to the great number of the trapped spores, this method is time-consuming and labor consumptive, and often leads to a great error. To find out a method for automatic counting of in-field trapped pathogen spores and improve accuracy and efficiency of spore counting, urediospores of Puccinia striiformis f. sp. tritic, the causal agent of wheat stripe rust, was trapped (via indoor simulation) using transparent tapes, glass slides with vaseline and Eppendorf centrifuge tubes in this study. And then the images of the trapped spores were acquired using a microscope camera. Finally, using MATLAB software, the urediospores were automatically counted and marked through a series of image processing including image zooming using the nearest neighbor interpolation method, image segmentation using K_means clustering algorithm, morphological image modification and watershed image segmentation. The satisfactory results for counting the trapped spores were obtained after processing the spore images acquired by using the three kinds of simulation methods. The average counting accuracy for the urediospores of P. striiformis f. sp. tritici trapped on transparent tapes, glass slides with vaseline and in Eppendorf centrifuge tubes was 98.5%, 98.7% and 99.9%, respectively. Average counting accuracy for the urediospores of P. striiformis f. sp. tritici mixed with the conidia of Blumeria graminis f. sp. tritici, which could cause wheat powdery mildew, was 99.8%. The research provided a simple, fast, accurate and efficient method for automatic counting of in-field trapped pathogen spores.