A Method of Human Interaction based Semi-automatic Counting for Digital Microscope Image of Spore

The sporulation of plant pathogens is an essential index for epidemic prediction. Counting is a necessary and vital step for sporulation computation. However, it is still a challenge to achieve efficient and accurate counting under the view of a microscope. The traditional manual exam is costly and time-consuming. The most automatic counting program is fast, but most of them may have poor performance when the digital images are noisy due to some particles, such as agar block, mycelium, floating in the suspension of sporangia. Here, we proposed a method of human interaction based semi-automatic counting for digital microscope images of spore. The scheme includes the components of image input detection, image preprocessing, Hough transform-based automatic spore counting, and human interaction based rectification. Mainly, considering the complex background of the spore microscope image caused by impurities in sporangium suspension, the accuracy of spore detection is low when only using simple automatic counting, we added a human rectification step after the automatic counting into the new scheme. To verify the efficiency and performance of the proposed interactive spore counting method, we randomly selected 20 microscope images of spore in the experiments. The results show that the average accuracy of the spore number is 0.9031 by only using the automatic counting of our scheme. Furthermore, all spores are correctly rectified and counted when semi-automatic human interaction is applied. To explore the improvement of efficiency, we compared the time consumption of manual, automatic, and semi-automatic counting methods. The average time consumption of a single image process is 30.78s, which has over 60% time reducing compare with the 104.9s of manual examination. Our proposed method is slower than the automatic process but far faster than the manual process, and it can ensure the examiner gets satisfying results.