Recognition of dense fluorescent droplets using an improved watershed segmentation algorithm

Fluorescent droplet images acquired during droplet digital Polymerase Chain Reaction(PCR) detection have dense distribution, low brightness and low contrast, resulting in poor recognition accuracy. In order to correctly identify densely distributed fluorescent droplets, a fluorescent droplet recognition method based on an improved watershed segmentation algorithm is proposed. First, the image is preprocessed using histogram equalization and Gauss filtering, then the local adaptive threshold segmentation method is used to extract the targets from the background, thereby reducing the dependence on image gray level information. Finally, the algorithm combines the prior knowledge of the droplets with a circular and uniform size to define the droplet adhesion function, which reduces the error rate in the watershed segmentation. The experiment results show that compared with the traditional distance-based watershed segmentation method, the accuracy of the proposed algorithm is 97.34%, which is higher than the 85.9% accuracy of its counterpart.