A watershed segmentation algorithm based on an optimal marker for bubble size measurement

•A new watershed segmentation algorithm is proposed for measuring the bubble size.•An improved sub-image classifier based on SVM is introduced.•A lumped together marker checking method based on the skeleton is presented.•The optimal marker extraction method in the froth image is proposed. Bubble segmentation is the most popular method for bubble size measurement. However, due to the complexity of the froth image, the present image segmentation methods cannot get a satisfactory result. In this paper, a watershed segmentation algorithm with an optimal marker is proposed. The marker extraction method is based on the sub-images, and thus an improved sub-image classification model is built. To reduce the under-segmentation, a lumped together marker checking method is also developed based on the skeletons. Then, the optimal marker is obtained by the data fusions with the three marker regions, the light reflection character of the bubble and the class information of sub-images. The industrial experiments show the effectiveness of the proposed method, in which the accuracy is improved by 11.44% and 9.10%, and the robustness is improved by 41.48% and 57.95%, respectively, compared with the two other present methods.