Abstract: An improved marked watershed segmentation algorithm based on HOG descriptors is proposed for overlapping leaves of tomato plug seedlings. Firstly, the leaves region and background region are segmented by using the super green transform and the OTUS to determine the overall contour of the leaves. Then the gradient of the image is calculated by using the morphological color image gradient calculation method. Using the HOG descriptor of the image gradient, the region with leaves edges and the region without leaf edges are selected, the areas without leaves edges are used as markers for watershed segmentation after morphological manipulations. Finally, image gradients are reconstructed by markers and watershed segmentation is performed to obtain segmented results of overlapping leaves. The experimental results show that the proposed algorithm can accurately select watershed markers, realize blade segmentation, and provide support for the classification of plug seedlings.

Key words: histogram of gradient, watershed algorithm, color image gradient, overlapping leaves