Segmentation algorithm of overlapping tomato seedling leaves based on edge chaincode information

According to overlapping leaves while robotic transplanters extracting seedling leaves’ feature parameters with machine vision, a layer-by-layer segmentation algorithm was proposed based on overlapping leaves chaincode information. Firstly the chaincode was counted and corners and edges of overlapping leaves were detected based on edge binary image from original image preprocessing. Then the overlapping leaves were separated layer-by-layer through authentic corners and edges pairing, and the overlapping leaves segmentation was achieved by linear interpolation. The result of experiments showed that the segmented leaves images would not be changed by rotation, and the success rate of the algorithm was 100% and 96% respectively and the separating time was 0.835 and 0.99 s respectively for two types of plug size (72 cells and 128 cells). This segmentation algorithm has the advantage of high speed and high accuracy with rotation invariance, and can meet the requirements of automatic transplanting.