Plant seedling stage main stem detection method in phenotype imaging box environment

The invention discloses a plant seedling stage stem detection method in a phenotypic imaging box environment, plant stem detection is realized by utilizing the characteristic that a detection line segment set to which a plant stem belongs in an image has strong correlation in position and direction, and the method comprises the following steps: (1) detecting line segments through an LSD method; (2) clustering the line segments according to the center point positions and direction information of the line segments; and (3) calculating the confusion degree in each line segment class and determining the line segment class to which the main stem of the plant in the seedling stage belongs according to the confusion degree. The method has the advantages of high accuracy, strong robustness, less constraint conditions and the like. And particularly, the method has good adaptability to a shooting angle, and a plant main stem target does not need to occupy an obvious position and a relatively large proportion in an imag