Recognition Method of Grafted Seedling Characteristics Based on Machine Vision

Correct identification of direction of seedling cotyledon was one of critical technologies for grafting machine to achieve fully automated.White pumpkin seed stock cotyledon direction feature recognition method under natural light conditions was analyzed.Firstly，images of seedlings were pretreated to extract cotyledon boundary.Then minimum bounding rectangle of seedling cotyledons boundary was extracted successively using regional mark.Seedlings boundary was transformed through regularized Hough transform in the external rectangle to fit profile curve of two cotyledons.Finally，an ellipse mathematical model was established according to outline of cotyledon to strike direction of seedling cotyledons，to get seedling growing point position and to access to information of cotyledon leaf area and other features.100 seedling images were identified，reaching 85% success rate.The proposed method can identify direction characteristics of white pumpkin cotyledon effectively and can be applied to feature recognition of other varieties by adjusting parameters.