Study On Extraction Of Forage Shape Features And Recognition System Based On Image Processing

To monitor the grass species of pasture livestock has a great significance in studying the relationship between livestock and grass, grassland management as well as evaluation of the sheep’s growth. However, currently the monitoring is mainly accomplished by artificial observation, which not only has strong subjectivity but also is time-consuming and inefficient. Thus, to explore a kind of automatic method for identification of the grass species is of great significance.In this paper, aiming at the characteristics of desertification grassland, by taking grass leaf images collected from the grassland of Wangfuyidui in Siziwang County as the research objects, Phlomis umbrosa, Leymus Chinensis and potentilla anserine were selects out from dozens of common grasss, and50group images were respectively collected. The research was conducted from four aspects including image acquisition, feature extraction, image recognition and system implementation. By using the methods of building models under color space, morphological operator, BP neural network, the target leaves were successfully separated from background, the complete blade profiles were extracted, the four kinds of shape characteristic parameters are calculated, and the identification of grass species were realized and also the users’interface were designed by using neural network.Finally, the feasibility of the automatic identification of grass species based on image processing was verified by tests, the results have laid a foundation for future identifying the other species of grass, the whole-shaped grass, and establishing the grass species database.