A method for evaluating the common models for feature extraction of rice seedlings

Rice seedlings have different shapes at different growth stages. In south China the paddy fields often have duckweeds and cyanobacterias, whose colors are similar to the color of rice seedlings. This causes great difficulty in extracting the feature of rice seedlings. To solve this problem, four models including G-R model, ExG model, ExG-ExR model and S component model are used to extract the feature of rice seedlings. These models are used to convert the color images to gray level images. First, We analyze the gray level images converted by these four models. We also compare the advantages and disadvantages of each model. Then we present a method based on evaluating factor for selecting the most suitable model in different situations. The result proves that the presented method is effective in choosing the most suitable model.