Abstract: The neural network method for segmenting the true color of seedling image was studied. The adapted network is BP model. The feature vectors of image were obtained through analysis and experiments. In the network, the inputs are feature vectors of seedling image, the teacher signal is the image result of manual segmentation and BP algorithm is used. The effective parameter was achieved after a lot of training. The experimental results show that the segmentation of BP network method can gain better effect. This study establishes foundation for further seedling image processing.