The Identified Classification Of Korean Pin’s Male And Female Flowers Based On Bp Nerve Net

Recent years, the prediction of Korean pine cone production is still stopped at the method of manual analysis and experiences summary. In order to realize automation and high efficient of it, taking modernized information method has become the main researching direction at present. It’s the main purpose of this paper to extract the morphological characteristic features of male and female flowers which come from its images of Korean pine cone production transferred back by Field Server-based Remote, meanwhile to classify these images by manual nerve net, from that comes out the precise data for male and female flowers’ of Korean pine that can be taken as an important reference for the prediction of Korean pine cone production.This paper preprocessed these images including smooth denoising, grey scale enhancement and region segmentation. For finding the most appropriate template to smooth the images, abundant samples are experimented. The new images made by grey scale enhancement are convenient for region segmentation where the body and background of Korean pine male and female flowers are separated.By combined methods Ant Colony Algorithm and GVF Snake model to make the edge extraction of the preprocessed images, then to extract six characteristic features which can identify male and female flowers through analyzing the edge images. In the process of extracting characteristic features, as methods of extracting width to length improved, the width to length feature is precisely extracted. The optimization of the male and female flowers’ feature space is realized by simplified features into four main characteristic ones which contributed a lot in classification of Korean pine male and female flowers, they are width to length, circularity, elongation and density.BP nerve net model in this paper is optimized by particle swarm optimization which mainly focused on convergent rate in net learning, from that learning ability is improved. And by optimized selection of initial control to select net initial control feature, it places net in a relative optimized condition instead of trapping in local minimum at the beginning of learning. The male and female flower classifier designed in this paper is just using the optimized nerve net involved above, to train the net by abundant learning samples in experiment. After the train this classifier is tested and the results are analyzed, all that steps proved the classifier can be used in the classification of Korean pine male and female flowers at a more accurate level.