Study On Techniques Of Image-based Classification For Castanea Species

Species identification and classification is basically important for forest research and forestry production and management. According to traditional methods of species identification and classification, experts observe the samples to obtain the data of the external characteristics of them, and then classify them one by one based on species identification of Classification List, finally find out which Family, Genus, Species they belong to. With the rapid development of the image processing and computer vision technology, pattern-recognition technology has been widely used. Image segmentation is one of the key problems in computer vision. Since classical image segmentation techniques cannot satisfy the requirements of complex image segmentation due to their own limitations, model-driven image segmentation techniques come into being and become popular. Geometric Active Contour Model can easily handle the changes in the topology of the evolving contour, while it is difficult for Parameter Active Contour Mode. And the level set method has greatly promoted the development of the Geometric Active Contour Models. This paper explores how to apply the technology to the automatic identification of species by the example of Castanea species, thus enhancing efficiency and reducing the intensity of work.This paper mainly studies the level set method for image segmentation. It discusses the theory of curve evolution and level set method on the analysis of the status quo at home and abroad. As one of the major deficiencies of the traditional level set method is that in curve evolution Zero Level Set must be re-initialized in order to maintain or stay close to Signed Distance Function, there is no need to initialize Zero Level Set used in the fast level set algorithm, greatly simplifying the calculation process. In addition, due to the impact of light conditions, there is a shadow in the original image of Castanea, which is slight discrepancy between the target image. With the application of level set algorithm, if used around Zero Level Set to set Castanea goal and its shadow, it will inevitably lead to Zero Level Set in the shadow of the image area to be stopped by the target and form the shadow of a regional division. The paper amends it through determination of a little points in shadow region interactively and then interactive level set algorithm is applied to species Castanea image segmentation, and good results is achieved. According to the segmented target regions of Castanea images, the invariant moments and edge moments shape features extraction algorithm is applied to extract shape feature and the feature matrix is formed, and finally Support Vector Machine is applied on the fruits of Castanea species classification. The experimental results show that image identification of Castanea on the basis of the Support Vector Machine is better and its identification accuracy can reach 87.5%. It completes the integration of the prototype system and verifies the correctness of the algorithm and proves the accuracy of results by combining image-based classification system for Castanea species framework to Java and Matlab combination.