Research On Recognition And Classification Method Of Forestry Pests Based On Fractal Theory

Forestry is an important basic industry and public utilities, and it is also a valuable resource of our country. However, the forest often suffers from the invasion of forest pests and diseases in the long process of growth, which exerts a serious impact on forestry development and ecological environment construction process, and causes significant losses to the national economy. Therefore, it is necessary to design a forestry pest classifier, so that some non-professionals can also accurately identify forest pests, thus can get a timely forecast. For this purpose, selecting three types of forest pests as study targets, this paper constructs a forestry pest identification and classification system based on the fractal theory and SVM.In order to obtain more accurate characteristic parameters of forest pests image and have better identification and classification, certain preprocessing of the collected forest pests image have been done in the first place. Then on the basis of fractal theory, obtaining the edges and image divisions by judging the range of the H parameter for DFBR model, so as to extract the geometry shape features of forest pests images; and texture features are also extracted from pests images by using differential box counting to get their fractal dimension, and then the obtained characteristics will receive normalization and feature selection, among which the final selection of features will be used as the classifier input. Considering that the number of obtained forest pest image samples is not very large, and also the good performance of SVM in solving finite sample learning, nonlinear and high dimension learning, finally this thesis select the SVM method to design classifier in order to achieve forest pests image recognition and classification. And the category effects under different nuclear functions are compared through experiments.The result of this research shows that:the forest pests identification and classification system constructed on the basis of the combination of the fractal theory and SVM can effectively achieve the classification of different types of forest pests, and thus has great practical values in forest pests and diseases forecasts.