Support Vector Machine (SVM) Based On Intelligent Algorithm Combined With Wood NIR Application Research

Wood is a valuable resource for human survival and development. With the rapid development of economy and society in China, the demand for the quantity and quality of wood has increased year by year, and therefore, exploring an effective testing method on wood properties is of great significance to improve the quality of forest tree cultivation and wood genetic improvement technology and to help us make better use of wood. NIR (Near Infrared Reflectance) spectroscopy analysis technique is a new analysis technique. This new analysis can be used both quickly and accurately in non-destructive test on physical, mechanical and chemical properties of organic samples like solids, liquids, powders etc. SVM (Support Vector Machine, SVM) method based on statistical learning theory based on the theory of VC dimension and structural risk minimum principle, according to the limited sample information in the complexity of the model (i.e., on a particular learning accuracy of training samples) and learning ability (i.e., not wrongly’s ability to identify random sample) to seek the best compromise between, in order to get the best generalization ability. In solving the small sample, nonlinear and multidimensional pattern classification and regression forecasting has very obvious advantages.This paper first analyzes the principle and training process of support vector machine (SVM), deeply analyzes the influence on classification performance by the choice of c and y when using radial basis function as the kernel function of support vector machines (SVM). The intelligent optimization algorithm is applied to the support vector machine (SVM) model,which effectively accelerates the speed of support vector machine (SVM) and improves the accuracy.Quality of eucalyptus mywood content regression model based on particle swarm support vector machine is established.The model uses40eucalyptus near-infrared spectra of samples for the training set, eight samples for testing set, regression coefficient prediction is0.970956, root mean square error is0.0021545. Compared with the traditional support vector machine (SVM) regression model and partial least squares regression model,the results show that, PSO-SVM regression model has high accuracy and good stability in the the eucalyptus lignin content forecast.Wood classification model based on intelligent algorithm of support vector model is established.Support vector machine (SVM) model for binary classification is established,including eucalyptus and poplars, larch and pinus massoniana,mongolica and pinus massoniana. By the results it can be seen near infrared spectra of wood on the result of prediction, the greater the correlation, the more prone to the prediction error. Based on grid search method, genetic algorithm (GA) and particle swarm optimization (PSO) multivariate classification model of support vector machine (SVM) are established. Results show that the greater the search area of the grid search method, the higher the model prediction accuracy is. Genetic algorithm takes longer, and search effect is not stable; The particle swarm optimization (pso) is the best on searching the optimal parameters and the time is the shortest and the search effect is stable. The SVM model based on particle swarm optimization algorithm for wood identification effect is good, and has good development prospect.