Study On Variation Regulation Model Of Wood Ring Density

Along with the expansion of our country's forestation area and the improvement of the lumber science and technical level, it is one of the important issues which will be solved urgently for our country forestry science to evaluate the wood quality and wood properties scientifically. Wood density is an important index to evaluate the wood quality and it's also the external reflection of the comprehensive index of the wood interior factor. We can estimate almost all the mechanics intensity and the elasticity through the change of the wood density. So it is very important to find out the variation regulation of wood ring density in growth process. It has very significant sense for cultivating forests to research the variation regulation of wood ring density quantificationally and then to get to the relation between the basic density and the anatomical characteristic of tracheid.It is very complicated to describe the model because the factors to influence wood density variation are complex and multiply. We can set up model which reflects the variation of wood density along with the growth of ring using traditional statistical regression method, but the fitting precision and the correlation are low. So it is necessary to find out a new analysis method to research the radial variation of wood ring density. Neural network model is through studying actual input/output data to make the objective function minimum and then to induce the connotative relation between the input/output data, namely neural network model. In this thesis, researching from cell microscopic structure in rings, we set up the variation regulation model of wood ring density based on the neural network theory, picked up the anatomical character parameters of cells in rings (diameter of tracheid, tracheid wall percentage, wall-indiameter ratio) which have important influence on wood ring density variation and trained the neural network model using various parameters data got by Computer Vision system, then obtained the variation regulation model of wood density in rings.We set up variation regulation model for wood ring density based on neural network which provided the theory basis for further analyzing some questions about physics mechanical properties of timbers, for making effective measure to breed fast-growing, high quality timber and for forecasting material quality in early stage. In this thesis, we presented the neural network model structure and the training algorithm, realized the neural network data training with MatLab and confirmed the model which is accurate. The simulation result indicated that the model is effective and reliable. At the same time, comparing the model with traditional model by regression method, we obtained that the former is better.