The Method Research On Tree Species Identification Of Texture And Chromatism On Plate Surface

In our country, there are wide variety of trees with different properties. In order to make use of resources of timber efficiently and rationally, we need to identify the samples variety of trees accurately. But it is a complex process to identify the timber. And traditional wood identification methods with workers eyes are not adapted to wood industry automation and intelligent development trend. Aiming at actual condition, we use computer as the automatic detection platform, make the solid wood products selection on line as the background, integrate the experts experience with fuzzy logic method and the design ideal of active shape model, and build the model of wood species identification based on texture and color characteristics of the plate surface, through the means of restricting surface shape and veins of timber in advance.Aiming at classification problem of the shape of timber surface texture in the model of species identification, we choose2D fuzzy logic classifier base on the rule. Thereinto, the input texture features root in graylevel co-occurrence matrix. There are14primitive features in the graylevel co-occurrence matrix quadratic statistics. We choose4features nonlinear correlation to make combination with each other. Finally, we choose Energy and Correlation as input texture features of classifier. In addition, the rule of fuzzy logic classifier is checking the standard features distributing in the picture, getting distributing from the statistics experience, finishing the building of core classifier according to the character of additive fuzzy system.Aiming at identification problem of colors of timber surface in the model of species identification, we introduce the modeling idea of the global shape model of active shape model, and extract color histogram on the three channels of RGB color space in order to building a single species of the color space model, that is the identification of the species.Finally, we choose200pieces of pictures of oak as experimental samples to build the oak recognition model. In our simulation, we choose100pieces of pictures of oak and100pieces of pictures of Korean pine and we get the recognition rate that is92.5%on average. Base on the oak recognition model, the simulation results show the feasibility and scalability of this species identification methods, the development of automatic detection systems for wood processing and applications with reference value.