Recognition Research Of Wood And Bark Based On Image Processing

In the industrial of wood, bark and wood must be separated in order to improve the efficiency of the use of wood. This paper, using digital image processing techniques to recognition wood and bark of the elm, willow and pin, while using parameter values of processed image and BP neural network and SVM computer recognition technology to identify separation. Verify the validity of the identification parameters and in this paper puts forward a new identification parameter. Validation of BP and SVM for recognition effect of wood and bark, then established efficient recognition model of bark and wood.Paper main research results are as follows:1. It was found that the acquisition of the early treatment to the image can improve the accuracy of image data, and the rate increase beneficial for later experiment. In the experiment also find the best source of light and background color image acquisition.2. The experiment confirmed the parameters can be used in wood and bark identification recognition, and established the image parameter extraction system. In this paper proposed a new recognition parameter mean square ratio by analysis the image parameter data.3. Established a wood and back recognition system based on the BP neural network and SVM kernel function and inspects the identification parameters identification effect through BP neural network and SVM different kernel functions.①Identification wood and bark by BP neural network, and establish a BP neural network recognition model. The parameters in the BP neural network recognition rate are generally lower than70%, and only the new parameters mean square ratio reaches78.8%.②In research of wood and bark based on polynomial kernel function to SVM and recognition rate of brightness is97.2%while the brightness MSE and mean square ratio are87.8and97.7%.③The experiment using the SVM radial basis function of image recognition and found that the radial basis kernel function is a better recognition rate. The parameters of luminance variance, brightness gradient variance, variance and mean square ratio is reach a recognition rate of94.6%,95.6%,95.6%and94.6%.④Established an identification model by Sigmoid function in using SVM, and the mean square ratio is94.7%while luminance variance recognition rate is90.6%and luminance variance recognition rate is90.6%. Other parameters recognition rate above80%except the brightness of MSE recognition rate is37.1%. 4. The study found that the wood and bark mean square than the optimal identification of parameters, and the SVM can be used to obtain higher recognition rate.