Research On The Detection System Of Wooden Products Based On Color Feature

As an important index of the quality evaluation of wooden products, color affects the efficiency and effectiveness of processing production directly. Applying the surface color characteristics of wooden products to compartmentalize grade is an important part to perfect the means of quality detection and to improve the detection accuracy of small wooden products. In this paper, the color classifications detection of small wooden products was studied based on image processing and pattern recognition technology, which would provide the necessary theoretical and technical basis for industrializing realization of the automated color detection of wooden products production.Ice-cream sticks were selected as study objects, and divided into two levels in accordance with color. Two types of experimental samples database that contained 1000 images were used for the analysis and classification of color. First, using the median filtering method to eliminate the noise of the image. Then, using the threshold segmentation technology to divides the image to provide necessary base images for extracting color features.With analysis of several commonly used color spaces, comparing and combining the characteristics of color spaces, the HSV space model was fixed as the model to research surface color of small wooden products in view of the close approach advantages of HSV color space and human eye color perception. For the characteristics of homogeneous color distribution of two types of samples, just consider the overall distribution of colors, color histogram and color moment were chose as the two color analysis methods, and two sets of characteristic parameters was extracted respectively in the HSV color space:①H,S,V color components of a total of 18-dimensional color histogram feature parameters;②H,S,V color components of a total of 9-dimensional color moment feature parameters, which realized image color feature acquisition.In this paper, BP neural networks was used as a classifier, feature parameters of the color histogram and color moment were the input respectively, two different types of color samples were the output. L-M algorithm was selected to train the neural networks, the recognition rates of the system using the two sets of feature parameters were 98.2% and 97% while testing the 500 samples. The results showed that the difference of the two classification and recognition rates was not too large, therefore, the small dimension feature parameter of color moment was chose as the optimal parameter. The color detection system design of wooden products are given.The realization of this study would provide an effective way and powerful reference for the improvement of the quality detection and production efficiency of small wooden products.