Research On The Method Of Wood Fiber Morphology Detection Based On Skeleton Information

In the field of wood processing, wide application of wood fiber products have become a trend. With the increasing amount of its usage, the requirements for product quality and performance have become increasingly strict. Studies have shown that the morphological characteristics of wood fibers affect the quality of manufactured goods directly, so exploring method applied to wood fiber morphology detection is particularly important.Skeleton is a kind of method that describes object shape with reduced dimensions. Using skeleton information to obtain the morphological parameters of wood fiber has a positive effect on the realization of forestry intelligent processing and production. The traditional skeleton extraction algorithm is difficult to balance between the accuracy and the connectivity, and it is easy to be disturbed by boundary noise, so it has been optimized in this paper. We proposed a new skeleton extraction method based on vector inner product, then we used standard images of graphic database as experiment objects to carry out verification. Firstly, the Euclidean distance transform was carried out for experimental images. Then we regarded the changed times of boundary vector direction as a criterion to select skeleton points. Finally, the complete skeleton was generated by extending process based on regression analysis. Experimental results have shown that the algorithm can guarantee connectivity and integrity of skeleton, and the average accuracy rate of location has reached 92.96%.In order to overcome the disadvantages of high cost, large error and high dependence on traditional fiber morphology detection methods, this paper proposed the measurement of fiber length and its degree of curvature by using skeleton information. The length of skeleton is approximated as the true length of wood fiber, and the projected length can be calculated according to the endpoint coordinates. Finally, the fiber’s curl index was obtained. This method can be better to avoid the influence of wood fiber morphological changes to the length measuring accuracy, moreover the curl index can be used to quantify the bending degree of wood fiber. In order to compensate for the lack of mathematical model support for the traditional study of fiber morphology, we have made function fitting on skeleton discrete points in this paper. Experimental data shown that using polynomial function to fit wood fiber morphology model is the most reasonable way, then according to the first derivative of fitting function can determine the type of wood fiber.For the eventual use of image processing techniques to obtain the morphological characteristics of wood fiber, we designed the morphology detection system of wood fiber. In the system, we finished lots of function, such as the skeleton extraction of wood fiber image, the measurement of wood fiber shape characteristic parameter and the classification of wood fiber shape type. It has a promoting effect on the realization of the intelligent detection of wood fiber morphology.