Research On The Wood Cell Image Segmentation Based On Computer Vision

Along with the developing of wood science, it has become an important direction of wood research to analysis wood cell image by Computer Vision (CV). Image segmentation is an important part of C.V, and an important way to measure cell area, cell number, and percentage of fiber cell wall, which are all important parameters to reflect wood property. It is also the foundation of calculation and analysis in morphology. The result of it affected analytical result of wood property directly.In this thesis, we introduced the principle and application of some typical image segmentation methods, analyzed and compared them, and then determined the methods adapting to wood cell image segmentation.In the third chapter, we introduced the principle of the typical edge detection operators and used the method to segment the wood cell image. We found out that the most adapted operator to each image, the difference among them and the advantage and disadvantage of these operators for wood cell images.In the next chapter, we introduced the geometric curve evolution to segment the wood cell image to solve the questions that the typical edge detection operators could not solve. We introduced the principle of geometric curve evolution method, the image segmentation method based on level set method that was first introduced by Osher and Sethian, its fastingalgorithm------Narrow Band, Mumford—Shah model, the solution for M-S model based onlevel set method that was first introduced by Chan and Vese. We used these methods to segment the wood cell images and explained the advantage and disadvantage through the experimental results.At the last of the thesis, we improved the M-S model according to the characteristics of softwood. We changed the parameter c₁(inside mean gray) into C_ding (core gray value of certain object, springwood or latewood). Results of the experiments indicated that this improved method reduced blindness of the model, increased efficiency, improved the effectiveness of segmentation, and helped to build the foundation for classification.