摘要：

针对纹理及复杂背景干扰下的木材单板缺陷识别问题，结合几何活动轮廓模型和无边活动轮廓模型各自的优点，提出一个改进的活动轮廓模型。采用对偶化的带权TV范数进行求解，得到该模型的快速实现算法。实验结果证明，该模型可实现复杂纹理背景下单板多节子目标的检测，算法计算量较小，单板缺陷的识别速度较快。

关键词: 木材单板, 活动轮廓, 多目标识别, 对偶化方法, 复杂背景

Abstract:

Aiming at the problem of wood veneer defect recognition with a texture, noise jamming and complex background, this paper proposes an improved active contour model. This model is combined with the respective advantages of the Geometric Active Contour(GAC) model and the active contours without edges model. It solves the model with a dual formulation of the weighted TV-norm and obtains a fast algorithm. Experimental results show that the model can be competent for the detection of veneer which contains multi knots and complex texture background. The algorithm has less computation, faster identification of veneer defects.

Key words: wood veneer, active contour, multiple target recognition, duality method, complex background

中图分类号: 

• TP391