用于原木材积检测的图像处理与分析算法

Image processing and analysis algorithms for measuring log volume

基于原木材积自动化检测系统的检测原理 ,研究并设计了关键的图像处理与分析算法 .针对原木端面为类圆形状这一特点 ,对其进行Hough变换 ,然后结合数学形态学中的骨架提取算法 ,在骨架极值对应的最大圆盘中搜索Hough变换的峰值 ,使每一峰值对应一个原木端面 ,从而实现了原木端面图像的粗略分割 .将活动轮廓模型变换到极坐标系中 ,以梯度的极值作为端面轮廓的候选点 ,可以利用动态规划法来获得能量泛函的全局极值 ,从而实现了精确提取端面轮廓 .

The key image processing and analysis algorithms were investigated and designed based on operating principle of the automatic system for log volume measurement. Hough transform (HT) was applied to the image since log ends are quasi-circular. Peaks of the HT were detected in maximal balls corresponding to each maximum skeleton point. Consequently, the image was coarsely segmented by each HT peak corresponding to one log end. The active contour model was transformed into the polar coordinates. And the maxima of image gradient were selected as candidate points of log end contour. Global minimum of the energy functional can be found using dynamic programming. With this approach, the precise contours of the log ends can be extracted. Experiment shows that the proposed algorithms are of significance.