基于数学形态学的树木图像分割方法

该文针对智能整枝机视觉系统的需要提出一种以数学形态学为主的图像分割方法,解决了智能整枝机视觉系统中背景复杂下树木图像分割困难的问题.该方法分成5个步骤:①利用小波变换将图像进行压缩;②亮度矫正;③用分水岭算法对图像进行分割;④用区域合并法解决分水岭算法中的过度分割问题;⑤用中值滤波器滤除分割过程产生的孤立点.结果表明:该方法比Roberts算子、Sobel算子及Prewitt算子对树木图像的分割效果好,为立木整枝机视觉系统的深入研究打下了基础.

Method of tree image segmentation based on mathematical morphology

This paper presented an efficient scheme of tree image segmentation to meet the need of the machine vision system of intelligent pruning machine. The scheme can segment the tree from the complicated background, which is mainly based on mathematical morphology. The scheme includes five steps: 1) image compression by wavelet transformation; 2) lightness rectification; 3) image segmentation by watershed algorithm; 4) merging redundant regions; 5) removing the insular pixel by median filter. Experimental results demonstrate that the scheme is more effective than Roberts arithmetic operator, Sobel arithmetic operator and Prewitt arithmetic operator in tree image segmentation, and sets a foundation of the in-depth research of the machine vision system of pruning machine.