复杂背景中的阔叶轮廓提取技术

树叶轮廓提取是植物图像识别的一个重要环节。设计一种复杂图像分割的综合算法。首先割取图像的目标区域利用Canny算子查找目标的边缘，然后边缘集合中查找所有的连通分支，利用二阶倒数查找连通分支长度与数目的关系，最后求出线条长度阈值，保留长度比阈值大的连通分支，进而得到目标图像的边缘。以阔叶楠树树叶为例进行实验，证明该算法可行有效，能够很好地过滤到背景的噪点和页面中的纹路干扰。

A Broad-Leaf Contour Extraction Technology in Complex Background

To extraction the leaf contour from an image is an important process in recognize plant by image. Proposes an algorithm based on integrat-ed technology. First, selects the target area and finds the edge by Canny operator. Second, finds all the lines, builder the relationship be-tween line numbers and line length, finds the threshold by second derivative. Finally, deletes the line that length less than threshold, Re-mainder is the contour of target. Experimental results of broad-leaf image shows that this algorithm is feasible and effective. This algo-rithm can filter the background noise and branch edges of leaf internal.