普通光照下叶片图像特征信息抽取

在光照不均匀，存在阴影以及存在背景小杂色块干扰的图像中准确辨识出叶片图像，并将其显著特征抽取出来是叶片图像特征的研究重点。对实际叶片图像的处理，提出了先综合利用图像边界探测算法以及连接线、连通域抽取及变形算法确认叶边缘和叶脉图像，去除了光影，杂色轮廓的干扰，接着综合利用Hough变换、角点检测等算法来抽取树叶叶形，叶脉特征。实验中利用SVM（Support Vector Machine，支持向量机）算法对抽取特征进行分类测试，分类正确率超过了90%。

Feature extraction from leaf pictures taken in normal illumination condition

Key point of processing leaf image is to extract distinguishable features of leaf from images containing both shadow and background noise interferences. To process the picture taken in real circumstance, a two-step method is proposed. It uses the combination of edge detection algorithm, connected lines and domains extraction method and shape modify algorithm to obtain the exact leaf edges and the leaf vein, eliminates the disturbance of shadow and background noise. It uses Hough line transform algorithm, Harris corner detector algorithm and other feature detecting algorithm to extract the features of leaf edge and leaf vein. When using the futures extracted to perform a SVM（Support Vector Machine）classify algorithm, the result shows accurate is above 90%.