复杂背景下的大豆叶片识别

利用计算机视觉和图像处理技术对叶片进行识别在农业领域逐步得到应用,但是,将叶片从具有复杂背景的图像中准确识别出来还是很难。本文提出了一种能有效地从具有复杂背景的图像中识别大豆叶片的算法。该方法首先利用基于HSI空间的三次标记分水岭算法提取出目标叶片,而后计算出目标叶片的形态参数,最后利用训练好的概率神经网络分类器对大豆叶片进行识别。对大豆叶片120幅样本图像能达到85.37%的识别成功率,证明了此方法的有效性。

Soybean leaves recognition of images with complicated background

The computer vision and image processing techniques have been widely used to recognize the plant leaves in the field of agriculture.However,it is still a challenging problem to accurately recognize the leaves from the images with complicated background.Is presented in this paper.An effective algorithm to recognize the soybean leaves from the complicated background images.In this algorithm,the marker-controlled watershed method based on HSI space is applied to extract the target leaf from the complicated background image firstly.Secondly,the morphological parameters of the target leaf are computed for representation.Finally,the trained Probabilistic Neural Networks is ap-plied to classify the soybean leaves into three classes.85.37% of successful recognition rate on 120 test soybean leaf images demonstrates the effectiveness of this algorithm.