中文摘要：

为实现对护坡植被根系图像的轮廓提取和边缘检测，提出一种遗传算法和误差反向传播算法相结合的混合算法来训练前馈人工神经网络。将描述边缘的特征向量，作为神经网络的输入信号，构造三层前馈网络。在网络训练过程中，构造的边缘特征量抗噪性佳，且可有效提取边缘的真实信息。仿真实验表明：设计的神经网络抗噪性强，边缘定位精度高，提取后图像更接近根系真实轮廓图像，更有利于植物固坡研究中根系形态的监测。

英文摘要：

A hybrid algorithm which combined with genetic algorithm and back-propagation algorithm to train a feed-forward arti- ficial neural network is presented in order to realize the contour extraction and edge detection of the images of plant roots for slope protection. The built eigenvectors for describing the edge are used as input signal of a three-layer feed-forward neural network. The built edge eigenvectors are robust against noise and the genuine information of edge can be extracted effectively in the process of network training. The experimental results illustrate that the designed neural network achieves excellent performance. It is noise robust and accurate in genuine edge positioning. The contour extracted by this method is closer to the practical contour; therefore it is more beneficial to the monitoring of root morphology of vegetations for slope protection research.