【摘要】 提出了一种应用混合神经网络进行颗粒图像边缘检测的方法 .混合神经网络由用于对边缘候选图像的二值输入模式进行聚类特征提取的自组织竞争子网络 (ASCSNN)和用于获取颗粒图像边缘矢量信息的BP子网络 (BPSNN)组成 .边缘候选图像是通过采用基于灰度极小值算法提取边缘候选像素获得 ,神经网络以边缘候选图象中的边缘候选像素及其邻域像素的二值模式作为训练样本 .对经过噪声污染的图像进行实验表明 ,该方法获得的边缘图像封闭性较好、边缘描述真实、抗干扰能力较强、适用于颗粒图像的边缘检测更多还原

【Abstract】 Presents a hybrid neural network which consists of a self organizing competitive neural network used for feature extraction from binary input pattern of edge candidate image and a BP neural network used for deducing edge vectors, points out a logical judgement algorithm is used to get edge candidate image, and the edge pixel candidates and its neighbor pixels are used as the binary samples of the hybrid neural network, and concludes from the results of experiments on image corrupted with Gaussian noise that the image segmented by this method has good edge closeness and true edge and the method is suitable for segmentation of cumulate particle image.更多还原