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·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像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，适用于颗粒图像的边缘检测· A method of detecting the edge of the particle image by using the hybrid neural network is proposed. The edge candidate image is obtained by extracting the edge candidate pixel based on the gray minimum algorithm. The neural network takes the edge candidate pixel in the edge candidate image and its neighborhood Pixel binary pattern as a training sample Experiments on noise-contaminated images show that the edge image obtained by the method has better closedness and true edge description, which is suitable for the edge detection of particle images