论文部分内容阅读
针对颗粒图像的特点，提出一种基于神经网络的边缘混合检测方法。该方法包含边界候选象素提取和神经网络边缘检测两部分，神经网络由用于图像信息压缩与图像信息编码的自组织竞争子神经网络（ＡＳＣＳＮＮ）和用于获取图像边缘矢量信息的基于径向函数子神经网络（ＲＢＦＳＮＮ）组成。实验结果表明，该方法分割颗粒图像得到的边缘图像封闭性好、边界描述真实，适用于堆积颗粒物料图像的边缘检测。
In view of the characteristics of particle image, an edge detection method based on neural network is proposed. The method includes two parts: pixel extraction of boundary candidate and edge detection of neural network. Neural network is composed of self-organizing competitive sub-neural network (ASCSNN) used for image information compression and image information coding and radial basis Functional sub-neural network (RBFSNN). The experimental results show that the edge image obtained by the segmentation of particle image has good sealing property and real boundary description. It is suitable for the edge detection of the image of the accumulated particle material.