本文提出了一种基于随机场特性的神经元网络算法,以实现皮革纹理图象的最佳分类。该算法包括SAR随机场自适应建模、聚类特征的最优统计度量、三层B-P神经元网络与知识学习机制四部分,有效地表达了基元非规则性强,聚类分离程度低和传统纹理分析技术不易描述的皮革纹理聚类特性,有助于系统对人类专家经验性知识的有效获取,已应用于羊皮皮革纹理的计算机辅助分类。 In this paper, a neural network algorithm based on random field characteristics is proposed to achieve the best classification of leather texture images. The algorithm includes four parts: SAR random field adaptive modeling, optimal statistical measure of clustering features, three-layer BP neural network and knowledge learning mechanism, effectively expressing the strong non-regularity of primitives and the low degree of clustering separation Traditional texture analysis technology is not easy to describe the characteristics of leather texture clustering, contribute to the system of human experts to gain an empirical knowledge effectively, has been applied to the sheepskin leather texture computer-aided classification.