摘要
针对中国古代小说图像的特性,提出了一种利用人工神经网络获取图像语义的方法。利用人工神经网络在图像的纹理、形状等低层视觉特征和高层语义特征间建立映射关系,利用改进的遗传算法确定人工神经网络的参数和权值,利用训练后的人工神经网络获取图像的语义。实验结果表明,所提出的方法具有理想的图像语义获取效果,能充分反映人对图像内容的理解,具有很好的应用价值。
For the character of Chinese ancient storybook image, a semantic acquisition approach is proposed in this paper. By neural network, the mapping between the primitive image features, such as textures and shapes, and semantic features is established. Through ap- plying improve genetic algorithm, some parameters and weights of neural network may be obtained. The trained neural network is used to ac- quire semantic of Chinese ancient storybook image. Experimental results show that the proposed approach has ideal effective, which can re- fleet understand for image content, and it has very good application value.
出处
《计算机与数字工程》
2012年第5期95-97,共3页
Computer & Digital Engineering
基金
教育部人文社会科学研究规划基金项目(No.11YJAZH040)资助
关键词
神经网络
图像语义
遗传算法
语义获取
neural network, image semantic, genetic algorithm, semantic acquisition