摘要
基于内容的图像检索技术研究的是根据图像的自身底层特征,在图像库中智能地、快速地、准确地检索出相似或者相近的图像的方法。文章把图像的颜色和纹理特征作为训练样本集,采用添加了附加动量因子和自适应学习率的BP神经网络对样本集进行相应的训练,来提升图像检索的自适应与自学习能力,增强图像检索的容错率,从而提高图像检索的精度和效率。
Content-based image retrieval technology is related to more intelligent, swift and accurate ways of retrieving similar images from the image database in line with underlying features of the images.This paper takes the color and textual features of images as the sample set, which is then trained by BP neural network with additional momentum factor and adaptive learning rate. Thus, the adaptive and self-learning ability of image retrieval can be improved, its fault-tolerant rate be enhanced, and its accuracy and efficiency be increased.
引文
[1]章毓晋.基于内容的视觉信息检索[M].北京:科学出版社,2003.
[2]龙飞,王永兴.深度学习入门与实践[M].北京:清华大学出版社,2017.
[3]张娜,韩美林,王园园,杨琳.基于改进的BP神经网络的车牌识别技术研究[J].计算机与数字工程,2018. 46(10):2094-2098.
[4]朱振国,田松禄.基于权值变化的BP神经网络自适应学习率改进研究[J].计算机系统应用2018.27(7):205-210.
[5]吴陈,王和杰.基于改进的自适应遗传算法优化BP神经网络[J].电子设计工程2016.24(24):29-32