基于改进U-Net卷积神经网络的数字图像智能分类方法
An Intelligent Classification Method for Digital Images Based on an Improved U-Net Convolutional Neural Network
梅光1
作者信息
- 1. 南昌大学共青学院,江西九江 332020
- 折叠
摘要
环境噪声和干扰可能会对数字图像智能分类方法的性能产生负面影响,导致分类速度慢.为此,研究基于改进U-Net卷积神经网络的数字图像智能分类方法.通过去噪、增强和标准化处理,提高数字图像的质量;利用改进U-Net卷积神经网络提高预处理后的图像的维度,使网络能够学习到更丰富的图像信息;采用分割算法将图像划分为多个区域,通过关键点精确定位技术,准确识别出图像中的关键特征点;对比待分类图像与已知类别的图像相似度,实现智能分类.实验结果表明:与传统的分类方法相比,新方法在分类速度更快,实际应用价值更高.
Abstract
Environmental noise and interference may have a negative impact on the performance of intelligent digital image classification methods,leading to slow classification speed.To this end,a digital image intelligent classification method based on an improved U-Net convolutional neural network is studied.Improve the quality of digital images through denoising,enhance-ment,and standardization processing;Utilizing an improved U-Net convolutional neural net-work to enhance the dimensionality of preprocessed images,enabling the network to learn richer image information;Using segmentation algorithms to divide the image into multiple regions,u-sing precise keypoint localization techniques to accurately identify key feature points in the im-age;Compare the similarity between the image to be classified and the image of a known catego-ry to achieve intelligent classification.The experimental results show that compared with tradi-tional classification methods,the new method has faster classification speed and higher practical application value.
关键词
改进U-Net卷积神经网络/数字图像/智能分类/图像分类Key words
improving U-Net convolutional neural network/digital image/intelligent classifica-tion/image classi fication引用本文复制引用
出版年
2024
NETL
NSTL
万方数据