基于改进ShuffleNetV2的织物颜色恒常性算法  

Cloth Color Constancy Algorithm Based on Improved ShuffleNetV2

作　　者：杨必成 张团善[1] YANG Bicheng;ZHANG Tuanshan(School of Mechanical and Electrical Engineering,Xi'an Polytechnic University,Xi'an 710048,China)

机构地区：[1]西安工程大学机电工程学院,陕西西安710048

出　　处：《沈阳大学学报（自然科学版）》2023年第3期216-223,272,F0002,F0003,共11页Journal of Shenyang University：Natural Science

基　　金：国家自然科学基金资助项目(51735010);西安现代智能纺织设备重点实验室项目(2019220614SYS021CG043)。

摘　　要：针对工业生产中织物色差检验存在色偏、部署在计算机资源有限的移动终端的算法适配性不强、实时准确校正织物的色偏准确率不高的问题,提出一种基于改进ShuffleNetV2的轻量级织物颜色恒常性算法。以ShuffleNetV2的框架为基础,使用H-Sigmoid激活函数代替ReLU激活函数,引入注意力机制,使用多通道置信加权估计出全局场景光源。然后,在Gehler-shi和NUS-8数据集上的进行实验。实验结果表明,相较于已有的颜色恒常性算法,所提的轻量级网络各项评价指标提高约0.2,可用于织物的颜色校正任务中。Aiming at the problems of color shift in fabric color aberration inspection in industrial production,weak adaptability of algorithms deployed in mobile terminals with limited computer resources,and low accuracy of real-time accurate correction of fabric color cast,a lightweight fabric color performance algorithm based on improved ShuffleNetV2 was proposed.Based on the framework of ShuffleNetV2,the H-Sigmoid activation function was used instead of the ReLU activation function,the attention mechanism was introduced,and the global scene light source was estimated using multi-channel confidence weighting.Then,experiments were performed on the Gehler-shi and NUS-8 datasets.Experimental results show that compared with the existing color constancy algorithm,the evaluation index of the proposed lightweight network is improved by about 0.2,which can be used in the color correction task of fabrics.

关 键 词：颜色恒常性 轻量级 ShuffleNetV2 通道注意力机制 颜色校正 

分 类 号：TP312[自动化与计算机技术—计算机软件与理论]