微型电脑应用2025，Vol.41Issue(9)：31-35,40,6.
基于ResNet50的棉花颜色级分类方法
Color Grade Classification Method for Cotton Based on ResNet50
摘要
Abstract
The traditional cotton color grade classification method exists some problems,such as high classification cost and slow classification speed,a prediction method based on convolutional neural network is proposed.The proposed method uses convolutional neural network to improve the classification speed and introduces an attention mechanism to improve the classifi-cation accuracy.Due to the lack of a publicly available cotton sample dataset,this paper uses a CMOS camera to take images of cotton with different conditions and different categories,and after enhancement,the dataset is divided into training and test sets with a 3∶2 ratio.The proposed cotton color level prediction method based on feature attention mechanism is trained and tested on the proposed dataset.The inference speed of the model is optimized by using TensorRT,and the optimized model is de-ployed to the Jetson Nano edge computing platform.Experimental results show that the convolutional neural network introdu-cing the feature attention mechanism improves the accuracy of 3.94％compared with the original network.The average running time of the model optimized by using TensorRT on Jetson Nano is 0.2379 s,which meets the needs of practical application scenarios.关键词
棉花颜色等级/深度学习/图像处理/注意力机制/卷积神经网络Key words
cotton color grade/deep learning/image processing/attention mechanism/convolutional neural network分类
轻工业引用本文复制引用
董飞龙,支双双,高润超..基于ResNet50的棉花颜色级分类方法[J].微型电脑应用,2025,41(9):31-35,40,6.基金项目
国家自然科学基金(61502537) （61502537）
陕西省2023年重点研发计划(2023-YBSF-203) （2023-YBSF-203）
西安工程大学大学生创新创业训练计划项目(202110709041) （202110709041）
