摘要
针对棉花中异纤检测率低的问题,提出了改进的FasterRCNN异纤目标检测网络。根据棉花异纤较小的特征,引入了残差网络作为特征提取网络,并将特征金字塔与残差网络结合;使用VGG16、ResNet50、ResNet50+FPN三种特征提取网络对3800张异纤图进行试验对比。结果表明:ResNet50+FPN特征提取网络的精确度和召回率为97.6%、82.4%,相比VGG16网络和ResNet50网络均有所提高。认为:基于改进的Faster RCNN棉花异纤识别能够满足检测精度和速度的要求。
Aimed at lower detection rate of foreign fiber in cotton,detection network aimed at foreign fiber of improved Faster RCNN was put forward.Based on smaller foreign fiber in cotton,residual network was introduced as feature extraction network.And feature pyramid and residual network were combined.Three kinds of feature recognition networks including VGG16,ResNet50 and ResNet50+FPN were used to test 3800 foreign fiber pictures.The test results showed that the accuracy and recall rate of ResNet50+FPN feature extraction network were 97.6%and 82.4%separately.The two data indicators were improved compared with VGG16 network and ResNet50 network.It is considered that the foreign fiber recognition by the improved Faster RCNN can meet the requirements on accuracy and speed.
作者
张越
张守京
冯中强
ZHANG Yue;ZHANG Shoujing;FENG Zhongqiang(Xi'an Polytechnic University,Xi'an,710048,China;Xi'an Key Laboratory of Modern Intelligent Textile Equipment,Xi'an,710600,China)
出处
《棉纺织技术》
CAS
北大核心
2022年第5期37-41,共5页
Cotton Textile Technology
基金
国家重点研发计划项目(2019YFB1707205)。
