| 一种基于深度学习图像超分的环形靶标稳定检测方法 |
| |
| 引用本文： | 崔海华,徐振龙,杨亚鹏,孟亚云,王宝俊.一种基于深度学习图像超分的环形靶标稳定检测方法[J].中国机械工程,2021,32(23):2861-2867. |
| |
| 作者姓名： | 崔海华 徐振龙 杨亚鹏 孟亚云 王宝俊 |
| |
| 作者单位： | 1.南京航空航天大学机电学院，南京，211106
2.中航西安飞机工业集团股份有限公司制造工程部，西安，710089 |
| |
| 基金项目： | 国家重点研发计划（2019YFB2006100，2019YFB1707501）；
江苏省自然科学基金（BK20191280）；
南京航空航天大学研究生创新基地（实验室）开放基金（kfjj20200519） |
| |
| 摘 要： | 为提高远距离、大倾角条件下环形靶标的识别率与定位精度，提出了一种基于深度学习图像超分的环形靶标稳定检测方法。通过真实图像与合成图像的混合数据集来构建多角度、多距离的图像空间集合，采用像素损失与感知损失来改进图像超分辨率模型的损失函数，从而实现图像的高分辨率重建，丰富靶标轮廓的图像细节，利用已训练好的图像超分模型重建图像，最后使用传统的检测算法识别与定位环形靶标。实验结果表明，环形靶标识别率可提高40%，靶标定位精度可提高8.47%。
|
| 关 键 词： | 环形靶标 超分辨率 深度学习 目标识别 |
A Circular Target Stability Detection Method Based on Deep Learning Image Super-resolution |
| |
| CUI Haihua,XU Zhenlong,YANG Yapeng,MENG Yayun,WANG Baojun.A Circular Target Stability Detection Method Based on Deep Learning Image Super-resolution[J].China Mechanical Engineering,2021,32(23):2861-2867. |
| |
| Authors: | CUI Haihua XU Zhenlong YANG Yapeng MENG Yayun WANG Baojun |
| |
| Affiliation: | 1.College of Mechanical & Electrical Engineering，Nanjing University of Aeronautics and Astronautics，Nanjing，211106
2.Manufacturing Engineering Department，AVIC Xi'an Aircraft Industry Group Company Ltd.，Xi'an，710089 |
| |
| Abstract: | In order to improve the recognition rate and location accuracy of circular targets under the conditions of long-distance and large deflection angle， a circular target stability detection method was proposed based on deep learning image super-resolution. The multi-angle and multi-distance image sets were constructed through a hybrid data set of real images and synthetic images， the pixel loss and perceptual loss were used to improve the loss function of image super-resolution model， so the super-resolution reconstruction of images might be realized and the image details of target contours might be enriched. By using the pretrained super-resolution model， the images were reconstructed. Finally， traditional detection algorithm was used to recognize and locate the circular targets. The experimental results show that the circular target recognition rate is increased by 40%， and the target location accuracy is increased by 8.47%. |
| |
| Keywords: | circular target super-resolution deep learning object recognition |
|
| 点击此处可从《中国机械工程》浏览原始摘要信息 |
|
点击此处可从《中国机械工程》下载全文 |
