您好,欢迎来到login_星欧(官方)代理招商直营娱乐!

首页> 《中国测试》期刊 >本期导读>改进型卷积神经网络无人机工地识别

改进型卷积神经网络无人机工地识别

348    2024-06-26

¥0.50

全文售价

作者:潘昱辰1, 徐浩1, 钱夔1, 徐伟敏2, 徐腾飞21. 南京工程学院自动化学院,江苏 南京 210044;
2. 南京睿捷智慧交通科技研究院,江苏 南京 210044

作者单位:1. 南京工程学院自动化学院,江苏 南京 210044;
2. 南京睿捷智慧交通科技研究院,江苏 南京 210044


关键词:工地识别;卷积神经网络;损失函数;数据增强


摘要:

现有通用目标检测算法在无人机工地识别任务中容易产生精度低下等问题,针对该问题,该文提出一种卷积神经网络模型,用于复杂环境下相似目标检测。该模型首先利用无人机高空拍摄图片作为数据集,通过高斯模糊、图像变换等方法进行数据增强,为模型泛化能力的提高提供数据支撑。然后基于Darknet-53特征提取网络实现多尺度特征融合,通过在网络模型中添加SPP-net(spatial pyramid pooling networks)应对模型中特征易消失问题。最后优化损失函数,解决模型正负样本不均衡问题。实验结果证明该模型mAP值达到84.94%,可为城市内土地规划、施工和违章搭建监管等领域提供技术支撑。


Construction site recognition of UAV based on improved convolutional neural network
PAN Yuchen1, XU Hao1, QIAN Kui1, XU Weimin2, XU Tengfei2
1. School of Automation, Nanjing Institute of Technology, Nanjing 210044, China;
2. Nanjing Ruijie Intelligent Transportation Technology Research Institute, Nanjing 210044, China
Abstract: Existing general target detection algorithms are prone to produce low accuracy problems in UAV site recognition tasks. To solve this problem, this paper proposes a convolutional neural network model for similar target detection in complex environments. Firstly, aerial images taken by UAV are used as the data set, and the data are enhanced by Gaussian blur and image transformation, which provide data support for improving the generalization ability of the model. Then, multi-scale feature fusion is realized based on Darknet-53 feature extraction network, and SPP-net (spatial pyramid pooling networks) is added to the network model to solve the problem of feature disappearing easily. Finally, the loss function is optimized to solve the imbalance of positive and negative samples. Experimental results show that the mAP value of the model reaches 84.94%, which can provide technical support for urban land planning, construction and supervision of illegal construction.
Keywords: construction site recognition; convolutional neural network; loss function; data enhancement
2024, 50(6):191-196 收稿日期: 2022-05-07;收到修改稿日期: 2022-07-13
基金项目:
作者简介: 潘昱辰(1999-),女,江苏徐州市人,硕士研究生,专业方向为机器视觉。
参考文献
[1] GUO Y, LIU Y, OERLEMANS A, et al. Deep learning for visual understanding: A review[J]. Neurocomputing, 2016, 187: 27-48.
[2] QIAN K, TIAN L. A topic-based multi-channel attention model under hybrid mode for image caption[J]. Neural Comput & Applic, 2022: 2207–2216.
[3] 刘桂雄, 张瑜, 蔡柳依婷. 机器视觉检测图像拼接融合技术研究进展[J]. 中国测试, 2020, 46(1): 1-6.
LIU G X, ZHANG Y, CAI L Y T. Overview of image blending technique in machine vision detection[J]. China Measurement & Test, 2020, 46(1): 1-6.
[4] 郭雨青, 曾庆军, 夏楠, 等. 图像增强水下自主机器人目标识别研究[J]. 中国测试, 2021, 47(11): 47-52.
GUO Y Q, ZENG Q J, XIA N, et al. Research on target recognition of autonomous underwater vehicle based on image enhancement[J]. China Measurement & Test, 2021, 47(11): 47-52.
[5] GIRSHICK R, DONAHUE J, DARRELL T, et al. Richfeature hierarchies for accurate object detection and semantic segmentation[C]. // IEEE Conference on Computer Vision and Pattern Recognition. IEEE Computer Society, 2014: 580-587.
[6] GIRSHICK R. Fast R-CNN[J]. Computer Science, 2015: 34-37.
[7] LIU W, ANGUELOV D, ERHAN D, et al. SSD: single shot multibox detector[C]//Computer vision-ECCV2016. Amsterdam. The Netherlands: Springer, 2016: 21-37.
[8] REDMON J, FARHADI A. YOLO9000: Better, faster, stronger[J/OL] . CoRR, 2016: 14-26. https://arxiv.org/abs/ 1612.08242
[9] 王鑫, 陈浩, 陶天宝. 一种无人机图像车辆识别系统设计[J]. 电子世界, 2021(18): 121-123.
[10] 车梦凡. 基于机器学习的无人机图像目标识别算法研究[D]. 金华: 浙江师范大学, 2021.
CHE M F. Research on UAV image target recognition algorithm based on machine learning[D]. Jinhua: Zhejiang Normal University, 2021.
[11] HUANG G, LIU S, VAN DER MAATEN L, et al. Condensenet: An efficient densenet using learned group convolutions[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018: 2752-2761.
[12] 王瑶. 基于视觉的无人机目标识别及跟踪[D]. 哈尔滨: 哈尔滨工业大学, 2019.
WANG Y. Vision-based UAV target recognition and tracking[D]. Harbin: Harbin Institute of Technology, 2019.
[13] 林海, 庞妙珍, 刘天成. 基于改进DarkNet网络的轻量型车型识别方法[J]. 现代计算机, 2021, 27(24): 100-104.
LIN H, PANG M Z, LIU T C. Lightweight vehicle identification method based on improved DarkNet network[J]. Modern Computer, 2021, 27(24): 100-104.
[14] BOCHKOVSKIY A, WANG C Y, LIAO H Y M. YOLOv4: optimal speed and accuracy of object detection[J/OL]. arXiv preprint, arXiv: 2004. 10934[2020-12-19].https://arxiv.org/ abs/2004.10934
[15] HE K M, ZHANG X Y, REN S Q, et al. Spatial pyramid pooling in deep convolutional networks for visual recognition[J]. IEEE Trans Pattern Anal Mach Intell, 2015, 37(9): 1904-1916.
[16] 刘桂雄, 刘思洋, 吴俊芳, 等. 基于深度学习的机器视觉目标检测算法及在票据检测中应用[J]. 中国测试, 2019, 45(5): 1-9.
LIU G X, LIU S Y, WU F J, et al. Machine vision object detection algorithm based on deep learning and application in banknote detection[J]. China Measurement & Test, 2019, 45(5): 1-9.
[17] 程海博, 熊显名. 基于GIoU的YOLOv3车辆识 别方法[J]. 桂林电子科技大学学报, 2020, 40(5): 429-433.
CHENG H B, XIONG X M. YOLOv3 vehicle recognition method based on GIoU[J]. Journal of Guilin University of Electronic Technology, 2020, 40(5): 429-433.
[18] 宋建辉, 饶威, 于洋, 等. 基于Focal Loss的多特征融合地物小目标检测[J]. 火力与指挥控制, 2021, 46(1): 20-24.
SONG J H, RAO W, YU Y, et al. Multi feature fusion small target detection based on Focal Loss[J]. Fire Control & Command Control, 2021, 46(1): 20-24.
[19] 徐义鎏, 贺鹏. 改进损失函数的Yolov3车型检测算法[J]. 信息通信, 2019(12): 4-7.
XU Y L, HE P. Yolov3 vehicle detection algorithm with improved loss function[J]. Changjiang Information & Communications, 2019(12): 4-7.