改进YOLOv7的城市小型无人机目标检测方法

doi:10.3778/j.issn.1002-8331.2308-0196

摘要/Abstract

摘要： 针对“低小动”无人机的反制技术已成为低空空域安全防御的重要手段，然而实时检测与准确识别是实施有效反制的前提条件与关键基础。针对城市低空环境下，目标检测算法对不同背景下小尺度无人机目标检测精度低，容易出现漏检误检且易受外界因素干扰等问题，提出了一种基于改进YOLOv7的“低小动”无人机目标检测方法。首先采集大量不同环境、不同背景下的无人机样本构建数据集，并采用ViBe（visual background extractor）算法进行预处理；其次引入坐标注意力机制与SPDConv（space-to-depth convolution）模块改进和优化YOLOv7的网络结构；最后提出融合ViBe和改进YOLOv7的二级检测架构，将改进后的YOLOv7作为网络模型检测经ViBe处理后的图像。依据原图与处理图像的位置大小关系，将检测出的目标坐标映射回归至原图片，从而完成目标检测提取。实验结果表明，所提目标检测方法检测精度达96.5%，较原YOLOv7方法提高了15.8个百分点，显著提升了“低小动”目标的检测精度，能够满足低空无人机的实时精准检测的需求。

关键词: ViBe算法, 反无人机, YOLOv7, 坐标注意力机制, 小目标检测, SPDConv

Abstract: Countermeasures against “low and small moving” UAVs have become an important tool for low altitude airspace security defense, but real-time detection and accurate identification are the prerequisite and key foundation for effective countermeasures. Aiming at the urban low-altitude environment, the target detection algorithm has low accuracy in detecting small-scale UAV targets in different backgrounds, is prone to omission and misdetection, and is susceptible to interference from external factors, etc., a “low and small moving” UAV target detection method based on the improved YOLOv7 is proposed. Firstly, a large number of UAV samples from different environments and backgrounds are collected to build a data set and are pre-processed by ViBe (visual background extractor) algorithm. Secondly, the coordinate attention mechanism and SPDConv (space-to-depth convolution) module are introduced to improve and optimize the network structure of YOLOv7. Finally, a secondary detection architecture is proposed to fuse ViBe and improved YOLOv7, and the improved YOLOv7 is used as the network model to detect the images processed by ViBe. Based on the position size relationship between the original image and the processed image, the detected target coordinates are mapped back to the original image, so as to complete the target detection and extraction. The experimental results show that the detection accuracy of the proposed target detection method reaches 96.5%, which is 15.8?percentage points higher than that of the original YOLOv7 method, significantly improving the detection accuracy of “low and small moving” targets and meeting the demand for real-time accurate detection of low-altitude UAVs.

Key words: ViBe algorithm, anti-drone, YOLOv7, coordinate attention mechanism, small target detection, SPDConv

崔勇强, 李嘉轩, 侯林果, 梅涛, 白迪, 陈少平. 改进YOLOv7的城市小型无人机目标检测方法[J]. 计算机工程与应用, 2024, 60(10): 237-245.

CUI Yongqiang, LI Jiaxuan, HOU Linguo, MEI Tao, BAI Di, CHEN Shaoping. Improved YOLOv7 Target Detection Method for Small Urban UAVs[J]. Computer Engineering and Applications, 2024, 60(10): 237-245.

参考文献

[1] KAPOULAS I K, HATZIEFREMIDIS A, BALDOUKAS A K, et al. Small fixed-wing UAV radar cross-section signature investigation and detection and classification of distance estimation using realistic parameters of a commercial anti-drone system[J]. Drones, 2023, 7(1): 39.
[2] 罗俊海, 王芝燕. 无人机探测与对抗技术发展及应用综述[J]. 控制与决策, 2022, 37(3): 530-544.
LUO J H, WANG Z Y. A review of development and application of UAV detection and counter technology[J]. Control and Decision, 2022, 37(3): 530-544.
[3] DIDKOVSKYI V, KOZERUK S, KORZHIK O. Simple acoustic array for small UAV detection[C]//Proceedings of the 2019 IEEE 39th International Conference on Electronics and Nano-technology, 2019: 656-659.
[4] VAN NGUYEN H, REZATOFIGHI H, VO B N, et al. Online UAV path planning for joint detection and tracking of multiple radio-tagged objects[J]. IEEE Transactions on Signal Processing, 2019, 67(20): 5365-5379.
[5] AI X, ZHANG L, ZHENG Y, et al. Passive detection experiment of UAV based on 5G new radio signal[C]//Proceedings of the 2021 Photonics & Electromagnetics Research Symposium, 2021: 2124-2129.
[6] REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: unified, real-time object detection[C]//Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition, 2016: 779-788.
[7] REDMON J, FARHADI A. YOLO9000: better, faster, stronger[C]//Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition, 2017: 7263-7271.
[8] REDMON J,?FARHADI A. YOLOv3: an incremental improvement[J]. arXiv:1804.02767, 2018.
[9] BOCHKOVSKIY A, WANG C Y, LIAO H Y M. YOLOv4: optimal speed and accuracy of object detection[J]. arXiv:2004.10934, 2020.
[10] WANG C Y, BOCHKOVSKIY A, LIAO H Y M.YOLOv7: trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[C]//Proceedings of the 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023: 7464-7475.
[11] 戚玲珑, 高建瓴.基于改进YOLOv7的小目标检测[J].计算机工程, 2023, 49(1): 41-48.
QI L L, GAO J L. Small object detection based on improved YOLOv7[J]. Computer Engineering and Applications, 2023, 49(1): 41-48.
[12] 齐向明, 柴蕊, 高一萌. 重构SPPCSPC与优化下采样的小目标检测算法[J]. 计算机工程与应用, 2023, 59(20): 158-166.
QI X M, CHAI R, GAO Y M. Algorithm of reconstructed SPPCSPC and optimized downsampling for small object detection[J]. Computer Engineering and Applications, 2023, 59(20): 158-166.
[13] BARNICH O, VAN DROOGENBROECK M. ViBE: a powerful random technique to estimate the background in video sequences[C]//Proceedings of the 2009 IEEE International Conference on Acoustics, Speech and Signal Processing, 2009: 945-948.
[14] SUNKARA R, LUO T. No more strided convolutions or pooling: a new CNN building block for low-resolution images and small objects[C]//Proceedings of the 2022 Joint European Conference on Machine Learning and Knowledge Discovery in Databases. Cham: Springer, 2022: 443-459.
[15] HOU Q, ZHOU D, FENG J. Coordinate attention for efficient mobile network design[C]//Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021: 13713-13722.
[16] 成浪, 敬超. 基于改进YOLOv7的X线图像旋转目标检测[J]. 图学学报, 2023, 44(2): 324-334.
CHENG L, JING C. X-ray image rotating object detection based on improved YOLOv7[J]. Journal of Graphics, 2023, 44(2): 324-334.
[17] 汤旻安, 王晨雨. 基于改进ViBe算法的静态场景运动目标检测[J]. 激光与光电子学进展, 2021, 58(14): 216-224.
TANG M A, WANG C Y. Moving object detection in static scene based on improved vibe algorithm[J]. Laser & Optoelectronics Progress, 2021, 58(14): 216-224.
[18] 赵元龙, 单玉刚, 袁杰. 改进YOLOv7与DeepSORT的佩戴口罩行人跟踪[J]. 计算机工程与应用, 2023, 59(6): 221-230.
ZHAO Y L, SHAN Y G, YUAN J. Wearing mask pedestrian tracking based on improved YOLOv7 and deepsort[J]. Computer Engineering and Applications, 2023, 59(6): 221-230.
[19] LIU W, ANGUELOV D, ERHAN D, et al. SSD: single shot multibox detector[C]//Proceedings of the 14th European Conference on Computer Vision, Amsterdam, Oct 11-14, 2016. Cham: Springer, 2016: 21-37.
[20] REN S, HE K, GIRSHICK R, et al. Faster R-CNN: towards realtime object detection with region proposal networks[C]//Advances in Neural Information Processing Systems 28, 2015.