Article Details

Title · Authors · Abstract · Keywords · DOI · References

Optimization and Comparative Verification of Target Recognition Algorithm Based on YOLO

LI Zewang, MA Xiaodan

(The People's Liberation Army Rocket Force University of Engineering, Xi'an 710025, China)

Section：Computer Technology · Vol. 2026, Issue 20 · Pages：111-117 · 2096-4706（2026）20-0111-07 · CLC number：TP391.4 · Document code：A · DOI：10.19850/j.cnki.2096-4706.2026.20.020

Abstract

As intelligent warfare becomes the key to victory on future battlefields, embedded development systems play a crucial role in improving military operational efficiency. Therefore, this study proposes a lightweight and high-precision target detection algorithm based on Raspberry Pi. The proposed algorithm replaces the C3 module of YOLOv5s with the lightweight C3Ghost module and adds the CA attention mechanism to the backbone network. It also adopts TensorFlow Lite for model inference and deploys the model on Raspberry Pi. Comparative experiments are conducted on a custom dataset. Test results indicate that, on embedded devices, the mAP@0.5 and mAP@0.5:0.95 of the optimized algorithm are increased by 0.8% and 0.5% respectively compared with the original YOLOv5s. The memory occupation is greatly reduced and the FPS is significantly improved. The proposed algorithm maintains a good balance between accuracy and speed and exhibits comprehensive advantages over mainstream models such as YOLOv8 and YOLOv26.

Keywords

target recognition; YOLOv5s; raspberry pi; CA Attention Mechanism; C3Ghost

References

  1. [1] Wang Qi, Guo Hao, Zhou Qingyang, etc The Impact and Response Strategies of Informationization, Unmanned, and Intelligent Warfare in Modern Warfare [J]. Military Abstracts, 2026 (1): 7-11
  2. [2] Mizeng, Lian Zhe A review of YOLO methods for general object detection [J]. Computer Engineering and Applications, 2024, 60 (21): 38-54
  3. [3] Xue Peiyou, Liu Lei, Liu Ruiyang, etc Vehicle detection based on Raspberry Pi and YOLO algorithm [J]. Journal of Liaocheng University (Natural Science Edition), 2025, 38 (6): 840-852
  4. [4] Qin Weicai, Qin Peiliang, Liang Zhenglong A lightweight strawberry maturity detection algorithm based on improved YOLOv5s [J]. China Agricultural Machinery Chemistry Journal, 2026, 47 (2): 86-93
  5. [5] Cheng Jiapei, Wang Xusheng, Lin Qibin, etc Improving pedestrian and vehicle multi-target recognition, tracking and detection under YOLOv5 [J]. Modern Information Technology, 2025, 9 (17): 73-77, 82
  6. [6] HAN K，WANG Y，TIAN Q，et al. GhostNet: More Features from Cheap Operations [C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition，2020：1580-1589.
  7. [7] Meng Shuai Research on Traffic Sign Detection and Recognition Based on Improved YOLOv5s [D]. Dalian: Dalian Jiaotong University, 2023
  8. [8] Lin Xianghui, Xie Benliang Real time object detection based on TensorFlow Lite platform [J]. Intelligent Computers and Applications, 2021, 11 (2): 80-83, 87
  9. [9] Mao Xinrong, Xu Xiao Small Object Detection Algorithm for Aerial Images Based on Improved YOLOv8s [J]. Microelectronics and Computer, 2026, 43 (3): 75-87
  10. [10] Liu Yingbo, Ma Shuai Qi, Cheng Huan Bird target detection algorithm based on improved YOLOv11n [J]. Advances in Laser and Optoelectronics, 2026, 63 (16): 226-239

About the authors

Li Zewang (2000-), male, Han ethnicity, from Handan, Hebei, undergraduate student, research interests include command information system engineering; Ma Xiaodan (1985-), female, Han ethnicity, from Yinchuan, Ningxia, associate professor, master's student, research interests include AI data science.