Military Target Detection Method Based on Improved YOLOv5

• Published in: 2022 International Conference on Cyber-Physical Social Intelligence (ICCSI) pp. 53 - 57
• Main Authors: Du, Xiuli, Song, Linkai, Lv, Yana, Qin, Xutong
• Format: Conference Proceeding
• Language: English
• Published: IEEE 18.11.2022
• Subjects: CBAM; Computational modeling; Decision making; Feature extraction; Military computing; military target detection; MobileNetV2; Object detection; Reconnaissance; Stem block module; Weapons; YOLOv5
• DOI: 10.1109/ICCSI55536.2022.9970675

Aiming at the requirement of military target detection under the condition of limited resources of weapon hardware platform, this paper proposes a military target detection method that takes into account network lightweight, mean average precision (mAP) and detection speed. This method is based on the You Only Look Once Version 5 (YOLOv5) algorithm. First, the Stem block module is used to replace the Focus module, which can effectively improve the feature expression ability and reduce the amount of parameters and computation of the network model. Second, a MobileNetV2-Convolutional Block Attention Module (MNtV2-CBAM) structure is designed with MobileNetV2 integrated into the CBAM mechanism. The amount of network parameters and computation is reduced, while the detection performance of the model is improved. The experimental results show that compared with the YOLOv5 algorithm, the mAP value of the method in this paper is increased by 1.3%, and the amount of parameters and the amount of calculation are decreased by 67.45% and 73.17% respectively, which can be better applied to the resource-constrained weapon equipment platform. In this way, the reconnaissance and analysis capabilities of military intelligence can be improved, the decision-making time of the commander can be shortened, and the combat capability of the troops can be greatly improved.