轻量化对抗增强的物流违规操作识别方法

目的 针对复杂强噪背景下物流违规操作难以有效识别的问题，提出一种轻量化对抗增强的物流违规操作检测方法。方法 以YOLOv5为基础框架，提出轻量化的GhostC3模块，运用对抗学习的思想提出轻量对抗模块，将原有结构中的C3模块修改为轻量化的GhostC3模块，Conv模块修改为轻量对抗模块，并将定位损失修改为CIOU损失。结果 通过实验验证可知，本文方法针对复杂强噪背景下物流违规操作具有优异的检测效果，其中本文方法相较于YOLOv5方法的检测平均精度均值提高了1.69%，模型参数量降低了45.14%，检测速度提高了2.46%。结论 本文提出的方法具有参数量低、检测速度快和精度高等特点，针对复杂强噪背景下物流违规操作的检测具有一定的先进性和实用性，充分满足物流违规操作检测需求。

Recognition Method of Improper Operation in Logistics with Lightweight and Enhanced Countermeasures

The work aims to propose a method of detecting improper operations in logistics with lightweight and enhanced countermeasures in view of the difficulty in effectively identifying improper operations in logistics under the background of complex and strong noise. Based on YOLOv5, the lightweight GhostC3 module was offered, and the lightweight countermeasure module was proposed with the idea of countermeasure learning. The C3 module in the original structure was changed into the lightweight GhostC3 module, the Conv module was changed into the LAconv module, and the positioning loss was changed into CIOU loss. Finally, through experimental verification, the method proposed had an excellent detection effect against improper operations under the background of complex and strong noise. Compared with YOLOv5, the average detection accuracy of the method proposed increased by 1.69%, the number of model parameters decreased by 45.14%, and the detection speed was improved by 2.46%. The method proposed has the characteristics of a low number of parameters, fast detection speed, and high accuracy. It is advanced and practical for the detection of improper operations in logistics under the background of complex and strong noise, and fully meets the detection needs of improper operations in logistics.