基于改进C3D模型的料仓视频分类识别方法  

Video Recognition Method for Silo Target Based on Improved C3D Model

作　　者：曹庆园 朱建鸿 CAO Qingyuan;ZHU Jianhong(Key Laboratory of Advanced Process Control for Light Industry,Jiangnan University,Wuxi 214122,China)

机构地区：[1]江南大学轻工过程先进控制教育部重点实验室,江苏无锡214122

出　　处：《计算机测量与控制》2025年第2期161-167,183,共8页Computer Measurement &Control

基　　金：国家自然科学基金项目(61973139)。

摘　　要：在自动上料控制系统中,针对传统电感式传感器容易受到外界复杂环境干扰,且需要进行繁琐校准工作等问题,提出了一种基于改进C3D模型的料仓视频视觉分类识别方法;基于实验需求,设计了合作标靶和建立了料仓识别视频数据集;将初始C3D模型作为主干网络进行改进,将该模型第3、4、5层卷积层进行精简,使得模型参数量大幅降低,有利于加快推理速度;在轻量化后的C3D模型上融合SE注意力机制,C3D模型从时空两个维度中提取特征,SE注意力机制可以有效在复杂场景视频帧中找出标靶显著区域,在兼顾时序信息的同时能够高效提取特征进而提高识别准确率;实验结果表明,SE-C3D识别模型准确率达到99.61%,与初始C3D模型相比,准确率提高2.48%,与其他典型三维卷积模型对比,各项性能指标也均有明显提升,对未来智能化上料系统的发展具有重要意义。In automatic feeding control systems,in order to solve the problems that traditional inductive sensors are easy to be disturbed by external complex environments and need to be frequently calibrated,a visual classification and recognition method for silo video based on improved C3D model is proposed.Based on experimental requirements,a cooperative target is designed,and a video dataset for silo identification is established.The initial C3D model is improved as the backbone network,and the 3rd,4th,and 5th convolutional layers of the model are simplified,which greatly reduces the model parameters,and it is beneficial for speeding up the inference speed.A SE attention mechanism can effectively be fused on the lightweight C3D model,the features of the C3D model are extracted from the dimensions of time and space.The SE attention mechanism can efficiently find out the area of the target in the complex scene video frames,while taking into account the time series information to improve the recognition accuracy.Experimental results show that the accuracy of the SE-C3D recognition model reaches 99.61%,which is 2.48%higher than that of the initial C3D model.Compared with other typical 3D convolution models,this model also significantly improves the performance indicators,which is of great significance for the development of intelligent feeding systems in the future.

关 键 词：上料系统 3D卷积神经网络 视频分类 SE注意力 模型轻量化 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]