基于CNN-LSTM的实时空降空投状态识别系统设计  

Design of Real-time Airborne Airdrop State Recognition System Based on CNN-LSTM

作　　者：汤鸿源 王利恒[1] 张海龙 康超 TANG Hongyuan;WANG Liheng;ZHANG Hailong;KANG Chao(School of Electrical Information,Wuhan Institute of Technology,Wuhan 430205,China;No.722 Research Institute,China State Shipbuilding Corporation,Wuhan 430205,China)

机构地区：[1]武汉工程大学电气信息学院,武汉430205 [2]中国船舶集团有限公司第七二二所,武汉430205

出　　处：《自动化与仪表》2024年第12期56-60,共5页Automation & Instrumentation

基　　金：武汉工程大学研究生教育创新基金(CX2023555);东湖高新区“揭榜挂帅”项目(2024KJB305)。

摘　　要：在空降空投的许多应用场景中,需要装置自动识别状态并及时做出相应反应。然而,传统方法的性能可能存在限制,难以灵活地应对各种情况。为了解决这一问题,该文提出了一种基于卷积神经网络和长短期记忆网络结合(CNN-LSTM)的神经网络模型,并且将其部署于RV1106芯片,从而搭建实时空降空投状态识别系统。首先,采集空降空投过程的原始数据以及MATLAB仿真数据作为训练样本;然后,通过改进的CNN-LSTM算法训练出网络模型;最后,将网络模型部署到终端系统。通过多维度特征融合,该算法能够更精确地确定当前状态。实验结果表明,该文提出的方法比单独的LSTM算法识别率提高了3.1%,实用率提高61%,能够满足飞行与伞降状态识别的需求。In many application scenarios of airborne airdrops,devices are required to automatically recognize the status and react accordingly in a timely manner.However,there may be limitations in the performance of traditional methods that make it difficult to respond flexibly to various situations.To solve this problem,this paper proposes a neural network model based on the combination of convolutional neural network and long short-term memory network(CNN-LSTM)and deploys it on the RV1106 chip to build a real-time airborne airdrop state recognition system.First,the raw data of airborne drop process and MATLAB simulation data are collected as training samples,then the network model is trained by the improved CNN-LSTM algorithm,and finally the network model is deployed to the terminal system.Through multi-dimensional feature fusion,this algorithm can determine the current state more accurately.The experimental results show that the method proposed in this paper improves the recognition rate by 3.1%and the utility rate by 61%over the LSTM algorithm alone,which can meet the requirements of flight and parachute state recognition.

关 键 词：LSTM 实时 多维度特征融合 RV1106芯片 AI模型部署 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]