基于小波变换和PSO-LSTM的智慧教学机器人抓取识别方法  

Intelligent teaching robot grasping recognition method based on wavelet transform and PSO-LSTM

作　　者：徐文[1,2] 李婷 XU Wen;LI Ting(Xi’an Siyuan University,Xi’an 710038,China;International college,National Institute of Development Administration,Bangkok,10240,Thailand)

机构地区：[1]西安思源学院,西安710038 [2]泰国国家发展管理学院,泰国曼谷10240

出　　处：《自动化与仪器仪表》2025年第3期149-153,共5页Automation & Instrumentation

基　　金：西安思源学院2023年校长基金科研项目《新疆南疆地区基础教育发展研究》(XASYZX-MGJ2302)。

摘　　要：针对传统教学机器人抓取识别精度低,识别效率不高的问题,提出一种基于小波变换与粒子群算法(Particle Swarm Optimization algorithm,PSO)优化长短时记忆神经网络(Long Short-term Memory Networks,LSTM)的智慧教学机器人抓取识别方法。首先,采用小波变换方法对物体移动信号进行特征提取;然后以LSTM神经网络作为基础识别网络,并采用PSO对LSTM神经网络进行优化,搭建一个基于PSO-LSTM的智慧教学机器人抓取识别模型;最后将提取特征输入至该模型中进行抓取识别。实验结果表明,本方法的抓取识别mAP分别取值为96.84%,相较于传统的SURF抓取识别方法和YOLOv5抓取识别方法,mAP分别高出了17.46%、19.04%,且本方法的抓取识别所用时间仅为8.46 s,对比于另外两种方法分别降低了13.59 s和21.17 s。由此说明,本方法能够提高抓取识别精度和效率,可为智慧教学提供参考借鉴。In view of the problems of low accuracy and low recognition efficiency of traditional teaching robots,an intelligent teaching robot grasping recognition method based on wavelet transform and particle swarm algorithm(Particle Swarm Optimization algorithm,PSO) to optimize long and short time memory neural network(Long Short-term Memory Networks,LSTM) is proposed.First,adopt the wavelet transform method;then use the LSTM neural network as the basic recognition network,and use PSO to build a grasping recognition model based on PSO-LSTM;finally,the extracted features are input to the model for grasping recognition.The experimental results show that the grasp recognition mAP of this method is 96.84%,compared with the traditional SURF grasp recognition method and YOLOv5 scratch recognition method,mAP is 17.46% and 19.04% higher,respectively,and the grasp recognition time of this method is only 8.46 s,which reduces 13.59 s and 21.17 s compared with the other two methods.This shows that this method can improve the accuracy and efficiency of grasping recognition,and can provide reference for intelligent teaching.

关 键 词：智慧教学 小波变换 粒子群优化算法 LSTM神经网络 抓取识别 

分 类 号：TP392[自动化与计算机技术—计算机应用技术]