基于Adaboost-BP神经网络模型的姿势识别研究

打开文本图片集

摘要：姿势识别是计算机视觉中的重要研究方向。本文基于BP神经网络模型构建姿势识别模型，并采用Adaboost迭代学习对BP神经网络模型的预测效果进行提升。在双手叉腰、单臂张开、跑步与散步等动作的识别上，比CNN卷积模型具有更好的效果，并且比直接采用BP神经网络模型能够更为精确地识别各类姿势。

关键词：机器视觉;姿势识别;Adaboost-BP模型

doi：10.3969/J.ISSN.1672-7274.2022.06.017

中图分类号：TM 715                 文献标示码：A                文章编码：1672-7274（2022）06-00-03

Research on Pose Recognition based on AdaBoost BP Neural Network Model

ZhU Rui1， XUE Wenhua 2， LI Wenai 3， WU Zhaoqi 4， QIU Xue1

（1. Chengdu Jiabaili Technology Co.， Ltd.， Chengdu 610000，China; 2. Pitong No. 1 primary school， Pidu District， Chengdu 611730，China; 3. West Branch of Chengdu Caotang primary school， Chengdu 610073，China; 4. No. 8 primary school， Karamay 834000，China）

Absrtact： Pose recognition is an important research direction in computer vision. This paper constructs a pose recognition model based on BP neural network model， and uses AdaBoost iterative learning to improve the prediction effect of BP neural network model. It has better effect than CNN convolution model in the recognition of movements such as hands on hips， one arm opening， running and walking， and can recognize all kinds of postures more accurately than directly using BP neural network model.

Key words： machine vision; posture recognition; Adaboost BP model

1  研究背景

姿势识别是计算机研究领域的重要问题，在现实中具有广泛的应用场景。（剩余3987字）