面向盲视群体的改进YOLOv3场景理解算法在交互景观设计中的应用研究  

Research on the Application of Improved YOLOv3 Scene Understanding Algorithm for Blind Groups in Interactive Landscape Design

作　　者：陈蕊[1] 张宁[1] 黄文华 CHEN Rui;ZHANG Ning;HUANG Wenhua(Xian Yang normal university,Shaanxi Xianyang 712000,China)

机构地区：[1]咸阳师范学院,陕西咸阳712000

出　　处：《自动化与仪器仪表》2024年第6期99-103,共5页Automation & Instrumentation

基　　金：陕西省社会科学基金年度项目《黄河流域陕西段传统村落文化景观的生态保护与活化传承研究》(2023J049);咸阳师范学院博士科研启动费(项目)(10502003613)。

摘　　要：为了提升盲视群体的社会融入能力和增强他们与社会的联系,研究利用改进YOLOv3场景理解算法与Kinect传感器来设计音乐交互景观体验系统。首先利用聚类方法对YOLOv3算法进行改进,提升其识别和分类能力;然后将YOLOv3算法与Kinect传感器进行融合,用于构建音乐交互设计中;最后利用对比方法和实际应用去验证音乐交互设计系统的性能。结果表明在系统参数结构中,YOLOv3+Kinect的性能均优于对比方法。其损失值、数据召回率、mAP%和数据测试平均耗时分别高出对比方法21.58%、15.13%,32.7%、4.37%,43.86%、34.21%,0.98 s、0.15 s。这说明研究构建的系统在音乐交互景观设计中具有较高的可行性和实用性,为盲视群体提供了一种全新的音乐交互体验。In order to enhance the social integration ability of blind individuals and enhance their connection with society,the study utilizes an improved YOLOv3 scene understanding algorithm and Kinect sensor to design a music interactive landscape experience system.Firstly,the YOLOv3 algorithm is improved using clustering methods to enhance its recognition and classification capabilities;Then,the YOLOv3 algorithm is fused with Kinect sensors to construct music interaction designs;Finally,use comparative methods and practical applications to verify the performance of the music interaction design system.The results indicate that YOLOv3+Kinect outperforms the comparison method in terms of system parameter structure.Its loss value,data recall rate,mAP%,and average data testing time are 21.58%,15.13%,32.7%,4.37%,43.86%,34.21%,0.98 seconds,and 0.15 seconds higher than the comparison method,respectively.This indicates that the system constructed in the study has high feasibility and practicality in the design of music interaction landscapes,providing a brand new music interaction experience for blind groups.

关 键 词：盲视群体 YOLOv3算法 音乐 交互景观 Kinect传感器 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]