仿真假体视觉下的空间导航研究-《中国医学物理学杂志》

文章信息/Info

Title:: Spatial navigation in simulated prosthetic vision

文章编号:: 1005-202X（2026）07-0879-10

作者:: 崔晓晨1; 赵瑛1; 张岩2; 代婷婷1; 1.内蒙古科技大学数智产业学院，内蒙古包头 014010；2.齐鲁理工学院计算机与信息工程学院，山东济南 250200

Author(s):: CUI Xiaochen1; ZHAO Ying1; ZHANG Yan2; DAI Tingting1; 1. School of Digital and Intelligent Industry, Inner Mongolia University of Science and Technology, Baotou 014010, China 2. School of Computer and Information Engineering, Qilu Institute of Technology, Jinan 250200, China

关键词:: 仿真假体视觉; 空间导航; 场景认知; 路线回溯; 图像处理策略

Keywords:: Keywords: simulated prosthetic vision spatial navigation scene cognition route-retracing image processing strategy

分类号:: R318.18；TP391.9

DOI:: DOI:10.3969/j.issn.1005-202X.2026.07.007

文献标志码:: A

摘要:: 目的：为帮助视觉假体佩戴者在有限视野下探索陌生环境，本文提出仿真假体视觉下的图像优化处理策略，并开展相应的空间导航任务研究。方法：基于Unity并借助感知深度融合语义分割与边缘增强表示的图像处理策略构建出仿真的假体视觉下的城市场景。招募20名被试作为研究对象，在仿真假体视觉下的虚拟城市场景中完成路线回溯任务与图片、序列识别任务。采用非参数统计方法分析两种策略下行为绩效与识别得分情况。结果：本文提出的图像处理策略与未经策略处理的方法对比发现，在任务效率、路径精准度方面表现均有显著优势。经过图像处理后被试在路线回溯试验中碰撞次数中位数减少5次，碰撞次数的离散程度收窄68.2%。完成任务总时长减少59.27 s（离散程度收窄54.5%），决策时长减少19.49 s（离散程度收窄35.0%）。总路程减少44.45 m（降幅16.1%），路程偏差率下降57.9%，离散度缩小73.1%。路径偏差增幅下降45%。在场景认知方面，图片识别测试得分提升17.19%，序列识别得分提升10.71%（离散程度下降22.6%）。结论：本文提出的图像处理策略能够显著提升被试在空间任务中的整体表现，降低个体间的行为数据波动，尤其针对任务完成耗时较长的被试，该策略展现出更强的定向优化效果。本研究的结论可为同类任务的策略选择提供参考。

Abstract:: Abstract: Objective To assist prosthesis users in exploring unfamiliar environments with restricted fields of view, this study proposes an image optimization strategy for simulated prosthetic vision and validates it through relevant spatial navigation tasks. Methods A series of urban scenes under simulated prosthetic vision were constructed in Unity via an image-processing strategy integrating semantic segmentation and edge-enhanced representation based on depth perception fusion. Twenty participants were recruited to complete a route-retracing task and picture- and sequence-recognition tests in the urban scenes under simulated prosthetic vision. Behavioral performance and recognition scores acquired under the proposed strategy were compared with those obtained from an unprocessed visual condition using nonparametric statistical analyses. Results The proposed image optimization strategy outperformed the unprocessed method in terms of navigation efficiency and path fidelity. After image processing, the median collision count for the route-retracing task was reduced by 5, and the dispersion of collisions (interquartile range, IQR) was reduced by 68.2%. The total task completion time was shortened by 59.27 s, with a 54.5% reduction in dispersion and the decision-making time was reduced by 19.49 s, with a 35.0% reduction in dispersion. The total travel distance was decreased by 44.45 m, with a 16.1% reduction route deviation rate was declined by 57.9%, and the dispersion of route deviation was reduced by 73.1% and the increase in route deviation was reduced by 45%. For scene cognition, there was a 17.19% increase in picture recognition scores and a 10.71% increase in sequence recognition scores, with a 22.6% reduction in dispersion. Conclusion The proposed image optimization strategy significantly enhances spatial navigation performance, reduces inter-subject variability under simulated prosthetic vision, and exhibits more prominent targeted optimization effects, particularly for participants with prolonged task completion duration. These findings can serve as a reference for strategy selection in similar tasks.

《中国医学物理学杂志》[ISSN:1005-202X/CN:44-1351/R]

文章信息/Info

相似文献/References:

备注/Memo

常用功能

导航/Navigate

工具/Tools

统计/Statistics