融合语义理解的航站楼显示设备故障检测方法

A Fault Detection Method of Terminal Display Equipment Fusing Semantic Understanding

针对航站楼内显示设备出现的故障画面,提出一种融合语义理解的故障检测方法。首先设计滚动文字拼接技术提取界面文字信息;然后根据机场业务背景融入语义规则,训练得到故障分类模型,实现对显示设备出现的非正常显示界面和显示信息歧义等故障的智能检测;最后使用神经网络模型压缩技术将模型轻量化并部署在SOM-RK3399嵌入式设备上。实验表明,融合语义理解模块的检测方法的分类准确率达到88.74%,该方法能够有效解决传统故障检测技术的不足,提高故障检测效率,减少人工检验情况。

A fault detection method that integrates semantic understanding is proposed to address the fault images of display equipment in terminal buildings.Firstly,a rolling text stitching technique is designed to extract interface text information.Then,based on the airport business background,semantic rules are integrated to train a fault classification model,which realizes intelligent detection of abnormal display interfaces and display information ambiguities in display devices.Finally,the neural network model compression technology is used to lightweight the model and deploy it on SOM-RK3399 embedded devices.The experiment shows that the classification accuracy of the detection method integrating semantic understanding module reaches 88.74%.This method can effectively solve the shortcomings of traditional fault detection techniques,improve fault detection efficiency,and reduce manual inspection situations.