视觉注意力机制在货运列车车号定位中的应用  

Application of Vision Attention Mechanism in Freight Car License Location

作　　者：葛莉[1] 赵歆波[1] 

机构地区：[1]西北工业大学计算机学院,陕西西安710072

出　　处：《微电子学与计算机》2014年第8期34-39,共6页Microelectronics & Computer

基　　金：国家自然科学基金(61171156);西北工业大学研究生创业种子基金(Z2013138)

摘　　要：为了定位货运列车车厢上的车号信息,提出并实现一种基于视觉注意力机制的定位新方法.该方法利用眼动跟踪技术获取真实眼动数据,结合提炼的车厢图像显著特征,建立车号注意力机制模型,并利用该模型预测车号显著区域,最后采用图像处理技术完成车号区域的分析定位.实验表明,相比传统基于灰度的处理方法,该算法更能适应不同光照条件下各类车厢的车号定位,具有更好的鲁棒性和普适性.To locate license number of every freight car ,a novel localization method is presented and implemented based on the visual attention mechanism .The method utilizes eye-tracking technology to collects real eye movement data ,combines eye-tracking data with salient features extracted from car images to establish a visual attention model of car license .T hen ,the model is used to predict salient license areas and .Finally ,image processing technology is applied to analyze and locate the license area .As experimental results show ,this algorithm is more insensitive to different types of car and illumination variance ,and performs better in robustness and universality than traditional methods based on intensity .

关 键 词：眼动跟踪 车号定位 视觉注意力 显著性 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]