基于图片显著性的安检视觉搜索研究

Visual Search in Security Screening Based on Stimulus Salience

应用显著性视觉注意模型分析安检员和新手的注意分配与机场X光行李图片显著性之间的关系。记录安检员和新手参加模拟安检实验搜索图片时的眼动数据,经分析发现安检员能比新手更快更准确地识别危险品。通过分析安检员和新手的初始和整体选择性注意,发现经验不是影响注意分配的主要原因。初始注视点和整体注意分配与图片显著性的相关性说明自下而上的注意驱动机制在安检视觉搜索中起到了重要作用。

A saliency-based visual attention model was used to examine the correlation between attention allocation of human and salience of X-ray luggage images.The eye movements of screeners and native observers were recorded when they searched X-ray images of air passenger luggage for potential terrorist threat items such as guns,knives and improvised explosive devices.Eye position data revealed that screeners were faster to fixate on target areas and detect threats.Initial visual selective attention and overall attention allocation were examined and it found that experience was not an important determiner of attention allocation.The correlation between fixation locations and stimulus salience was found to be significantly greater than chance alone.The results indicate that bottom-up,stimulus-driven attention is very important in security screening visual search.