We investigated differences in visual search of dangerous events between security experts and naïve observers during the observation of large scenes, typically encountered on the grandstand of stadiums during soccer matches. Our main technical objective was the reduction of computational effort required for the detection and recognition of such events. To overcome the scarcity and legal issues associated with real footage, we designed a new algorithm for the synthesis of crowd scenes with well-controlled statistical properties. We characterize the relative importance of saliency and expert knowledge for the generation of correct detections and the visual search strategies for both security experts and naïve observers. We found that during the first few seconds of this search task, experts and naive observers look at the scenes in a similar fashion, but experts see more. We compare the results with theoretical models for saliency and event classification. We show that the recognition model can deliver reasonable classification/detection performance even when operating under real-time constraints. When real-time operation is not a concern, performance can be improved further by allowing the model to grow.
展开▼
