Object-Based Image Analysis for Identifying Small Scale Elements During Large Scale Public Events

Abstract

Major public events such as open-air festivals are characterized by large concentrations of people in relatively small areas. These events represent challenging situations for public authorities concerned with organization, traffic and security and for festival organizers. In the research project VABENE++* an end-to-end monitoring system has been developed in order to provide situational and traffic related information in near real-time based on terrestrial and airborne platforms. In the present study, the potential of object-based image analysis (OBIA) for the extraction and classification of small scale features in the context of large scale public events is examined. The features to be identified involve different types of vehicles and tents. A first step involves a quantitative evaluation of segmentation quality and the investigation of the relationships between segmentation and machine learning algorithms. Methods are evaluated considering computational cost, thematic accuracy, robustness/transferability and the amount of user interaction. Finally, a methodological framework is developed as a basis for the development of (semi-)automatic processors to be implemented in the existing airborne monitoring system of VABENE++. Even though methods are predominately tested and compared using very high resolution aerial imagery, it is also assessed if and to what extent lower resolution imagery from optical satellites (e.g., WorldView-2) represent appropriate alternatives for the same objectives. Possible application scenarios in the context of large scale events may involve the monitoring of the occupancy of parking and assembly areas, camping sites and the estimation of the number of overnight guests and the assessment of the accessability of emergency and escape routes.