Car detection in aerial thermal images by local and global evidence accumulation

This paper introduces an approach to automatic vehicle detection fromaerial infrared images of approximately 1 m resolution. On one hand,the extraction relies on a local description of cars, i.e. blob-likestructures; on the other hand, because many other objects in urbanareas have a similar appearance, the local model is extended by amore global description that incorporates knowledge about the appearanceof cars as repetitive patterns in dense traffic situations or infilled parking lots.The implemented system is intended to detect vehicles independentof their current state (stationary or moving). Since vehicle velocityis not included as feature for detection, the system can be usedas first module for vehicle tracking but also for analyzing trafficcongestions or counting cars in parking lots. The only requirementto be met is that cars appear as repetitive patterns which consistof blob-like structures.     «

This paper introduces an approach to automatic vehicle detection fromaerial infrared images of approximately 1 m resolution. On one hand,the extraction relies on a local description of cars, i.e. blob-likestructures; on the other hand, because many other objects in urbanareas have a similar appearance, the local model is extended by amore global description that incorporates knowledge about the appearanceof cars as repetitive patterns in dense traffic situations or infilled parking lots.The imp...     »