| Abstrakt: |
Abandoned baggage poses a potential threat to public safety, which needs to be monitored to avoid catastrophic effects. Identifying left baggage, the owner of the baggage, and the intention of the owner leaving the baggage are the challenging tasks in an automated video surveillance system. This paper aims at detecting abandoned objects by recognising and tracking the person carrying the object with context sensing. Here, a well-organised background modelling strategy and background subtraction in the HSV colour plane is proposed that yield a complete foreground image. The retrieved foreground blobs were identified as pedestrians, luggage, or other moving items using the prior model built with deep features. Further, deep features with kernel canonical correlation analysis and the cosine similarity index are used for tracking the person by re-identification in successive frames. The gait energy image generated for the tracked individual is identified for the condition of the carried baggage, and thus the owner of the bag is recognised. The duration of stay of the person carrying an object and the behavioural cues displayed by the person serve as the defining features for threat assessment. The PETS 2006, AVSS 2007, and ABODA data sets were used to evaluate the performance of the proposed method, and the experiments demonstrate promising results comparable to state-of-the-art techniques. The results suggest that the proposed method can effectively detect abandoned objects in public places by recognising and tracking the person carrying the object and using context sensing. The method's performance is comparable to state-of-the-art techniques and can be used in real-world scenarios to enhance public safety. [ABSTRACT FROM AUTHOR] |
Full text from SpringerLink