This paper presents a semi-supervised intelligent surveillance system
that detects anomalies in a parking lot. The proposed methodology extracts
speed characteristics and change of direction from the trajectories to represent
and classify the behavior of objects. The training phase aims to create
a model of the objects normal behavior from a set of manually labeled trajectories.
Then, the classification of the behavior is performed through similarity
measure between the extracted features and the model. The results showed that
the proposed system presented fast response and good accuracy in the anomaly
detection.