This paper presents an approach for indexing a large set of videos by considering the cinematic behaviour of local visual features along the sequences. The proposed concept is based on the extraction and the local description of interest points and further on the estimation of their trajectories along the video sequence. Analysing the low-level description obtained allows to highlight semantic trends of behaviours and then to assign labels. Such an indexing approach of the video content has several interesting properties: the low-level description provides a rich and compact description, while labels of behaviour provide a generic and semantic description, relevant for selective video content retrieval depending on the application. The approach is firstly evaluated for Content-Based Copy Detection. We show that taking these labels into account allows to significantly reduce false alarms. Secondly, the approach is experimented on particular applications of video monitoring, where selective labels of behaviour show their capability to improve the analysis and the retrieval of spatio-temporal video content.