In this article, semi-automatic approaches are developed for wide area situation assessment in near-real-time. The two-step method consists of two granularity levels. The first entity assessment uses a new multi-target tracking algorithm (hybridization of GM-CPHD filter and MHT with road constraints) on GMTI data. The situation is then assessed by detecting objects of interest such as convoys with other data types (SAR, video). These detections are based on Bayesian networks and their credibilistic counterpart.