This progress report describes the activities carried out, and the results produced in the second year of the three year research project Multi-sensor surveillance systems, which is funded by the Swedish Armed Forces. The project examines and evaluates how different configurations of collaborative and network-based multi-sensor systems should be designed to automatically detect, classify and track threats in a difficult complex environment. The project also investigating how the data from the individual sensor systems can be distributed to a central unit which creates a common situational awareness, and how multi-sensor systems can be developed for blue force tracking in complex environments. The long-term objective of the research is to provide an improved military situational awareness. We have studied how a multi-sensor network should be designed for a robust and sustainable manner to support surveillance in mainly military scenarios. This year, the project has examined how a system based on a relatively large number of low-resolution sensors for surface or limit coverage in the complex environment can be supplemented with drones for alarm verification and other limited-time reconnaissance missions. Based on a scenario in which a military base on the battlefield is guarded against hostile activities, the architecture for an autonomous sensor systems is described briefly. The sensor system solves the problem of effectively monitoring a large terrain area with limited staff effort. The work represents a first step towards a simulator prototype, where the autonomous processes can be developed and validated in future work. Within the work package Distributed target tracking we examine how a multi-sensor system for surveillance of wide and complex environments can distribute their algorithms and be able to monitor people and vehicles for longer periods. The algorithms implemented in the previous year have now been tested on simulated and real data. The results have been verified against simple motion models and more complex movements, both in simulation and on real data. It turned out that the relatively unknown method Safe Fusion, developed at Linköping University, gave the best results. The work package Anomaly Detection, have studied how methods and systems for monitoring the urban environment should be designed for early detection of threatening events, particularly in crowds. The implemented methods automatically sorts large amounts of image data and identifies the important adverse events, which can alert an operator. During the year, we have studied how to improve the anomaly detector by filtering out noise and unwanted signals. A new anomaly detector has been developed, where the parameter learning of the normal situation is automatic, and where the anomaly detection is based on groups of individuals, which make is more robust in some applications.