Drosophila larvae are an insightful model and the automated analysis of their behaviour is an integral readout in behavioural biology. Current tracking systems, however, entail a disturbance of the animals, are labour-intensive, and cannot be easily used for long-term monitoring purposes. Here we present a novel monitoring system for Drosophila larvae which allows us to analyse the animals in cylindrical culture vials. By utilizing frustrated total internal reflection in combination with a multicamera/ microcomputer setup we image the complete housing vial surface and thus the larvae for days. We introduce a calibration scheme to stitch the images from the multi-camera system and unfold arbitrary cylindrical surfaces to support different vials. As a result, imaging and analysis of a whole population can be done implicitly. For the first time, this allows to extract long-term activity quantities of larvae without disturbing the animals. We demonstrate the capabilities of this new setup by automatically quantifying the activity of multiple larvae moving in a vial. The accuracy of the system and the spatio-temporal resolution is sufficient to obtain motion trajectories and higher level features like body bending. This new setup can be used for in-vial activity monitoring and behavioural analysis and is capable of gathering millions of data points without both disturbing the animals and additional labour time. In total we have analysed 107, 671 frames resulting in 8, 650 trajectories, which are longer than 30 seconds, and obtained more than 4.2 ⇥ 106 measurements.