In a scenario where two UAS can benefit from cooperation, but where one of the UAS is non-cooperative and only broadcasting very limited state information, there is a need for the second UAS to estimate the intentions of the first UAS. When the second UAS is capable of this they will be able to solve joint tasks faster even though the cooperation is only "one way". In this work we investigate a method for such an estimation. It is based on a short and a long term prediction of the intentions of the non-cooperative UAS, combined in a Kalman filter to provide a quantitative estimate of the intentions. The method is demonstrated on two helicopters semi-cooperating on covering a number of hot spots in field of crops.