Abstract: It is important to locate the oil spill and distribution of oil pollution timely and accurately after the oil spill accident. UAV (unmanned aerial vehicle) multi-spectral detection is one of the most effective methods for monitoring the oil spillage at sea surface, which can compensate the lag and low precise satellite remote sensing with flexibility and real-time advantages. In this study, a multispectral oil pollution detection system for UVAs was built and airborne detection experiments were carried out. Multi-spectral images were classified by maximum likelihood, minimum distance and spectral angle mapping. In addition, the classification results of the three methods were analyzed. It demonstrated that the total classification accuracy of the spectral angle mapping exceeded 90%. The system was used to detect oil spill in a fishery harbor and the original images were classified based on the spectral angle mapping. The experiment shows that the oil spill detection method has advantages of low cost, non-contact, simple operation and high classification accuracy. Thus it can meet the needs monitoring the sea oil spill quickly and has a good application prospect.