Presented paper is focused on fast near surface anomaly detection in potential data. Our aim is to find fast and semi-automated anomaly detection technique for the near surface anomalies with defined geometry. The proposed algorithm is based on the shape recognition. The edge and line detection is used on acquired data to detect the typical shape of the anomaly. Shape geometry parameters are converted into the anomaly parameters and location information. The technique was tested using a set of noise-free and noisy synthetic gravity data; satisfactory results were obtained.