Microseismic monitoring from large arrays using migration-based detection and location techniques suffers from detections of false positives. Semblance has been considered to differentiate between false positive and true events. However semblance is not suitable for variable signals, e.g. source radiation. We present a new methodology for event detection and location using source mechanism corrected semblance suitable for multichannel processing of microseismic datasets acquired with large arrays. The main novelty of this methodology is that amplitudes are corrected by the radiation pattern of an inverted source mechanism before the semblance computation. We show that source mechanism correction is the key procedure maximizing the value of semblance and makes the detection based on semblance superior to commonly used simple stacking. We show results of this methodology for a large surface star-like array for synthetic as well as for field data with various level of noise.