Festivals and big scale events are becoming more and more popular, they can attract thousands of spectators. Ensuring the safety of the crowd has become a top priority to many organisers after the multitude of dramatic accidents that resulted in losses in human lives. Monitoring the crowd via smartphones is a relatively new technique that emerged recently with the capabilities of mobile phones to transmit their GPS location data. We present a novel approach, based on the local crowd pressure, combined with the detection of groups in a crowd, to detect critical situations and propose evacuation plans that does not separate groups of people that are together. Groups were detected using DBSCAN clustering algorithm with 80 % accuracy. Location acquisition was tested during the Campus Fever event, and 87 % of the collected data had an accuracy lower than 10 m while 29 % of the total data had 5 m of accuracy. During 2 h of monitoring, activity of the application, reduced the battery of 20 %.