Road surface monitoring is a critical key factor to serve the purpose of road safety and driving comfort. Recently, many efforts have been made in developing approaches to detect road surface anomalies using smartphone sensors. However, detecting road surface anomalies from smartphone sensors face considerable number of challenges due to the various factors affecting detection rate. By aggregating data from a large number of users (i.e., concept of crowdsourcing), the accuracy of detection can be increased, and the potential false positive and false negative detection rates raised from every single source (i.e., user) can be detected and filtered. In this paper, a novel probabilistic-based crowdsourcing technique is proposed to classify and combine road surface anomalies (i.e., dynamic events) detected from various smartphones on-board vehicles. The proposed approach can integrate detected events from multiple users which are not an absolute binary scenario primarily caused by different sensing capabilities of various participators’ smartphone sensors and diversity in mechanical properties of vehicles. Furthermore, this approach considers the spatiotemporal behaviour of reported road surface anomalies from different users in different times and locations. The experimental results show that the proposed crowdsourcing method improves the accuracy and rate for detecting road surface anomalies.