Abstract: In this paper, we describe a new approach to real-time detection of bedload transport rate using image recognition technology, based on the characteristics of bedload transport. This method can be used to determine the particle composition of bedload and its transport rate of each compositional component, and thus the dynamic transport rate, which is difficult to measure by the traditional methods, can be directly measured in real-time. The flume experiment indicates that the method of bedload transport rate measurement based on image recognition has good reliability. Application to real-time dynamic detection shows that the transport of fine particles is stable with continuous variation while the transport of coarse particles is characterized by obvious intermittency. The non-contact detecting method presented herein can achieve dynamic monitoring on the transport process of non-uniform bedload, real-time tracking of the particle composition and each component's transport rate, and tracking of the strip transport rate. It can also lower the ill conditions of curve fitting in calculation of transport rate. Thus, the method would enrich bedload measurement and help the study of non-uniform bedload transport.