Abstract: In order to study the effects of non-motorized traffic flow on the visual behaviors of cyclists at intersections, this paper conducts visual tracking experiments through an eye tracker under different traffic flow conditions. Eye movement data of e-bike riders at intersections are collected, such as eye movement time, the number of gaze, pupil change, and gaze point locations. The cyclists’ eye movement characteristics, visual load, gaze shift probability and gaze smooth probability are compared and analyzed.The results show that the number of cyclists’ gaze duration, gaze number and gaze sweep number at intersections increase with the number of non-motorized vehicles. The cyclists are divided into two groups according to non-motorized traffic flow, and the degrees of visual load and gaze behavior indexes of the cyclists in the peak group are significantly higher than those of the cyclists in the flat group, indicating that the risk perception ability of the cyclists under peak traffic flow conditions is enhanced. In terms of the probability of smooth gaze, the gaze shifting behavior of the peak group riders is concentrated and stable, while the flat group riders have a more flexible gaze shift pattern and are able to take into account of the traffic information on left and right sides. The results can provide theoretical support for non-motorized traffic safety research.