Complex indoor scene localization has problems such as signal occlusion and multi-path effect, this paper analyses the topology formed by the motion trajectory of IMU and the coordinates of the UWB base station, and based on this topological constraints, it implements the judgement of line-of-sight/non-line-of-sight UWB measurements, and selects line-of-sight measurements to participate in localization computation. The heuristic extended Kalman filter algorithm (H-EKF) is used to establish the error model of IMU/UWB dual sensors, and the UWB and IMU are tightly coupled for fusion processing, and the Kalman filter information is corrected according to the results of the selection of range values to make full use of the useful information of multiple moments, improve the fusion accuracy, and achieve the alleviation of the non-visual range error.