Abstract: With the integrated perception system of IR camera and 3D lidar, the calibration based on feature point matching was studied and the thermostat heating hollow net was designed to provide feature points. Infrared image feature points were extracted by Harris corner detector. To decrease the influence from mixed pixels, plane fitting of heating net and plane conformity check of laser points on net were used to locate the edges in depth image. Edges'average value at corner was calculated to extract depth feature point. At last, NMSM-EM method was adopted to optimize the result. The research results show that the integrated perception system can accurately perceive mobile robot's running environment in dim light.