For the loss of GNSS information in the field transportation scene, we constructed a NRTK-GNSS/INS/Visual multi-sensor fusion system based on the crawler carrying vehicle platform. Defects of state estimation for single sensor, using the Error State Kalman Filter algorithm, establishes a GNSS/INS fusion state estimation system to improve the position accuracy and stabilize the output fusion positioning results. We study the Visual/INS fusion state estimation in the case of abnormal information reception of GNSS, using (Visual Inertial Odometry, VIO) technology, using artificial labeling methods to increase the features, and collect the offline dataset for simulation analysis in MATLAB. Finally, it was tested in field road, the results show that in short-term GNSS signal abnormalities, the navigation position calculation method realized by VIO has a good position estimation accuracy, the motion estimation error of 50 meters could be controlled within 0.5 meters, meet the short-distance positioning and navigation requirements.