To improve the positioning accuracy of inspection robots， a monocular vision-based simultaneous localization method is proposed. A monocular camera was installed on the inspection robot to capture images of the calibration board from multiple angles to obtain the camera''s internal parameter matrix. The inspection space coordinate system was transformed into the pixel coordinate system of the inspection two-dimensional image to obtain discrete pixel coordinate data. By comparing the pixel brightness of the video stream image frames， the corner points of the image were identified， and the position and direction of the points were calculated as feature point descriptors. The random sample consensus（RANSAC） algorithm was used to filter the matching results of the descriptors to remove the mismatched points. The perspective-n-point（PnP） algorithm was used to solve the pose of the monocular vision camera and convert it into the pose of the inspection robot to complete the simultaneous localization. Experimental results show that compared with the lidar odometry and mapping（LOAM） algorithm and light detection and ranging（LiDAR）， the average maximum path tracking offset of the proposed method is 0.89m. In the simulated environment 1， the accuracy is improved by 2.9% compared with the LOAM method and by 1.5% compared with the LiDAR method， the recall rate is improved by 1.3% compared with the LOAM method and by 1.7% compared with the LiDAR method， and the F1 measure is improved by 2.1% compared with the LOAM method and by 1.6% compared with the LiDAR method， demonstrating a relatively ideal positioning accuracy.