Laying of secondary cables is an important step in the secondary installation construction process of power transformer projects. Presently, the installation of secondary cables is predominantly executed manually, lacking effective mechanized construction means. Consequently, there are problems such as large workload, complex processes, low efficiency, high labor demand and high cost. Based on the application of cable-pulling robot in the laying of secondary cables, aiming at the localization problem of four-wheel cable-pulling robot during cable laying, this paper proposes a localization scheme of correcting inertial navigation posture by observing an ArUco marker fixed on the wall of the cable trench with a camera. This method solves the problem that localization methods based on vision and laser sensors and GPS-RTK cannot achieve high precision localization of cable-pulling robot in cable trench environment. The method is to identify the ArUco markers pre-arranged on the wall of cable trench during the movement of a four-wheel cable-pulling robot. It will select specific frames based on its pose in the camera plane. The pose information of the inertial navigation is modified by recognizing ArUco markers to realize the pose correction of four-wheel cable-pulling robot. The experimental results show that the method can enable the four-wheel cable-pulling robot to achieve real-time and accurate localization in the cable trenches. This further enhances the intelligence of the secondary cable laying process in power transformer substations, and enhances construction efficiency and quality.