To enhance the accuracy and environmental adaptability of pose detection systems during the automatic assembly process of shield segments, a highly robust multi-source positioning method integrating line laser sensors and an RGB-D camera is proposed. At the perception level, five line laser sensors were employed to extract edge features of the segments, and the coordinates of the intersection points between arc edges and straight edges were derived in combination with the geometric model to estimate the segment pose. Meanwhile, a target system was designed, and the RGB-D camera was used to capture images and depth information of the segments. Through coordinate transformation, the spatial posture of the target relative to the assembly robot's end was obtained. In view of the limitation that a single type of sensor was vulnerable to environmental disturbances, an improved adaptive weighted fusion (IAWF) algorithm was proposed, which dynamically adjusted weights based on observation support and measurement variance, thereby achieving high-precision and high-robustness fusion positioning. Finally, the simulation was conducted to verify the stability and superiority of the IAWF algorithm under abnormal sensor conditions. Meanwhile, a fusion positioning experimental platform was established to evaluate the positioning performance of the IAWF algorithm under various working conditions, including normal, low-light, and abnormal line laser measurement scenarios. The results showed that the proposed fusion positioning method significantly improved the anti-interference ability of the segment pose detection system in complex environments while maintaining high measurement accuracy. The average position error was reduced by 28.4%, and the average posture error was reduced by 34.2%, which was superior to the measurement effect of a single type of sensor. The maximum misalignment and maximum gap during segment assembly were 4.8 mm and 4.6 mm, respectively, which met the actual engineering requirements. The research results provide theoretical foundations and technical support for the design of intelligent perception systems for the automatic assembly of shield segments.