Abstract: Due to the long-term exposure of transmission towers to strong wind vibrations, uneven ground settlement, and material creep, the structural inclination presents a nonlinear evolution characteristic of gradual accumulation and sudden disturbance. Moreover, the monitoring site is subject to strong electromagnetic interference and complex terrain obstructions, resulting in low image signal-to-noise ratio and difficulty in feature extraction. This leads to large measurement errors of the inclination angle. Therefore, a non-contact online monitoring method for transmission tower inclination based on machine vision is proposed. The improved YOLOv8 algorithm is used to identify tower targets and eliminate background interference in the monitoring images, suppressing non-tower pixel energy and improving the image signal-to-noise ratio; the Canny edge detection operator is used to extract the tower skeleton contour, combined with the Hough linear transformation and standardized cross-correlation template matching, to locate the anchor point at the lower end of the insulator as the reference feature point, accurately quantifying the tower inclination angle and the magnitude of the offset vector; a perspective projection mapping model is constructed to compensate for the pitch and yaw angle errors of the camera installation, and a mapping relationship between image features and the actual posture of the tower is established, achieving non-contact high-precision monitoring of transmission tower inclination based on the mapping model. Comparative experimental results show that the measurement error of the inclination angle at each measurement point is stably controlled within ±0.05°, with the maximum absolute error being approximately 0.04°, effectively solving the problem of large measurement errors in inclination angle.