Transmission line is a vital component in a power system and damages to transmission lines pose serious threat to the safety of power system. In this paper, we provide an intelligent detection method for untwisted strands in transmission lines using image sequences captured from the Unmanned Aerial Vehicles (UAVs). Considering that transmission line detection is a challenging problem under illumination variations and cluttered background, we propose to extract transmission lines using quaternion phase congruency (QPC) model and cascade filters of shape filtering and color filtering. Compare with current methods based on gradient computation, we can extract the region of transmission lines with higher accuracy due to invariance to noises and contrast changes. To meet real-time requirement of untwisted stands fault detection, we establish statistical histogram of edge orientation around region of transmission line and the divergence of edge orientation is used as a metric to locate untwisted strands faults. In contrast to current methods, experimental results show that the proposed method can obtain an improvement of detection accuracy of 9.8% for the untwisted strands faults in transmission lines.