Abstract: In foggy weather conditions, the low visibility of transmission lines makes it challenging for unmanned aerial vehicles (UAV) to guide the split-type spacer installation robot for accurate on-line. To address this issue, this paper proposes a wire detection and on-line control method based on a lightweight dehazing network and YOLOv8 with collaborative optimization. First, a lightweight dehazing module is constructed to enhance the clarity and contrast of foggy images through end-to-end feature mapping. Then, the dehazed results are fed into the YOLOv8 detector, where a collaborative optimization mechanism is employed to strengthen the feature representation of transmission lines, enabling stable detection under foggy conditions. Finally, an on-line control strategy is designed according to the geometric relationship between the detection box center and the camera imaging model, converting pixel deviations into UAV velocity and attitude commands to achieve precise on-line of the robot’s upper module. Experimental results demonstrate that the proposed method achieves detection accuracies of 89.65%, 86.07%, and 81.76% under light, medium, and heavy fog conditions, respectively, representing an average improvement of about 5 percentage points compared with the original YOLOv8 baseline. Moreover, by introducing depthwise separable convolutions, the inference time is reduced from 11.8 to 8.3 ms, achieving a 30% improvement. Meanwhile, the proposed method achieved stable landing points for the upper part of the robot in the online guidance experiment, meeting practical operational requirements.