Tiny target detections, especially power line detection, have received great attention due to its critical role in ensuring the
flight safety of low-flying unmanned aerial vehicles (UAVs). In this paper, an accurate and robust power line detection method is proposed, wherein background noise is mitigated by an embedded convolution neural network (CNN) classifier before conducting the final power line extractions. Our
proposed method operates in three steps: 1) extract edge features of power lines from a testing image, 2) employ a CNN classifier to remove the background noise, 3) use a HoughTransform (HT) based fine-selection module to locate power lines. Comprehensive experiments demonstrate the superiority of the proposed method, compared to the state-of-the-art methods.