Abnormal object detection of transmission lines plays a very important role in improving the safety, reliability and stability for transmission system, but existing object detection has not been effectively designed for large scale changes, many small objects, dark light, partial occlusion of abnormal objects on the line, resulting in recognition speed slow, susceptibility to environmental interference, and frequent false alarms. In response to the above problems, this paper adopts a two-stage deep network. FPN is used to extract multi-scale features so that backbone network can better adapt to multi-scale changes of object, and feature enhancement is performed through global network to obtain clearer and representative multi-scale object features. A feature-guided region proposal generation network is proposed in RPN, which can generate sparse and arbitrary-shaped anchors, generate tighter mask bounding boxes and improve detection performance of abnormal objects. In detection stage, a multi-task loss function is used to improve prediction accuracy and generalization ability of network, and to improve detection performance of abnormal objects. Ablation experiment and performance comparison on MS COCO dataset prove the effectiveness and advancement of proposed method. The detection accuracy of abnormal objects on transmission line dataset reaches 77%, which is better than mainstream deep learning object detection methods.