To address the challenges related to obstacle recognition and distance perception in the context of unmanned aerial vehicle (UAV) automatic inspection of distribution lines and maintenance involving insulation layer coating, a method combining visual images and three-dimensional point clouds for obstacle identification is proposed. Data augmentation preprocessing is applied to enhance the dataset of images. A deep learning approach based on feature extraction is introduced for model training, enabling the identification of obstacle categories and orientations. Distance measurements for the detected targets are obtained by integrating three-dimensional point cloud information. Experimental results demonstrate the practical effectiveness of this approach, which combines the advantages of two types of sensor data. The maximum recognition error is 2.356%. The proactive obstacle identification method contributes to enhancing the obstacle perception capabilities of UAVs and coating robots, thus ensuring operational safety.