Tower-Type Detection of UAV Aerial Image Based on YOLOV5 Network Model

With the application and promotion of UAVs in power system, the needs of information mining or target recognition of UAV aerial image is also increasingly stronger. UAV aerial image has advantages of high inspection efficiency and low strength; it can facilitate the establishment and processing of various databases. Due to the different tower types having different detection requirements and image collection processes, it needs to identify and classify the type. The tower-type recognition network method is established in this paper, firstly, the role of YOLOV5 model Input, Backbone, Neck, and Prediction in the role of UAV aerial image tower type recognition is analyzed; secondly, the UAV aerial image model recognition test environment is built, then the training effect can finally classify the data collected by the UAV aerial image by real-time identification tower type verification method and can establish a standardized detection database.