The invention belongs to the technical field of contact network inspection and positioning, and discloses a method for identifying and positioning a contact network rod and components thereof in an inspection process, which adopts an algorithm based on a deep neural network to realize simultaneous positioning of a plurality of targets, and comprises an image preprocessing step, a neural network processing step and a target coordinate conversion and output step, and the selected CNN network structure can meet the requirements of quickly, efficiently and accurately identifying the overhead linesystem and components thereof at the same time in use.
本发明属于接触网巡检定位技术领域,公开了一种采用基于深度神经网络的算法实现多个目标同时定位的、在巡检过程中对接触网杆及其组件进行识别定位的方法,包括图像预处理步骤、神经网络处理步骤和目标坐标转换输出步骤,选用的CNN网络结构,使用中可以满足快捷、高效、准确的对于接触网及其组件同时进行识别。
A method for identifying and positioning a contact network rod and components thereof
接触网及其部件视觉识别定位方法
2021-01-12
Patent
Electronic Resource
Chinese
Vibration reduction type contact network positioning mechanism and design method thereof
European Patent Office | 2024
|Vehicle whistling identifying and positioning system and method
European Patent Office | 2022
|