The application provides a kind of based on
deep learning's marine converter
station warm air professional three-way identification method, comprising the following steps: obtaining the dwg file of target warm air professional CAD drawing, constructs graph structure: graph
data annotation based on prior knowledge: model training and topological
feature learning based on graph neural network;To the graph structure that is well constructed, all subgraphs containing 3 lines are intercepted by sliding window, as the input of the trained graph neural network, the subgraph with confidence degree greater than or equal to threshold is screened and classified, and is determined as three-way, and the three-way area boundary box and the three interface center point coordinates are output.The application constructs graph structure with lines as nodes, directly models the core topological relationship of three-way, avoids the dependence of prior art on form details and pixel
noise;By adjusting the angle range,
distance threshold and other parameters, the warm air three-way of different specifications can be adapted, or migrated to other industrial pipeline components (such as four-way,
elbow) identification scene, without reconstructing
network architecture, reducing technical iteration cost.