Detection of Electrical Diagram based on QueryDet
Shiyi Pang, Feng Guo, Zhenrui Yang, Shilu Zhao, Zijie Yang · 2023
A significant challenge in the field of grid power system is the need for scheduling and maintenance personnel to manually draw and import CAD design drawings, which can be a labor-intensive and error-prone process. To address this issue, we proposal a electrical diagram detection method based on Querydet, which significantly improves the efficiency and accuracy of detection. By utilizing the Cascade Sparse Query (CSQ) mechanism and sparse convolution methods, this method can rapidly identify electrical components on high-resolution feature maps, overcoming the difficulty of detection components in high-resolution diagrams. Next, contour tracking algorithms are used to extract the contours of connected regions of the diagram. Finally, the topological relationships of the electrical diagram are constructed based on the contours and the positions of the electrical components. Testing was conducted on a dataset of actual power grid electrical diagrams exported from the scheduling system, and the results demonstrate that the proposal method can effectively detect various electrical components and their topological relationships in electrical diagrams, and is robust to complex interference factors.