Significant Target-Guided Feature Extraction Algorithm Based on Optimized Superpoint

In order to address the problems of low efficiency and poor accuracy of feature extraction in the process of heterogeneous image alignment caused by the complex environment of the substation and the heat interference of various power equipment. This paper proposes a significant target-guided feature extraction algorithm optimizing SuperPoint based on the SuperPoint algorithm. Firstly, based on the original SuperPoint algorithm framework, an image preprocessing link is added, edge features are extracted using the Canny algorithm, and the extracted edge features are fused with the original to strengthen the contour information of the image. Secondly, a significant target-guided feature extraction strategy is proposed, which uses the K-means image segmentation algorithm to mark the significant target region in the image and the process of feature extraction, the feature extraction is mainly carried out on the significant target region, to reduce the percentage of the number of feature points in the background, and to improve the efficiency of feature extraction and accuracy. The experimental results show that the feature extraction algorithm proposed in this paper has higher accuracy and robustness. On the CAO-C2F dataset, the alignment rate of this algorithm after feature extraction is 87.31%, and the time consumed is 112.58ms, which is 12.04% higher than the original SuperPoint algorithm and 23.09ms lower than the original SuperPoint algorithm, which proves that the algorithm proposed in this paper is superior for the feature extraction in the process of alignment of heterogeneous images of substation equipment.