Intelligent registration techniques of power equipment's using data fusion of contour-based infrared and visible data

This paper proposes an automatic registration method for power equipment images using contour centreline main orientation features. Addressing low accuracy caused by scale/viewpoint variations in multi-modal images (infrared and visible), the method detects contour corner points as feature points, assigns scale/viewpoint-invariant orientation features through centreline analysis, and employs improved scale-invariant feature transform (SIFT) descriptors with connecting-line consistency matching to determine transformation parameters. Experimental results demonstrate registration errors of 2.742 and 2.543 in scenarios with subtle and significant viewpoint differences respectively, outperforming traditional SIFT, speeded-up robust features (SURF) and partial intensity invariant feature descriptor (PIIFD) methods. This approach effectively resolves complex-scenario registration challenges, providing an effective solution for intelligent monitoring of power equipment.