An efficient technique for object recognition using fractional Harris-Stephens corner detection algorithm

J. E. Lavín-Delgado

doi:10.1007/s11042-023-16428-0

2023-08-12

PRE

J. E. Lavín-Delgado

J. F. Gómez‐Aguilar *

Daniel Urueta

Zizilia Zamudio

José Andrés Alanís-Navarro

DOI：10.1007/s11042-023-16428-0

原文链接

原文求助

摘要

En 中文

In this research, a fractional-order technique for corner detection and image matching based on the Harris-Stephens algorithm and the Caputo-Fabrizio and Atangana-Baleanu derivatives is proposed and experimentally tested. It focuses on three main ideas: 1) To suppress image noise more effectively while maintaining better image fidelity, a fractional Gaussian filter based on the Atangana-Baleanu derivative is designed. 2) The image derivatives and consequently the Hessian matrix are generalized through the Caputo-Fabrizio derivative, which has a high capability to preserve texture details in low-contrast regions. 3) An image-matching scheme that combines our fractional corner detector with the SURF algorithm is developed so that the corner extraction and the accuracy of matching images are improved. The proposed technique is compared experimentally with the conventional Harris-Stephens algorithm and some other methods reported in the literature. Experimental results on test images validated this approach in terms of more corners detected and matching accuracy improvement. In addition, the proposed operator is implemented for image processing of concrete structures images, i.e., for the identification and analysis of cracks in this kind of structure. Implementation results on images with different types of cracks prove the advantages of our operator over other methods since it can detect more pixels corresponding to cracks, improving their identification and the way they propagate, that is, their patterns of propagation.

Keyword:

Fractional calculus

Corner detection

Fractional hessian matrix

Fractional gaussian filter

Harris-Stephens operator

Concrete crack