An improved Harris corner detection method for honeycomb structures

Structural defects and cell irregularity significantly impact the performance and safety of honeycomb structures. Thus, various image processing techniques have been employed to evaluate the cell shape and detect structural defects within these structures. This paper proposes an improved Harris vertex extraction method for detecting honeycomb structures. Initially, the influence of parameters on Harris detection results is examined, and an empirical equation is derived to directly determine the optimal parameter values. Subsequently, a thinning process is applied to generate a honeycomb skeleton with single-pixel lines, ensuring consistent parameter settings across different cellular images and eliminating scale discrepancies. A Gaussian filter is then incorporated to smooth the noise and create a controllable multilevel grayscale transition, allowing the Harris corner detector to accurately identify corners. Comparative experiments demonstrate that the proposed method outperforms conventional algorithms, achieving an accuracy of 99.5%. In addition, a cell reconstruction approach is introduced to form a measurable honeycomb Y-shaped structure, accompanied by a regularity evaluation method based on the consistency of the side lengths and internal angles. Test results confirm that the proposed method accurately determines the side lengths and angles with an error margin of less than ±2% compared with manual measurements, effectively evaluating the regularity of the honeycomb. This methodology enhances the applicability and reliability of the structural performance evaluation through image processing.