Proposed Algorithm for Detecting Displacement of Points of Interest in SAE1020 Steel Element Under Tension

Resumo

The use of non-destructive structural health monitoring methods has gained increasing attention due to the need to ensure the integrity of structures without compromising or interrupting their operation. Among these methods, computer vision-based monitoring emerges as a promising alternative. However, this approach faces challenges in accurately detecting points of interest on critical surfaces, especially in environments with varying lighting conditions, contrast, and surface texture, in addition to the need to ensure the stability of the setup during image acquisition. This study presents a method for detecting points of interest in images captured during a uniaxial tensile test, in accordance with ISO 6892-1, with the goal of estimating the elongation of an SAE 1020 steel specimen. Image acquisition was carried out using a fixed tripod under controlled lighting. Pre-processing, segmentation, and feature extraction techniques were applied to facilitate point detection using algorithms from the OpenCV library, such as GaussianBlur, thresholding, and Canny edge detection. Data handling and storage were performed using auxiliary libraries including NumPy, Pandas, and Matplotlib. The data analysis showed that detection is highly dependent on the contrast and uniformity of the region of interest. On surfaces with tonal or textural variations, a decrease in both the quantity and quality of detected points was observed, requiring parameter adjustments and manual delimitation of the analysis area. In uniform and well-lit regions, the method demonstrated high reproducibility and good precision in point localization. The influence of physical imperfections such as cracks and wear was also observed, affecting the consistency of the results. It is concluded that the proposed method is reliable for detecting points on standardized specimens, offering satisfactory accuracy when compared to experimental results. Nonetheless, the system requires specific care in preparing the testing environment and in tuning detection parameters.