Sub-Pixel Corner Detection Based Full-Surface Deformation Measurement for Specimens

The research team of Xi'an University of Architecture and Technology has successfully addressed the bottleneck of deformation measurement in traditional geotechnical triaxial tests by virtue of the full-surface deformation digital image measurement system, with the relevant findings published in a CAS Zone 1 TOP Journal! This breakthrough marks a critical advancement of image measurement technology in the high-end scientific research field of civil engineering.

The patented technology named "Digital Image Measurement Device and Method for Specimen Surface Deformation Based on Sub-pixel Corner Detection" (hereinafter referred to as the image measurement technology) is applied to measure specimen deformation and is equipped with the function of dynamic volume change measurement. The implementation process of this technology is as follows:

1. Error Correction and Calibration of Image Measurement

The sources of comprehensive image errors mainly include camera distortion, as well as the influences of medium, glass, pressure intensity (high/low), and temperature (high/low).

The photogrammetric calibration method is adopted to obtain distortion parameters by photographing a standard calibration module with known image distribution. Various factors causing measurement errors are considered comprehensively, such as the distortion induced by the deformation of the pressure chamber under different confining pressures, the distortion caused by light passing through media with different refractive indexes, and the distortion resulting from lens manufacturing processes. Meanwhile, considering that different pressures may affect the shape and volume of the pressure chamber, strain calibration of the external deformation of the pressure chamber under the maximum pressure condition will be carried out to determine the maximum strain caused by the load and its potential impact on the glass window.

In addition, a metal rod verification method is adopted. Specifically, a certain metal rod with high elastic modulus is used to measure its elastic modulus, while the elastic modulus values under different pressures are calculated simultaneously. The metal rod is placed inside the pressure chamber, and image measurement is carried out to collect data under the action of pressure field. The collected data are then compared with theoretical data for verification.

2. Cylindrical Surface Unfolding and Image Mosaicking of Specimens

Computer vision technology is adopted to unfold the cylindrical surface images, with appropriate adjustments and transformations performed during the unfolding process, such as resizing, rotation, and translation, to ensure their correct positions and orientations in the composite image. Meanwhile, factors such as the curvature of the cylindrical surface and image resolution need to be taken into consideration to avoid image distortion or deformation.

Finally, the adjusted and transformed images are subjected to automatic mosaicking. For the mosaicked image, necessary image processing and refinement are carried out, including removing mosaic seams, adjusting brightness, contrast, and color balance, so as to optimize image quality and visual effects.

Step 1: Identify the Feature Points on the Specimen Surface and Record Their Pixel Coordinates

Step 2: Transform the Corner Points into the Plane Coordinate System with the Same Object Distance

Step 3: Mosaic the Corner Points by Means of Overlapping Columns

Step 4: 3D Reconstruction of the Specimen's Corner Points

3. Sub-pixel Corner Detection and Contour Plotting of Specimens

Pixel corner detection and specimen contour plotting technologies are widely used in the fields of computer vision and image processing, and can be applied to high-precision measurement, image recognition, 3D reconstruction, and other aspects. The technical processes of sub-pixel corner detection and specimen contour plotting are as follows:

A. Technical Process of Sub-pixel Corner Detection:

Image Preprocessing: First, preprocess the input images, and adopt operations such as filtering and enhancement to eliminate noise and improve image quality.

Feature Extraction: Extract sub-pixel level features such as corners and edges from the preprocessed images by means of the combined Sobel and Canny algorithm.

Sub-pixel Localization: After extracting the sub-pixel level features, achieve accurate sub-pixel localization of these features through interpolation and fitting.

Corner Detection: Identify the positions of corners in the images according to the results of sub-pixel localization, and conduct corresponding processing and application to realize functions such as image registration and target tracking.

Sub-pixel Corner Detection

B. Technical Process of Specimen Contour Plotting:

Image Acquisition: First, acquire the images processed by the sub-pixel corner detection method for contour plotting.

Image Segmentation: Segment different regions in the images and isolate the regions where contours need to be plotted.

Feature Extraction: Extract edge and texture features from the segmented regions.

Contour Fitting: Perform contour fitting by means of combined algorithms such as the least squares method and polynomial fitting based on the extracted features.

Contour Plotting: Plot the contours onto the images according to the fitting results, and use different colors or line types to represent different contour levels.

Fig. 1 Contour Plot of Surface Axial Displacement

Fig. 2 Contour Plot of Surface Axial Strain

4. Practical Application Effects of Image Measurement Technology

Full-Process Recording of the Shear Band Propagation Process in Triaxial Tests

The stress level field map is restored to the actual size of the specimen. Based on the region enclosed by the contour of S=1, the shear band width, shear failure angle, and shear band length at different moments can be estimated.