Automated Strain Measurements in Uniaxial Testing via Computer Vision

摘要

In standard unconfined strength testing, only the axial strain is measured directly, and the radial strain is calculated based on the assumption of uniform deformation. Alternatively, there are other assumptions, such as parabola and conical frustum. In fact, soft clays lose cylinder geometry and deform in an irregular geometry. This study aims to determine strains directly through computer vision using a specially developed setup and a computer code. An image acquisition module was integrated to a traditional unconfined compression test device. The time lapse images of the soil sample, which were placed between the backlight and the camera, were digitally acquired. Dimensions of the traced sample are measured so that axial and radial strains are calculated in real-time. The experimental program was carried out on 15 different cylindrical cohesive soil samples prepared in different moisture contents. Stress-strain curves were drawn according to the standard method, parabola assumption, conical frustum assumption, and a newly developed technique. The results were evaluated and compared. Standard method and parabola assumption calculations overestimate the stress, whereas the conical frustum approach gave close results with computer vision method.