Research on the feasibility of visual measurement using first-person perspective based on smartphones

Due to the lack of professional equipment, convenient deformation measurement methods are always needed, especially in damage assessments after extreme disasters. Different measurement and judgment methods require different reference data and corresponding processing methods. Considering the flexibility of smartphones, this paper proposes a rapid and simple process of visual measurement from the first-person perspective. By using smartphones fixed on the body, images can be synchronized with the scene seen by operators. With the help of trained convolutional neural networks, the contour line of the target component in the images can be extracted directly by tracking the key points of hands, and man-machine interactive operations are realized by gesture detection without interruption. For further measurement, the error is relieved by Bezier curve fitting, and the contour is transformed into the result of a list of uniformly sampled points from a series of uniform velocity circular orbits through the Fourier series. The feasibility of deformation measurement of structural components is verified by experimental data, and further expansions of this program are discussed.