Robust 3-D Reconstruction and Parameter Measurement of the Foot Using Multiple Depth Cameras

In recent years, 3-D reconstruction and measurement of the foot have been applied in the personalized customization of shoes’ customization. To solve the dependence of foot point cloud coarse registration and camera calibration on 2-D lenses and the problem that registration easily fails due to the small number of foot point cloud feature points and low overlap in all directions, this article proposes a 3-D reconstruction method for the foot. First, the depth cameras were used to obtain the foot point cloud from four directions, and a 3-D calibration plate was introduced as a tool for camera calibration and foot point cloud coarse registration, which was performed directly on the point cloud. Second, the normal vector and curvature information were introduced into the iterative closest point (ICP) to make the fine registration more accurate. Then, two methods were proposed for 3-D reconstruction and visualization to overcome the problem of difficult reconstruction due to the lack of a plantar point cloud, and the fly noise in the point cloud was eliminated using Euclidean clustering segmentation. Finally, a foot 3-D reconstruction and parameter measurement system was built, and a large number of experiments were carried out. Experimental results show that the 3-D reconstruction and visualization of the foot as well as the measurement of the parameters are good. The average errors for foot length, foot width, plantar circumference, and heel width were 1.104, 0.701, 1.590, and 0.451 mm, respectively. Moreover, the 3-D reconstruction and parametric measurements of the foot in this article are highly accurate and stable, which can meet the needs of personalized shoe customization.