Investigation of Image Reconstruction Algorithm Based on the Structured Light Method

Abstract

This paper proposes a new method of recognition of spatial objects on the basis of approaches related to technical vision. The developed algorithm is used to combine several point clouds by adding a target object to the scene on the example of a sphere. The algorithm introduces a restriction on some parameters, such as the sphere radius, to reduce the impact of noise in point clouds. Even for point clouds with overlapping regions, the proposed algorithm is more accurate because the corresponding points are selected from a continuous and ideal surface, instead of the actual measured points. Furthermore, the proposed algorithm is less sensitive to target size, point density, noise and the overlap factor, so it is more accurate than the algorithm of registration of the centers of the spheres and the ICP algorithm. The proposed algorithm can be used for scanning large, complex, curved surfaces where parts of the object overlap each other. Experiments and analysis of the proposed algorithm are carried out. To evaluate the proposed algorithm, three spheres were taken. Two simulations were carried out with overlapping data and with non-overlapping data. The size of point clouds is two-thirds of the hemisphere according to the field of view. Experiments were carried out on the basis of recorded data and real experimental data, respectively. The simulation was repeated several times with different points of extraction and reproduction of random noise. The statistical results were presented in the table. The practical significance lies in the possibility of using the results obtained in the study in the field of three-dimensional scanning.