An Approach for Estimation of Distance Information Between Two Persons from Single 2D Image

This paper presents a new approach for distance determination between two persons based on human height reference using model with single monocular 2D camera. Depth estimation from single 2D image is very difficult problem. Local and global features of the image are important for depth estimation. In this paper, human height feature is used as reference for distance estimation. First, the image is captured from the 2D camera and converted into grayscale image, and then the presence of a human is determined based on person detection. Separation of images according to the number of human and segmentation of each image is performed. Next process, the depth information estimation, is performed for each separated image (Fidan et al., IEEE Trans Ind Elect 60(12), 2013) [1]. Finally, distance is measured using human height and distortion value. For this depth estimation problem, supervised learning approach has been utilized and corresponding depth model is built.