DBSCAN Based Object Localization with Ultrasonic Spherical Scanning

Abstract

Object localization is an important problem in many areas such as robotics, automation and intelligent systems. Object localization can be used in areas such as indoor mapping, obstacle avoidance and non-line-of-sight (NLOS) object detection. In such studies, expensive technologies such as LIDAR are generally preferred. In this article, we address the object localization problem using scan data from a single ultrasonic sensor in spherical coordinates. Considering the noise and gaps in the measurement data, DBSCAN, a density-based clustering algorithm, was preferred in this study due to its robustness. In the proposed method, an ultrasonic sensor rotated by servo motors scans the environment by taking distance measurements from different angles. This set of spatial data points is analyzed by the DBSCAN algorithm to identify clusters representing physical objects. The method successfully identified both the quantity and spatial distribution of objects without requiring prior information. The sensor location and walls were also detected using this method. In experimental evaluations, success was achieved with objects placed in different numbers and positions, and it was seen that the method was particularly robust against measurement errors caused by noise. © 2025 IEEE.