Multiple Objects Localization With Camera-LIDAR Sensor Fusion

The objective of this work is to present a cost-effective solution for the precise detection and simultaneous 2-D localization of multiple objects based on the fusion of camera and Light Detection and Ranging (LIDAR) data in real-time 3-D space in the context of smart vehicles and robotics applications. The majority of approaches have concentrated on the utilization of sensors, such as 3-D LIDAR or stereo cameras, for the purposes of object detection and localization. However, these sensors are both costly, which limits their accessibility for large-scale applications, and face specific challenges. In particular, the existing literature has not sufficiently addressed the issue of multiobject detection and localization based on camera-2-D LIDAR fusion and its comparative analysis with the conventional stereo camera. In this study, an innovative method has been developed that enables the simultaneous 2-D localization of objects. In the object detection process, the You Only Look Once version 7 (YOLOv7) model was employed to achieve high-accuracy object detection. The bounding box information generated by YOLOv7 and LIDAR data were used for object localization. This study compares the proposed approach with a conventional stereo camera-based localization method. To ensure a fair evaluation of the two methods, a special mechanical design has been developed, integrating all sensors. The results demonstrate that the proposed method significantly improves localization accuracy compared to the stereo camera while offering a more cost-effective solution.