RetinaNet-Based Approach for Object Detection and Distance Estimation in an Image
Mohanad Alhasanat, Moath Alsafasfeh, Abdullah Alhasanat, Saud Althunibat
Abstract
Detecting, recognizing, and manipulating data in a digital image provides much information to the end-user. Translating this data into more meaningful information has paved the way for advancing a wide range of computer vision applications, such as autonomous vehicles, security systems, traffic monitoring, etc. In this paper, a novel method to estimate the distance between the camera and the captured objects in a scene is proposed. It is based on object detection using RetinaNet model and distance calculation based on the Similarity of Triangular. The system can successfully obtain the object distance with respect to a reference model with known size and a specific focal distance. This method provides a simple system with reliable distance estimation, which in turn makes a suitable solution for real-time embedded applications. The experimental results presented in this paper show an accurate method for camera-object distance estimation in a real-time environment. The error rate is around 5% for different experiment scenarios.