Object Detection At Night Time

In computer vision, object detection is an important task. Object detection has demonstrated utility in multiple fields, including autonomous driving, surveillance, robotics, smart cities, and more, however, nighttime object detection poses its own set of challenges such as low-light situations, shadows, noise, and diverse lighting conditions. In such cases, traditional methods do not succeed [4]. However, the end of 2019 has seen a rapid evolution in deep learning algorithms, especially Convolutional Neural Network (CNN), Region Based CNN (R-CNN), and You Only Look Once (YOLO) architectures, that impacted object detection in night time considerably better [2]. Furthermore, sensor fusion methods that combine infrared and visible light data have been investigated to overcome the limitations of singular sensors, thus enhancing detection performance and environmental awareness [1][5]. However, there are still some challenges, especially real-time processing, adaptability to changing conditions, and high detection accuracy across a range of environments [2].