Regresar

Object Detection and Tracking Based on Artificial Vision for a Single Board Computer (SBC)

Abstract:

In the field of computer vision, detecting and tracking objects is an area on demand.-For this reason, algorithms specialized in tracking any object have been developed. However, those algorithms are unable to initiate the detecting process automatically since users are required to manually draw a bounding box. A potential solution is to merge an object detection neural network with a tracking vision-based algorithm. Therefore, this research proposes an algorithm developed to enhance the capabilities of KCF, an object tracking algorithm, by combining it with SSD MobileNet V2, a neural network for object identification. The proposed algorithm is developed so it can be executed on a Single Board Computer (SBC). Hence, it is possible to run and process real-time video on a Raspberry Pi 4.