Design of Embedded Target Tracking System Based on MobileNet and KCF

Abstract

With the development of computer technology, people pay more attention to target detection and tracking technology. The computer vision technology can be used to detect and track the target in the monitoring scene. However, the tracking algorithms are usually programed on the computer platform and cannot be utilized for automatically tracking, which severely limits the application scope of the target tracking system. In this paper, we construct a target detection and tracking system based on the embedded development board and image analysis technology. The int8 quantized MobileNetSSD network model is utilized and the improved kernel correlation filtering (KCF) target tracking algorithm is transplanted to the embedded development board. The video data is collected through the external camera and the tracking information is output to the display screen in a real time. At last, an experimental platform is built to verify the feasibility of the tracking algorithm based on the embedded development board.