Visual mapping of target tracing methods based on CiteSpace bibliometrics

Abstract

Visual target tracking is a hot problem in computer vision research in recent years, and the application fields are gradually increasing, such as unmanned driving, intelligent surveillance, etc. In recent years, with the wide application of deep learning in the field of computer vision, the field of target tracking has also developed rapidly, and many scholars have improved and innovated the target tracking. Common target tracking areas include single target tracking, multi-target tracking, pedestrian re-identification, multi-target multi-camera tracking, pose tracking for many complex scenarios. Eventually the problem will also be attributed to single target tracking or multi-target tracking, which has become the focus of many scholars' research. In view of the rapid development of this field, this paper presents a review of visual target tracking research, including a review and analysis of single-target tracking methods, a review and analysis of multi-target tracking methods, and a summary of the shortcomings of these methods, including the lack of fusion based on target detection methods, the decrease of real-time accuracy, and the problem of target loss in long-term target tracking. . And according to the shortcomings, the following suggestions are made: combining traditional algorithms based on filtering and deep learning algorithms, focusing on the improvement of the ideas of deep learning frontier theories, and improving the long-term target tracking loss.