3D Object Tracking with Reduced Cumulative Error on Sparse Point Clouds

In 3D target tracking, the cumulative error generated in the sparse point cloud environment makes tracking more difficult, which is a key issue to track the target stably to allow the lost target to be retrieved. This paper proposes a Bayesian theorem-based 3D single-target tracking algorithm to improve the subsequent tracking performance by reducing the cumulative error caused by low-quality prior information. First, we use Bayes' theorem to classify the prior information to obtain the maximum a posteriori probability, and second, we suppress the proposals with poor prior information to re-duce the cumulative error. Finally, the algorithm proposed in this paper is applied to the dataset, and the study shows an improvement of about 23% in real-time performance over the PTT-Net algorithm and excellent performance in several tracking metrics.