Menglin Yin¹, Yong Qin^1,2,3,4,*, Jiansheng Peng^1,2,3,4

CMC-Computers, Materials & Continua, Vol.82, No.1, pp. 1329-1347, 2025, DOI:10.32604/cmc.2024.057460 - 03 January 2025

Abstract In dynamic scenarios, visual simultaneous localization and mapping (SLAM) algorithms often incorrectly incorporate dynamic points during camera pose computation, leading to reduced accuracy and robustness. This paper presents a dynamic SLAM algorithm that leverages object detection and regional dynamic probability. Firstly, a parallel thread employs the YOLOX object detection model to gather 2D semantic information and compensate for missed detections. Next, an improved K-means++ clustering algorithm clusters bounding box regions, adaptively determining the threshold for extracting dynamic object contours as dynamic points change. This process divides the image into low dynamic, suspicious dynamic, and high More >

Fan Li¹, Yang Yang

CMES-Computer Modeling in Engineering & Sciences, Vol.107, No.2, pp. 103-124, 2015, DOI:10.3970/cmes.2015.107.103

Abstract The shift of the camera leads to unsteadiness of backgrounds in video sequences. The motion of camera will results in mixture of backgrounds and foregrounds motion. So it is a challenge for targets detection in dynamic scenes. A realtime moving target detection algorithm with low complexity in dynamic scenes is proposed in this paper. Sub-block based image registration is applied to remove the global motion of the video frame. Considering the blocks in one frame have different motion vectors, the global motion of each block is separately estimated. Then, a neighbor-based background modeling is applied More >