High-altitude parabolic detection method based on GMM model and SORT algorithm
Yuntao Shi, Qi Luo, Tao Zhang, Cheng Yue Hao · 2022
With the accelerated urbanization process in China, the resident population is flocking to high-rise buildings. The number of high-altitude parabolic has increased significantly, resulting in many casualties and property damage accidents. To this end, a proposed technique for high-altitude parabolic detection based on Gaussian Mixture Model (GMM) and Simple Online and Realtime Tracking (SORT) is proposed. Firstly, the GMM-based background modelling method is used to separate the background and foreground from the image sequence to obtain the motion image of the dynamic target at the current moment. The image is subjected to mathematical morphological denoising to determine whether the dynamic target is a throwing object by SORT. Experimental results show that the proposed method can accurately detect the dynamic target of overhead throwing objects with good stability and can effectively reduce the false detection rate.