Moving Object Detection Based on Improved Gaussian Mixture Model and Three-Frame Difference
Lihua Li, Yang Yang, Maohua Liu, Xiaowei Lv, Yongrong Li, Ning Yang · 2023
The proposed algorithm is based on an improved Gaussian Mixture Model (GMM) and a combination of three-frame difference for addressing the issue of low detection accuracy in complex backgrounds such as abnormal closure of doors, suspended objects, and unauthorized personnel intrusion. Firstly, the improved three-frame difference method can rapidly distinguish between background and targets in the video, and it can also address illumination changes. Furthermore, the algorithm incorporates adaptive selection and background updating strategies into the Gaussian Mixture Model (GMM), which significantly improves modeling efficiency and accelerates the elimination of "ghosting" artifacts. Lastly, morphological operations are applied to the results to obtain moving objects, and the performance of the algorithm is validated through comparative experiments with simulations. The experimental results demonstrate that the proposed algorithm can effectively detect moving objects in complex environments with high accuracy.