Improved Mixed Gaussian Model for Background Subtraction Based on Color Channel Fusion

Motion object detection, which has a wide range of applications in video surveillance systems, extracts motion objects from the video. Gaussian Mixture Model (GMM) has achieved widespread success in motion object detection because of its good performance. However, it simplifies color channel information or directly uses grayscale information. If the multi-color channel information can be jointly used, it is expected to obtain better results in complex scenes. Therefore, we design the multi-color channel voting GMM by jointly using the multi-color channel information and introduce soft voting based on soft decision to further strengthen the use of information. Experimental results show that the multi-color channel voting GMM proposed in this paper can well detect motion objects in complex scenes. Compared with original GMM algorithms, multi-color channel voting GMM has a better F-measure metric in the complex scenarios.