CNN Based Moving Object Detection from Surveillance Video in Comparison with GMM

Abstract

Automatic moving object classification has become more important in modern intelligent detection and visual monitoring systems. Moving object recognition from pictures and vehicle type categorization is essential for an effective visual surveillance system. Moving object movement must be limited as much as feasible. Most of the research projects that have been done so far have simply looked at how to increase the percentage of forecasts, which has poor real-time performance and uses more processing power. A network categorization system used by CNN can swiftly recognize moving objects. Convolution neural networks (CNN) based on deep learning neural networks are compared with Gaussian mixture model (GMM) based object detection in this paper to increase the accuracy and F-score of the object detection. The mean accuracy for CNN is 95.69% and the mean accuracy for GMM is 93.22%. The mean F-score for CNN is 87.02 and the mean F-score for GMM is 84.77.