For video surveillance systems, this work proposes a method for detecting a moving human using the background subtraction method. This paper focused on the major challenges and its applications. We present an effective scheme for motion-based human detection using statistical parameters that are based on the background subtraction method. The suggested method developed the proposed scheme using varying threshold and learning rates. The threshold is generated automatically for each pixel in every frame; this method assigns the learning rate depending upon two different weighted parameters. This methodology resolves the problem of using constant threshold and learning rates. These parameters are updated adaptively. Using Internet of Things (IoT)-based devices and cameras, this work is applicable in and IoT cloud environment. The proposed method runs more quickly and achieves better accuracy with minimum false alarms. The experimental result and analysis established better performance of the developed method against the peer methods.