Falls at home is one of the most common health issues facing elderly people who live alone.Fall detection is an active research area in the smart home and smart healthcare system field. The sensor-based devices that the senior can wear or keep inside the pocket are available in the market.While they are easy to use, people forget to wear them due to their age factor. Senior people are now being monitored for their safety at home using a video surveillance system. But the lacking of the system isthat it doesonly monitors and not detect the fall.Due to this, the system does not help people to get help on time.Otherwise, a person should monitor the videoscontinuously, but it is a very tedious task. This work proposes a new vision-based real-time fall detection algorithm to address this issue. It is cost-effective and ensures the safety of senior citizens who live alone.In this instance, the fall is detected and an alert is made based on sudden changes in the human object model and the idle state continuing for more than 180 frames.The publicly available video datasets are utilized for implementation. The simulation result shows that the new vision-based approach achieves a high detection rate of 96.77%within less processing time.