Falls are an important issue for the elderly, hence identifying them and responding to them without delay are often the most effective preventive measures. In no case should the life or well-being of a senior citizen be threatened. In this work, we propose a novel approach for fall detection using video sequences as input, which is based on position estimation. Our method consists of two main stages: Metrabs pose estimation and shadow suppression. Metrabs pose estimation is an AI-based system that accurately tracks and detects body postures and movements in video sequences. A good way to evaluate human posture nature is suppressing the shadows so as to reduce reflections noise. Shadows are not advanced in the video frame quality control technology, and this limits their effectiveness. Since dark regions have the highest levels of shadows and reflections, our shadow suppression method is designed to minimize false positives from such regions and the still frames containing them, we thus reduced false alarm rates. The process involved extensive testing of numerous videos with cases of falls and non-fall events. This was done to test its feasibility. In addition, we carefully chose our dataset so that we have a representative sample of different real-world situations as well as meteorological factors affecting such scenes from within it. We also ensured that the chosen dataset reflects various real-life scenarios based on which meteorological conditions interact. According to our test, falls can be identified with a 95% accuracy basis. The high detection rate indicates that this technique can be employed for fall detection through extraction of human locations from images. In addition, this study examines how computer vision techniques integrated with machine learning algorithms can promote elderly safety thereby reducing fall injuries Our research presents an original fall detection system that estimates human positions from video sequences. With Metrab’s posture estimation together with shadow removal procedures, our system is capable of recognizing falls even in harsh environments.