Lightweight fall detection system based on Orangepi 5B

Fall detection plays an important role in the healthcare of older adults. Falls are reported to be the leading cause of death in people over the age of 65. For fall detection, computer vision-based methods are currently the best performing methods. This paper proposes a fall detection system with a Orangepi 5B development board as the core processor and a shape and motion change method. First, we use the VIBE detection algorithm to extract human silhouettes from the video, and improve the effect through certain post-processing. Then, the information obtained by ellipse fitting and the vertical and horizontal projections in the projection map are used as features to distinguish different postures of the human body. The abnormal downward movement speed measured by the optical flow method was used to preliminarily judge the fall, and then the occurrence of the fall was determined by the auxiliary judgment of the posture feature. Experiments show that our fall detection system achieves good performance in terms of both accuracy and error.