Human Recognition with Yolo to Reduce False Alarms in the Internet-of-Things-Based Motion Detection System

Abstract

Security is a serious matter that must not be ignored. When dealing with a Smart Home or Office, an Internet of Things-based approach is preferable to safeguard the physical system. Smart security devices offer many features to help homeowners to secure their houses. However, there are several issues in terms of storage and motion detection since motion detection cannot recognize what or who caused the movements, and unencrypted data is vulnerable to theft. In fact, the increasing exposure of the detected motions produces false alarms and false reports. Therefore, this study implements human recognition by adding redundant Blockchain and YOLO algorithms in motion detection. The findings have shown in the evaluation process that the proposed model only needed 58.362% of the CPU and 110.123MB of memory allocation. Meanwhile, the proposed model temperature reached 32.33 degrees Celsius on average. As a result, the execution of the proposed model, which is the human recognition with the YOLO algorithm successfully lowered the false alarms over the previous model that is deployed without the YOLO algorithm.