Human Detection Surveillance Camera Based on Face Recognition for House by Using Movidius Neural Compute Stick 2
Keywords:Movidius neural compute stick, HD-SC, SMS, deep learning
At present, video surveillance is not new for house owners to have. It is essential due to the increasing crime and theft rates around the community. It can provide video footage, whether live or recorded, allowing the house owners to keep an eye on their property while they are away. It also helps the security personnel to identify and solve the crimes by reviewing the recorded videos. Unfortunately, reviewing the footage took a long time. Thus, our device (HD-SC) allows the house owner to be notified in real-time through the Internet and send short message service. Using the prototyping method, we created a preliminary design to build this device. With the aid of Movidius Neural Compute Stick, we managed to build a program structure of deep learning broader family of machine learning based on artificial neural networks. We also conducted a survey and follow-up questions that lead to a mini-interview, which enhanced our device based on the result we gathered from house owners that already own a surveillance camera on their house. Afterward, we refined and integrated our device based on data collected through quantitative analysis. On our device, we perceived the following substantial problems with the assessment innovation. 1.) Technical problems –Low CPU usage causes a bottleneck of programs. A poor Internet connection can causes a delay in notifications and camera quality that causes failed recognition. With these problems, the device cannot function well. Similar issues like this is cannot be avoided since our device is a prototype that needs considerable improvements; 2.) Follow-up Remarks –Good feedback from the house owners from having a surveillance camera that can identify any person that may have contact on the perimeter of the camera. Having SMS and smartphone notifications, they are satisfied with this innovation. Nonetheless, they also recommended some functions that we did not include recording video footage and saving video footage. The results we indicated were based on the assessments we observed and the comments of our respondents for our device to work efficiently. Our device can compare to other innovations of surveillance cameras such as facial recognition and smartphone integration. However, we have a SMS notification that can notify the user without an Internet connection on the phone. Hence, it is not enough to secure the house. Future researchers can use this study as a basis to implement the solution to the same problem that they want to enhance in the future.