This project is started to catch the person who roams near my house. I don't know my house is perfectly safer than other house. We need to protect by ourselves to keep family.
Basically CCTV is able to capture the image when some one comes and goes like black box in the car. But there are required to work during the day. How to detect the people during the night? Are you going to turn on the light? I don't think the way is good because a stranger will run away if the light is on during the night. In addition to that, If you buy the solution which includes all functionalities in the market, it should be pretty expensive. Therefore I made the solution with RaspberryPI2, Webcam, PIR sensor and Light sensor.
Here is description what it has the functionalities.
Motion DetectionPIR Sensor which can be detected when someone is walking or running in front of the sensor. It works based on Infrared sensor inside of PIR. Once I have tested it with 16 X 2 character LCD.
The sensor is popular to use in IoT area. It works using I2C communication between RaspberryPI and the sensor. If some one is detected, the solution will check the light status in order to turn on IR LED along to web cam. I have tested the mechanism using Light sensor and 3-color LED. According to Light value, the LED will be changed with different color(RED, GREEN, BLUE)
MS provides API's to people's face. Oxford project is ready to use for cognition. I also signed up the site and took the security key to call the API's.
The UI provides supervisor page to register user based on photo ID and there is the button on the bottom of the UI and able to test how much the photo is similar to users' face.
IR LED to take a photo at the nightThe solution is going to set up outside or in front of my house to keep checking. Even though the status of the brightness is too dark, it can be detected due to IR LEDs.
However web camera will not be detected to IR because of IR cut filter. You have to remove the filter inside of webcam.
Now, Let's check the result with all features in the solution.
Shape of the solution is here. All of stuff are not fixed but it works based on the scenario. If it covers with boxes, it will be done.
Between the matched rate during the day and night, it will not be lower around 10~15% at night rather than the day due to the brightness. However if I configure threshold at the night, it can be better than now. It someone is not in the user list, the message will show up "CAUTION" and save the image in the RaspberryPI.
No more I worry about my house from strangers.
Comments