Hardware components | ||||||
 |
| × | 1 | |||
 |
| × | 1 | |||
Fire Fight with Smart Drone
Author Roy Zang <roy.zang@nxp.com>
1. Overview
Wild fire can damage environment, property and even kill people. The recent Austrian fire is a disaster for local resident and wild life.
The modern drone technology provides new way to prevent, monitor fire. The information sent back by drone can provide guild for fireman to fight wild fire.
Drone can fly to the place where fire man can’t reach.
This project utilizes the NXP HoverGames drone kit, Melexis MLX90614 IR sensor, USB camera together with a companion computer to find a hot spot ( fire ) and use the companion computer to do human face recognition. The result can be used by fire department to make the decision.
The block diagram of the smart fire fight drone system describes as following:
2. Smart drone fire fighter assistant
Following the hover game build guild [1] to build the drone platform.
3. AI people detection
It is critical to get video stream of the fire. Extremely important, to find whether there is people that needs help in the fire.
To implement video processing, the processing power of M4 in the FMU is not enough. M4 is also running the PX4 flight stack, that is critical and real time task for drone. If M4 is also used for video stream processing, it potentially can impact the fly stack. So the project does not go with M4 as the video processing processor.
This project considers two candidates as companion computer:
· i.mx8mm evb board
· Raspberry pi 3 Model B
A compare between i.mx8mm evb and Raspberry pi 3 model B as following
i.mx8mm is more powerful and standalone test also proves that i.mx8mm performance is better than raspberry Pi 3.
However, considering the size and power supply, the raspberry pi 3 is chosen and powered by USB via a battery.
4. Camera
Pixy 2 camera[2] can work with raspberry pi directly with downloaded application. However, it does not appear as a Linux video device, for example /dev/videox.
Based on a better device driver support, a ELP 2.1mm 5megapixel Hd USB Camera[3] is selected to capture the video stream.
5. Capture result
The capture control panel can be accessed via wifi from computer or smart phone. The following pictures are static result around 5 meters between people and camera.
6. CPU usage
When doing the people search, around 50% raspberry pi cpu usage measured by mpstat command.
7. Conclusion
The project uses drone as fire fight platform equipped with thermal sensor and camera. Raspberry pi 3B board is used as companion computer based on its size and battery power supply. Video stream is processed locally on drone with raspberry pi. The people recognition result can be accessed remotely via wifi.
Multiple people can be captured and recognized by the smart drone platform.
8. Acknowledgement
Great thanks for the support from Iain Galloway and the hover games team providing the drone kit and answering stupid questions.
9. Reference
1. https://nxp.gitbook.io/hovergames/
4. https://openweb.co.za/how-far-will-your-wifi-signal-reach/
Comments