Forest Guard

The forest fire or agriculture fire is one of the biggest problem in the world today. The forest fire creating climate change effects, loss of lives of animals, health problem to the humans and bio diversity loss.

So it is necessary to timely detect the forest fire and notify the authorities, so quick action shall be taken.

In my project I connected a camera to Xilinx Kria board, where CNN based fire detection algorithm is running and it will identify the fire using the image detection.

We used reference project from the (Thomson, Bhowmik, Breckon), In Proc. International Conference on Machine Learning Applications, IEEE, 2020.) for the light weight and efficient computing of the Fire.

using the reduced complexity CNN running on the Kria board we will be able to achieve the accuracy of the 95% over the full frame binary classification and 97% using super pixelate classification.

If the fire is identified then the whatsapp notification is sent to the authorities using the Twillio API. It shall also send the image frames to the authorities.

the live streaming of the fire detection shall be made using is make using the OBS studio APP.

Test setup:

Results:

Future scope:

The project shall be be extended by adding Real time streaming using OBS studio, Extending the Image processing to the block chain based processing and Designing solar based power supply for deploying the solution in the forest and agriculture applications.