Early warning system for human-elephant conflict in Sri Lanka

Abstract

Elephants are essential indicators in an ecosystem. Human-elephant conflict (HEC) arises when elephants move to human livelihood. HEC is one of the significant problems in Sri Lanka. A variety of technologies have been developed to mitigate HEC. However, existing strategies are designed only to protect humans and their harvest from elephants. The proposed model identifies elephants before they invade villages. Therefore, save human lives, their property, and elephants. The model is built in combination with image preprocessing techniques and a convolutional neural network to detect the elephants. The model has achieved over 98.1% of accuracy with the test set. A trained model and classifier algorithm are applied to detect elephants in the video. After detecting elephants, a message is sent to relevant parties, and the ’bee sound’ is emitted. The developed model will further be improved to detect elephants in motion video and to apply the Internet of Things to HEC.