Development of early warning system for human wildlife conflict using deep learning, iot and sms

Abstract

Human-wildlife conflict is a significant challenge to communities living in areas close to wildlife game parks and reserves. It is more evident in the United Republic of Tanzania whose economy depends on agriculture and wildlife tourism as a significant source of income for her citizens and foreign exchange respectively. The proposed system is a low-power and low-cost early-warning system using deep learning, Internet of Things (IoT) and Short Message Service (SMS) to support human-wildlife conflict response teams in mitigating these problems. The proposed system comprises three basic units: sensing unit, processing unit, and alerting unit. The sensing unit consists of a Global Positioning System (GPS) module, a passive infrared (PIR) sensor, and a Raspberry Pi camera. The PIR sensor module detects animal nearby using its heat signature, the GPS collects and records the current system location while the Raspberry pi camera takes an image after the PIR sensor has detected the animal nearby using its heat signature. The processing unit with the main unit uses a Raspberry microcomputer to perform image inferencing using the “you look only once” (YOLO) algorithm and data processing. The last unit is an alerting unit that uses Global System for Mobile Communications module to send an alerting SMS message to the community response team leader and the human-wildlife conflict response team whenever wild animals are detected near the park’s border. Therefore, the system detects, identifies, and reports wild animals detected using SMS. General Packet Radio Service cellular network provides internet connectivity for the purpose of data collection to enable monitoring and storage in the cloud. An online visualization system was developed using google maps to show the location of wildlife detected by the camera trap. The park rangers track the wildlife online to acquire important information before the wildlife wanders

out of the park. This system was developed using the open-source Raspberry pi which is cost- effective even for low-income communities who are targeted by the system.