A Real-Time Thermal Image Processing Using Deep Convolutional Neural Network (DCNN) for Monitoring Intrusion of Elephant

Human Elephant Conflict (HWC) is a major threat in Valparai region of the Anamalai Tiger Reserves (ATR). Intrusion of elephant into the human settlements is more in Valparai and there is a need to get an early intimation to people regarding the animal movements and to be in a safer area. Visual animal detection is rapidly very useful as it’s cost-effective and non-invasive for monitoring wildlife. Camera trap usage holds a huge volume of data or collected images, with this manual processing it is possible to process further. In this work, we develop a framework for automatic recognition of elephant with the help of thermal imaging. Most existing system manually use input image for localization of animal and doesn’t scale well for large dataset. In this experiment, we collected 200 thermal images of elephant in different poses to analyze the performance of our method. The proposed approach uses a Deep Convolutional Neural Network (DCNN) for elephant detection. These simulation results showed that our approach can effectively improve elephant monitoring even if we get occluded images.