Night Vision Thermal sensor based Animal Movement Observation using CNN and YOLO v3

Abstract

Developing an automated system for wildlife detection and recognition using thermal cameras can be challenging nowadays, which have many applications in wildlife conservation and management. This includes many challenges like detecting false positives and negatives and also environmental factors like temperature and humidity. Developed a system for the classification of animals using one of the developed algorithms, the computer-aided CNN algorithm and Python Flask web application that loads a pre-trained convolutional neural network (CNN) model for image classification (cat or dog) and allows the user to upload an image for classification. The wildlife conservationists need a system, which uses a thermal images or videos and detects the animal presence in video. Temperature difference is used to distinguish the animals and can develop the system which uses images to detect the animals in a given video. Using this option, the conservationists will identify the animal using the developed system which is built with yolov3 (You Look Only Once, version 3) algorithm. Finally, if a pet animal is surrounded by a wild animal, or if both pet and wild animals are present in the video, an alarm indication is given in order to protect domestic animals from the wild animals and also safeguard the agriculture fields