Modernized Wildlife Surveillance and Behaviour Detection using a Novel Machine Learning Algorithm

Abstract

In a natural ecosystem, understanding the difficulties of the wildlife surveillance is helpful to protect and manage animals also gain knowledge around animals count, behaviour and location. Moreover, camera trap images allow the picture of wildlife as unobtrusively, inexpensively and high volume it can identify animals, and behaviour but  it has the issues of high expensive, time consuming, error, and low accuracy. So, in this research work, designed a novel wildlife surveillance framework using DCNN for accurate prediction of animals and enhance the performance of detection accuracy. The executed research work is implemented in the python tool and 2700 sample input frame datasets are tested and trained to the system. Furthermore, analyze whether animals are present or not using background subtraction and features extracted is performed to extract the significant features. Finally, classification is executed to predict the animal using the fitness of seagull. Additionally, attained results of the developed framework are compared with other state-of-the-art techniques in terms of detection accuracy, sensitivity, F-measure and error.