A Systematic Review of Animal Detection by Using Vision-Based Techniques

Abstract

Animal detection from videos has become a popular research area due to its wide range of applications in wildlife monitoring, conservation, and animal behavior studies. In recent years, computer vision and machine learning advancements have allowed the development of accurate and efficient animal detection algorithms. In this paper, we review the techniques used for animal detection from video streams, including object detection, feature extraction, and motion detection. We also discuss the challenges associated with animal detection, including variations in animal appearance, changes in illumination and background, occlusion, and limited training data. Moreover, we present a systematic review of the deep learning-based animal detection literature, highlighting the pros and cons of each category of approaches. We start by giving a concrete introduction to the topic by outlining the definition, background concepts, and fundamental notions of algorithms within this field of study. Subsequently, we summarize the datasets for training and testing animal detection algorithms, common challenges, and evaluation metrics. Finally, we present the future directions in this research area, including the use of multi-modal data and deep learning techniques.