A Systematic Approach for Object Detection in Unconstrained Environments

Object detection in unconstrained environments could be defined as identifying objects captured in challenging environments like smoky, wild, and cloudy environments. Object detection is widely used in computer vision applications. It also has many applications in security surveillance, where we often have to deal with unconstrained environments. Several algorithms and techniques have been developed for object detection over the past years. Overall, object detection approaches can be divided into two main types: conventional computer vision-based approaches and modern deep learning-based computer vision approaches. This study reviews different methods and algorithms developed in recent years for object detection in constrained environments. Most object detection methods in constrained environments produce good performance. However, object detection in unconstrained environments is still an open problem that requires unique methods. So far, only a handful of approaches have been proposed. This study also presents future development trends and research directions for object detection in unconstrained environments to produce automated intelligence security surveillance. This review's main conclusion is that object detection in unconstrained environments needs to be improved and well-established using unique computer vision methods based on modern machine learning and deep learning technologies.