Robust Motion Detection in Real-Life Scenarios

Our knowledge of the surrounding world is obtained by our senses of perception. Among them, vision is undoubtedly the most important for the information it can provide. In artificial systems, this discipline, known as Computer Vision, mainly tries to identify physical objects and scenes from captured images to be able to make useful decisions. For that, the processing and analysis of images, video sequences, views from multiple cameras, or multi-dimensional data like a medical scanner, are carried out.
In this context, motion plays a main role since it provides a stimulus for detecting objects in movement within the observed scene. Moreover, motion allows other characteristics to be obtained such as, for instance, object shape, speed or trajectory, which are meaningful for detection and recognition. Nevertheless, the motion observable in a visual input could be due to different factors: movement of the imaged objects (targets and/or vacillating background elements), movement of the observer, motion of the light sources or a combination of (some of) them. Therefore, image analysis for motion detection will be conditional upon the considered factors. In particular, in this work, there is a focus on motion detection from images captured by perspective and fisheye still cameras. As cameras are still, ego-motion is not considered, although all the other factors can occur at any time.
With that assumption, the work proposes a complete sensor-independent visual system which provides robust target motion detection. So, firstly, the way sensors obtain images of the world, in terms of resolution distribution and pixel neighbourhood, is studied. In that way, a proper spatial analysis of motion can be carried out. Then, a novel background maintenance approach for robust target motion detection is implemented. On this matter, two different situations are considered: (1) a fixed camera observing a constant background where interest objects are moving; and, (2) a still camera