Multi-sensor based object tracking using enhanced particle swarm optimized multi-cue granular fusion

In the discipline of computer vision, object tracking is one of the progressive and prominent areas of research with its application in the field of medical imaging, vehicle navigation, surveillance etc. Many of the proposed object tracking algorithms has shown success in the recent years. In this manuscript, we introduced a novel approach for object tracking that can develop an efficient framework of various features from different sensors. If we contemplate a RGB (red–green–blue) image it has better distinction of colors from human eye standpoint but is degraded by shadows and noise caused by illumination. Unlike RGB images thermal images are less receptive to such type of noise factors yet environmental condition can alter its distinction. To overcome this distinction issue of the two sensors a fusion of these two modalities is introduced, considering their interdependent advantages. This proposed technique is focused at enhancing the information collected from the fusion of visible imaging and thermal imaging sensors. It can also be implemented if the number of sensor are increased which in turn increases the number of features. With the use of features from two different sensors the proposed scheme utilizes the six information cues for the estimation of single output. The EPSO (enhanced particle swarm optimization) based particle filtering was adjusted with the concept of using multi-cue granular computing to weigh the particles and estimate the ultimate tracking result. After conducting attribute weight adaptation, the same approach is expanded to produce source-level fusion. The experimental performance of the method has been demonstrated on publicly available standard video sequences. After comparing it against state-of-the-art approaches, the findings show that it outperforms the trackers mentioned in the literature.