Even though thermal infrared images captured during night time are available in some publicly available datasets, such images acquisitioned in adverse weather conditions such as low light, dust, rain, fog etc. are not reported as yet to the best of our knowledge. Because of these deficiencies, object detection techniques applicable in weather affected night thermal infrared images have a very limited reporting in literature. In the present scope, we discussed the acquisition, creation, design, and ground truth annotation of a new video dataset consisting of nearly 60 videos representing 4 atmospheric conditions: low light, dust, rain, fog, named as Tripura University Video Dataset at Night time (TU-VDN) in adverse weather conditions, suitable for this purpose. The objective is to provide a night video dataset containing moving objects with annotated ground truth in the image frame sequences. Using TU-VDN a comparative study is made between the results of ten existing state-of-the-art moving object segmentation methods.
