||Resemblances between bee social structure and human society may be useful in human epidemiological research. To comprehensively record all bee movements within a hive, bees mounted with custom designed tags on the dorsal part of their thoraxes were filmed under infra-red light for the span of 5 days. In this work, two methods of processing the video are explored. First, background texture was learned using K-means clustering allowing the background to be removed while preserving bee edges. Edges are important for detecting communication and food exchange among bees. Second. a dataset of tag images was created. The background was estimated using Gaussian regression over the span of one hour and then removed. The resulting image was lighting compensated using a the blurred background estimate. Potential tags were initially detected using an adaptive Laplacian of Gaussian filter and classified by a human operator. The dataset was used in training a Viola-Jones Boosted Cascade to detect the tags in the raw video. A larger dataset need to be used for getting the accuracy above 90% level although the method looks promising at this stage.