Shelf detection via vanishing point and radial projection

Retailers need to monitor products on store shelves in order to maintain adequate stocking and in many cases to comply with a placement arrangement defined by a planogram. It is possible that such inspections be performed automatically or semi-automatically using computer vision techniques. Towards this goal, we propose an algorithm to detect shelves from images captured with a handheld digital camera. First, we identify the vanishing point and the associated line segments corresponding to the shelves. Second, we divide the image into equal-angle wedges centered at the vanishing point and project the associated line segments into the wedges. Finally, we identify shelves by analyzing the projections.