Customer's Flow Analysis in Physical Retail Store

Abstract Customer's Flow Analysis helps retailers to understand the actual behaviour and path of customers in their store, such as traffic, visited areas, customers’ paths, dwell time or loyalty. Online stores seem to “know” their customers better - correctly identifying them by name, remembering their last few purchases, or dynamically customizing the storefront to showcase relevant products. Retailers are now enabling their physical storefronts to provide a similar level of personalization. Although store cameras have historically been used as a form of surveillance to detect and deter shoplifting, stores are now tracking shoppers as they browse through a store to gather information about their target market, specifically what products consumers like and don’t like. In this paper, aim to developing physical retail store detail behavior and trajectory analysis. This paper proposed an in-store customers’ trajectory color-map system with Camera. It can also be applied to multi-flow to correspond with the demand for the practical application. First, we capture the images remotely. For obtaining the moving objects, this system uses Codebook Algorithm which applied to subtract the background dynamically. And do the morphological processing like erosion and dilation to the object and find the region of interest (ROI) often performed on using a mapping-based detection approach. Then apply Histogram of Oriented Gradients (HOG) on ROIs to sift the target out. After these previous steps, this system gathers statistics of the data and resample them. Therefore, we can obtain the color-map of customers’ trajectory according to the results. After all, our system provides a high quality in the results. And we are using this system now on those stores, like baby products store, woman's shoes store, and furniture store, etc.