This Small Business Innovation Research (SBIR) Phase II project aims at developing video mining techniques for automatically generating statistics about in-store shopping behavior to help retail enterprises. These statistics can provide valuable insights for supporting critical decisions in store layout design, merchandising, marketing, and customer service. Further, since it is automated, video mining can become a tool for monitoring the impact of all customer-facing elements in a store. The Phase II research will continue in cooperation with the proposing company's partners and customers, while addressing the remaining challenges for video mining. The proposed tasks include robust person detection, tracking people across multiple cameras, modeling and recognizing complex shopping behavior involving shopping groups and sales associates. The approach will be to use a variety of computer vision and statistical learning techniques under the constraints of a typical retail environment.
Retail enterprises today operate in a hyper-competitive environment characterized by blurring categories, eroding market shares and fickle, but more demanding customers. These challenges have prompted retailers to adopt customer-centered strategies focused on uncovering and matching the needs of customers to gain (retain) market share. These strategies rely heavily on obtaining deeper insights into shopper behavior. Current methods (human observation and manual video indexing) for analyzing shopper behavior are limited in their scope while being expensive and time-consuming. On the contrary, the shopper insights gained from the proposed video mining platform will enable more informed decision-making leading to improvements in retail productivity and business process optimization. The proposing company has plans to immediately incorporate the outcome of the SBIR research into its retail product line.
