The objective of this project is to design and build a intelligent shopping cart by appropriately utilizing the advantages of cloud services, embedded system, and mobile phone. The main functionalities to be realized are automatic bill recording, and intelligent shopping recommendation providing. The following block diagram can provide a brief elaboration of our system.
To make the shopping experience easier and more intuitive, we abandoned the bar code scanner solution, and adopted a combination of RFID, camera, and weight sensors. Packed items are identified by RFID tags, and items sold by weight, such as fresh fruits, are recognized by computer vision. Weight sensors can not only determine whether users are putting things in or taking things out, but also provide weight information to calculate the price of unpacked goods. This design allows bills of items to be automatically recorded when users adding or deleting items. The data and main computation components were set up in Amazon Web Services, which eases the management of data, as well as lowers the requirements on hardware performance. (Item recognition is powered by computer vision and recommendation is powered by machine learning. The computation force needed is not negligible.) The user interface was provided by Android or iOS mobile phone application, instead of LCD screen, which further decreases the cost of the shopping cart. The detailed cooperation between these three sub-systems and their workflow is further demonstrated in the following section.