At present, the size of the national supermarket dotted, supermarket queue phenomenon often occurs. The supermarket queuing process is lengthy and inefficient, which not only wastes the waiting time of customers, but also is not conducive to the profitability of the supermarket. To this end, this paper designs a supermarket intelligent shopping cart, comprehensive consideration of application requirements, to achieve face recognition login, self-service scanning code shopping, automatic positioning follow and operation interface visualization and other functions. Aiming at the design of the operation interface, the Raspberry Pi 4B board is the core controller of the operation software. The operation interface is developed based on PyQt technology and includes three functions: face recognition login, commodity activity recommendation and self-service code scanning and billing. The face recognition algorithm uses MobileNet as the backbone feature extraction network, based on the MTCNN framework, to achieve fast and accurate face recognition. The product activity recommendation is developed based on html, and the supermarket web page is embedded into the PyQt interface to realize the real-time promotion of supermarket activities. Self-service code scanning billing uses embedded code scanners and combines the created commodity forms to realize accurate billing of commodities. For the design of automatic follow, the Arduino mega 2560 is the core controller of the automatic follow system. Based on UWB ultra-wideband technology, the PDOA algorithm is used to achieve high-precision indoor positioning. The movement of the car uses the Nam wheel, controls the speed of the DC motor based on PID technology, and collects the obstacle avoidance information through the laser sensor in the process of moving forward. The main usage process of shopping cart is as follows: customers use the operating software on the touch screen to bind their accounts and enter the operating system; Wearing a positioning label, the car follows the target in real time through the positioning label position; The user takes the goods and aims the bar code of the goods at the scanner under the touch screen to realize self-service shopping. When checking out, click the checkout button on the operation interface to automatically jump to the Alipay payment interface. After testing, the single recognition time of face recognition algorithm is 0.6s, and the recognition accuracy is 95.42. The distance error of indoor positioning algorithm based on UWB is 6.5cm, the angle error is 4.25, and the maximum distance is 80m. Users can obtain good interactive effect by operating the software, realize self-service and easy shopping, and the hardware cost is low, which has a good market prospect.