Recently, NEC has developed a variety of object recognition technologies that can significantly increase the efficiency of product identification at the time of settlement by retailers such as supermarkets and convenience stores, and can perform images for all retail products including fresh food, daily necessities, and finished packaging products. Identification. This technology will make a huge contribution to the realization of unmanned settlement for retail stores.
This technology combines deep learning techniques and feature point matching technologies with different characteristics to identify various retail products with high accuracy, including items such as fresh foods with large individual appearance, as well as similar designs such as packaged products. Existing pipeline products. And even if these goods are stacked together, they can accurately identify all products at once.
NEC product image recognition
In recent years, with the continuous development of image recognition technology, the use of cameras to identify image recognition methods for retail products has made retail store settlements continue to improve in terms of manpower savings and unattended. However, due to differences in the characteristics of fresh foods and finished packaging products, The image recognition technology is difficult to uniformly identify all kinds of goods. In addition, in order to accurately identify these goods at one time, the goods need to be neatly placed, which will increase the burden on users.
With the latest technology developed by NEC, it is no longer necessary to read barcodes and RFID tags one by one. Simply place the goods on the cash register. This simple operation allows all products to be identified at one time. Reading efficiency.
NEC is actively committed to promoting 'social solutions business' globally, creating safe, secure, efficient and fair social values. Integrating advanced ICT technologies and knowledge to achieve a brighter, more colorful and efficient society A force.
'background'
In recent years, the shortage of labor in the retail industry has become increasingly severe. How to save manpower in store operations and realize unattended has become an extremely important issue for the time being. In particular, there is a heavy burden on the clerk (banking checkout), unmanned Demand is increasingly pressing.
In order to achieve unmanned settlement services, self-checkouts can be considered. Self-service checkout requires the customer to scan the barcode of the product one by one on the machine itself. However, because some customers are not familiar with how to use a barcode reading device, Therefore, it takes a lot of time to read barcodes. For products without barcodes, such as fresh foods, the correspondence at the time of self-checkout becomes more complicated and billing efficiency becomes a problem. In addition, although companies have started to give products, RFID tags are used for automated settlement, but it is not cost-effective to add electronic tags to low-priced goods. Therefore, it is still far from achieving a true sense of popularity.
The image recognition technology developed this time is a multi-object recognition technology that eliminates the need for barcodes and RFID tags, even in retail stores that mix fresh foods, daily products, and finished packaging products into a complex environment. , also can carry on the accurate identification. Through this technology, the customer only needs to put the purchased multiple commodity on the cash register, can read the commodity accurately once.
'Features of new technologies'
1. High-precision identification of retail products of various characteristics, including natural products and assembly lines
The image recognition technology developed this time incorporates different features in deep learning technology and feature point matching technology. First, it infers the recognition accuracy when using deep learning technology and feature matching technology to each commodity, and adjusts each technology. Use the ratio to maximize the recognition accuracy. Then, according to the ratio obtained after adjusting each product, integrate the recognition results of each technology. For natural products such as fresh fruits and vegetables and other fresh foods, the appearance of different individuals is huge. , Deep learning techniques that can discriminate between different individual differences will be used, and similarly designed industrial products such as beverages, snacks, boxed instant noodles, and grocery products will use feature points that can discriminate design configurations in detail. Matching technology.
For goods such as bento, cooked foods, and other daily necessities that have both natural (ingredients) and industrial (product label) characteristics, two technologies are used at the same time to identify the results and then combine the results. , The integration of the two technology identification methods, compared to the use of only a single technology, effectively improves the accuracy of identification for a variety of different retail products.
2. Accurately identify even complex environments where multiple commodities are mixed
By photographing each commodity individual, you can automatically generate a large number of pictures of complex environments in which multiple commodities are mixed. In this development of various object recognition technologies, mechanical learning of these images is required in advance to effectively enhance the complex environment. The recognition accuracy. This solves the problem of the need to shoot and prepare a large number of learning image data when mechanical learning, even if only a small amount of captured image data, can effectively improve the accuracy of commodity recognition in a complex environment where multiple commodities are mixed.
At the 'RetailTech JAPAN2018' Circulation Technology Exhibition held at the Tokyo International Exhibition Center in March 2018, NEC displayed a POS payment settlement demonstration system using this technology. At the same time, it plans to use this technology at NEC headquarters from March. 3. The verification experiment of unmanned settlement of the settlement business in the convenience store.
The NEC Group is committed to promoting the 'social solutions business' globally and creating social values that are safe, secure, efficient, and fair. Integrating advanced ICT technologies and knowledge to achieve a brighter, more colorful and efficient society Do one strength.
YORK New Media Coolest Technology News
Scan code win award
comment
* Netizens are not allowed to speak in the standing position, this site is not recommended in the comments section of any online store, dealer, Beware of being deceived!
