Self-selection restaurant automatic charging method

Abstract

The invention provides a self-selection restaurant automatic charging method based on computer vision technology. The system comprises a dinner plate sensing module, a dish division module and a dish identification module functionally. The dinner plate sensing module adopts an infrared sensing and static detection cascaded method, thereby ensuring high definition of pictures while realizing fast acquisition; the dish division module adopts an area generation network, thereby ensuring interference resistance for sheltering situations while realizing accurate positioning and segmentation; and the dish identification module adopts a depth twin measurement network, thereby overcoming dependence of a conventional deep learning method on the number of training samples. The system can directly identify dishes based on the images, so that when the system is applied to the restaurant, bowls and dishes do not need to be replaced, and the system is convenient to deploy and low in cost. The method can collect detailed and specific pictures and consumption information and the like, can provide sale analysis for the restaurant and diet health analysis for clients, and is an important link for realizing smart restaurant based on big data.

Description

technical field [0001] The invention is an automatic pricing method for self-selected restaurants, in particular relates to intelligent dish segmentation and identification, and belongs to the technical fields of computer vision technology and smart restaurants. Background technique [0002] Self-selected fast food is a fast food mode developed and extended on the basis of Chinese fast food. The self-service restaurant will neatly list batches of prepared meals, and customers will line up with trays to choose their favorite dishes, and they will be settled and served immediately. This model is convenient and fast, with a high degree of freedom. Its application scope covers almost all organizational units including industrial parks, enterprise factories, state agencies, schools, hospitals, and the military. The market scale continues to expand. However, the current self-selected restaurants still rely on manual pricing and settlement. On the one hand, cashiers need to enter...

AI Technical Summary

Problems solved by technology

The advantage of this type of method is that it is stable and reliable, but its limitation is that: on the one hand, the restaurant must replace differentiated dishes; employee workload

The advantage of the method proposed in this patent is that it does not need to replace dishes and reduce system cost. However, there are the following problems in practical applications: First, the Hough transform algorithm is used in this patent to detect the plate, which requires high integrity of the outline of the plate. , unable to cope with the situation that some dishes are easily covered in practical applications (such as stacking plates, hand occlusion); second, the patent uses a classification model based on convolutional neural network for dish recognition, and the parameters of this type of model are complex , the data volume requirements are very large. To identify a dish, hundreds or even thousands of dishes must be collected to achieve sufficient accuracy. Once a restaurant launches a new dish, it needs to be re-collected, retrained and updated. It brings inconvenience and difficulty to the practical application of the restaurant

Images