Object category recognition method and device

Abstract

The present invention proposes a method and device for class recognition of an object, wherein the method includes: determining a partial image showing the operated object according to the first object image collected before the object is operated and the second object image collected after the object is operated, According to the local map, the multi-frame continuous images collected during the operation of the object are tracked, the tracking area corresponding to the multi-frame continuous image is determined, and the object type is identified for the tracking area. In the embodiment of the present invention, the operation object is collected before and after Combined with the multi-frame continuous images collected during the operation of the object, the category of the object is determined comprehensively, which improves the accuracy of object category recognition and solves the problem of identifying objects in the prior art through two static pictures collected before and after the operation of the object. Category recognition, when the object being manipulated is occluded or placed in a corner, it leads to a technical problem of low recognition accuracy.

Description

technical field [0001] The invention relates to the technical field of computer vision, in particular to a method and device for class recognition of objects. Background technique [0002] With the advancement of computer vision technology, computer vision recognition has been widely used in various fields, for example, in unmanned retail scenarios, such as unmanned supermarkets or unmanned containers, etc., when the goods on the shelves change Finally, if the customer takes the product from the shelf or puts it down, it is necessary to accurately identify which product the changed product on the shelf belongs to through visual technology. [0003] In the prior art, the usual practice is to compare the static pictures before and after the customer's operation to judge the product operated by the customer, but sometimes the products in the aisle are superimposed on each other, or the product is in a corner position, which will cause the product to be under the customer's oper...

AI Technical Summary

Problems solved by technology

[0003] In the prior art, the usual practice is to compare the static pictures before and after the customer's operation to judge the product operated by the customer, but sometimes the products in the aisle are superimposed on each other, or the product is in a corner, which will cause the product to be in the customer's operation. The before and after static pictures are blocked by others or the features are not obvious. Only relying on the two static pictures before and after the customer's operation cannot accurately judge the condition of the product operated by the customer, resulting in low accuracy of product category recognition

Images