Coin identification method, device, and cash register

Patent metadata

Field	Value
Publication number	US-11354887-B2
Application number	US-202016825222-A
Country	US
Kind code	B2
Filing date	Mar 20, 2020
Priority date	Sep 20, 2019
Publication date	Jun 7, 2022
Grant date	Jun 7, 2022

Abstract

A coin identification method and device, as well as a coin register are provided, which are related to a field of image identification. A specific implementation includes: extracting, from a to-be-identified image including coins, image blocks corresponding to the coins, to form an input coin set; inputting the input coin set into a coin classification model, to obtain categories and a classification confidences of the image blocks of the coins in the input coin set, wherein the category includes a face value of the coin; incorporating an image block, with a classification confidence reaching a threshold, of a coin of the input coin set into a first coin set; and determining a face value indicated by a category of the image block of the coin in the first coin set as a final face value of the coin in the first coin set.

First claim

What is claimed is:

1. A coin identification method, comprising:
extracting, from a to-be-identified image comprising coins, image blocks corresponding to the coins, to form an input coin set;
inputting the input coin set into a coin classification model, to obtain categories and a classification confidences of the image blocks of the coins in the input coin set, wherein a category comprises a face value of a coin;
incorporating an image block, with a classification confidence reaching a threshold, of the coin of the input coin set into a first coin set; and
determining a face value indicated by the category of the image block of the coin in the first coin set as a final face value of the coin in the first coin set.

2. The coin identification method according to claim 1 , further comprising:
incorporating an image block, except the image block of the coin of the first coin set, of the coin of the input coin set into a second coin set;
acquiring an appearance feature for each face value of coins in the first coin set; and
for each coin to be identified in the second coin set, comparing the image block of the coin in the second coin set with the appearance feature for each face value of coins in the first coin set, to determine a final face value of the coin in the second coin set.

3. The coin identification method according to claim 2 , wherein the category further comprises an upward side;
the incorporating an image block, with a classification confidence reaching a threshold, of a coin of the input coin set into a first coin set comprises:
incorporating an image block of a coin with a first upward side of the input coin set into the first coin set, wherein the image block of the first coin set has the classification confidence reaching the threshold.

4. The coin identification method according to claim 3 , wherein the appearance feature comprise information on a diameter and/or a color, and the for each coin to be identified in the second coin set, comparing the image block of the coin in the second coin set with the appearance feature for each face value of coins in the first coin set, to determine a final face value of the coin in the second coin set comprises at least one of:
for an image block of each coin to be identified with the first upward side in the second coin set, determining a face value of an image block of a coin, having a diameter difference with the image block of the coin to be identified less than a first threshold, in the first coin set; and/or determining a face value of an image block of a coin, having a color difference with the image block of the coin to be identified less than a second threshold; and determining the face value as the final face value of the coin to be identified; and
for an image block of each coin to be identified having a second upward side in the second coin set, determining a first face value of an image block having a diameter difference with the image block of the coin to be identified less than a third threshold, in the first coin set; and a second face value of an image block having a color difference with the image block of the coin to be identified less than a fourth threshold, in the first coin set; and if the first face value is equal to the second face value, determining the first face value as the final face value of the coin to be identified.

5. The coin identification method according to claim 1 , wherein the extracting, from a to-be-identified image comprising coins, image blocks corresponding to the coins, to form an input coin set comprises:
extracting n image blocks from the to-be-identified image comprising n coins, to form the input coin set, wherein one image block corresponds to one coin in the to-be-identified image, and n is a positive integer.

6. The coin identification method according to claim 5 , further comprising: training a coin detection model by using a target detection model,
the extracting n image blocks from the to-be-identified image comprising n coins comprises:
inputting the to-be-identified image into the coin detection model, and detecting positions of multiple detection frames and a confidence for each detection frame comprising a coin;
according to the positions of multiple detection frames and the confidence for each detection frame comprising a coin, acquiring positions of n detection frames, each of which is determined to comprise a coin; and
extracting n image blocks from the to-be-identified image according to the positions of the n detection frames.

7. The coin identification method according to claim 6 , further comprising:
performing ellipse fitting on a detection frame for each coin, to acquire an ellipse detection frame; and
calculating a diameter of the ellipse detection frame as information on a diameter of the coin, and/or a color average in the ellipse detection frame as information on a color of the coin.

8. The coin identification method according to claim 2 , wherein the acquiring an appearance feature for each face value of coins in the first coin set comprises:
calculating a diameter average and/or a color average of one or more coins for each face value in the first coin set as information on a diameter and/or a color corresponding to the face value.

9. The coin identification method according to claim 1 , further comprising:
according to the final face value of each of the coins in the input coin set, determining a total face value and/or a number of all coins in the input coin set by statistics.

10. The coin identification method according to claim 1 , further comprising:
outputting information of the coins in an image form;
wherein the information of the coins comprises at least one of a position of each coin, a face value of each coin, a classification confidence of each coin, a total face value of the coins, and a number of the coins.

11. A coin identification device, comprising:
one or more processors; and
a storage device configured for storing one or more programs, wherein
the one or more programs are executed by the one or more processors to enable the one or more processors to:
extract, from a to-be-identified image comprising coins, image blocks corresponding to the coins, to form an input coin set;
input the input coin set into a coin classification model, to obtain categories and a classification confidences of the image blocks of the coins in the input coin set, wherein a category comprises a face value of a coin; and
incorporate an image block, with a classification confidence reaching a threshold, of the coin of the input coin set into a first coin set; and determine a face value indicated by the category of the image block of the coin in the first coin set as a final face value of the coin in the first coin set.

12. The coin identification device according to claim 11 , wherein the one or more programs are executed by the one or more processors to enable the one or more processors further to:
acquire an appearance feature for each face value of coins in the first coin set; and
for each coin to be identified in the second coin set, compare the image block of the coin in the second coin set with the appearance feature for each face value of coins in the first coin set, to determine a final face value of the coin in the second coin set.

13. The coin identification device according to claim 12 , wherein the category further comprises an upward side; and the one or more programs are executed by the one or more processors to enable the one or more processors further to:
incorporate an image block of a coin with a first upward side of the input coin set into the first coin set, wherein the image blo

Assignees

• Beijing Baidu Netcom Sci & Tech Co Ltd

Inventors

• Zhang Jiadong
• Liu Guoyi
• Li Chao

Classifications

• G06V20/66

  Trinkets, e.g. shirt buttons or jewellery items (recognising microscopic objects G06V20/69) · CPC title

• G06V20/00

  Scenes; Scene-specific elements (control of digital cameras H04N23/60) · CPC title

• G06V10/82

  using neural networks · CPC title

• G06V10/764

  using classification, e.g. of video objects · CPC title

• G06N20/00 Primary

  Machine learning · CPC title