US-12620196-B2 - Method for detecting an object in a search image, method for generating a pattern vector, and use of a method for determining a position and/or orientation of a security element of a banknote

US12620196B2US 12620196 B2US12620196 B2US 12620196B2US-12620196-B2

Abstract

A method for detecting an object in a search image, includes: a) providing a pattern vector which describes the object by means of coordinates of characteristic pixels; b) shifting the pattern vector over different positions of the search image; c) determining a success value at each of the different positions; and d) detecting the object at the position on the basis of the success value, wherein each characteristic pixel is assigned a first direction and a second direction that differs from the first direction. The first and second overall intensity values of one-dimensionally arranged pixels are determined in the first and second directions, respectively. A difference value between the first and second overall intensity values is determined in each case, and the success value is determined on the basis of the respective difference values.

Inventors

Shanchuan Su

Assignees

GIESECKE+DEVRIENT CURRENCY TECHNOLOGY GMBH

Dates

Publication Date: 20260505
Application Date: 20220302
Priority Date: 20210309

Claims (11)

1 . A method for detecting an object in a search image, having the following steps: a) providing a pattern vector, which describes the object by way of coordinates of characteristic pixels; b) displacing the pattern vector over various positions of the search image; c) defining a respective success value at the various positions; and d) detecting the object at the position on the basis of the success value, wherein a first direction and a second direction different from the first direction is assigned to each characteristic pixel by the pattern vector, wherein in each case a first intensity overall value of a first number of one-dimensionally arranged pixels in the first direction is determined and a second intensity overall value of a second number of one-dimensionally arranged pixels in the second direction is determined, wherein in each case a difference value is determined between the first intensity overall value and the second intensity overall value, wherein the success value is determined on the basis of the respective difference values, wherein the object is detected on a banknote.
2 . The method according to claim 1 , wherein the maximum difference value of the difference values leads to the success value, at which the object is detected.
3 . The method according to claim 1 , wherein the first direction and the second direction are assigned in opposition by 180°.
4 . The method according to claim 1 , wherein the first number and/or the second number is defined as greater than one.
5 . The method according to claim 1 , wherein the first number and/or the second number is defined as less than ten.
6 . The method according to claim 1 , wherein the first number and/or the second number is defined differently.
7 . The method according to claim 1 , wherein the pattern vector has at most four values for each characteristic pixel.
8 . The method according to claim 1 , wherein a security element of a banknote is detected as the object.
9 . A method for generating a pattern vector, in which the following steps are carried out: a) defining a contour of an object in a pattern image, wherein the object is detected on a banknote; b) defining a gradient direction for pixels of the contour; c) defining a gradient amount for pixels of the contour, d) defining pixels of the contour as characteristic pixels if the respective gradient amount is greater than an amount threshold value; e) defining a respective position of the characteristic pixels; f) defining a first direction and a second direction in dependence on the gradient direction of the characteristic pixels; and g) generating the pattern vector using the respective position of the characteristic pixels in the pattern image and the first direction and the second direction.
10 . The method according to claim 9 , wherein the pattern vector is generated using the first direction and/or the second direction relative to the associated characteristic pixel.
11 . The use of a method according to claim 1 for determining a position and/or orientation of a security element of a banknote relative to a substrate of the banknote.

Description

BACKGROUND The invention relates to a method for detecting an object in a search image. A pattern vector is provided, which describes the object by way of coordinates of characteristic pixels. The pattern vector is displaced over various positions of the search image. Furthermore, a respective success value is determined at the various positions and the object is detected at the position on the basis of the success value. The invention also relates to a method for generating a corresponding pattern vector. Moreover, the invention relates to a use of a corresponding method for determining a position and/or orientation of a security element of a banknote. Methods for detecting an object are known. This field is thus also generally known under the term pattern recognition or object recognition in digital images. One known method is the Viola-Jones method. Using the Viola-Jones method, noteworthy objects such as faces can be detected in a digital image. The core of the Viola-Jones method are Viola-Jones base patterns, which are partially based on so-called hair wavelets. Hair wavelets are single wavelets which are formed from the combination of two rectangular functions. It is intended in particular hereinafter that objects be detected under changing illumination conditions. One challenge in this case is that the same object is thus depicted once light against a dark background and another time dark against a light background. Known methods have proven to be slow or unreliable again and again under these conditions. SUMMARY The object is to detect an object which can be depicted in a digital image with different brightnesses depending on the illumination faster and more reliably. This object is achieved according to the invention by a respective method having the features according to the independent claims. Advantageous embodiments of the invention are the subject matter of the dependent claims. In a method according to the invention, an object, in particular an object of a banknote, is detected in a search image. The following steps are carried out: a) providing a pattern vector or a pattern matrix, which describes the object by way of coordinates of characteristic pixels;b) displacing the pattern vector over various positions of the search image;c) defining a respective success value at the various positions; andd) detecting the object at the position on the basis of the success value, in particular the maximum success value. One important concept of the invention is that, in particular by the pattern vector, each characteristic pixel is assigned a first direction and a second direction different from the first direction, wherein in each case a first intensity overall value of a first number of one-dimensionally arranged pixels in the first direction is determined and a second intensity overall value of a second number of one-dimensionally arranged pixels in the second direction is determined, wherein in each case a difference value, in particular and/or a ratio value, is determined between the first intensity overall value and the second intensity overall value, wherein the success value is determined on the basis of the respective difference values, in particular and/or the respective ratio values. The invention is based on the finding that sufficient features in order to detect the object or the position and/or orientation in the search image are provided solely by the coordinates, thus the position of the characteristic pixels and the directions to form the difference value. During the generation of the pattern vector or the classifier, in particular pixels on edges or at points having a high frequency, i.e., a transmission from light to dark or dark to light, are selected as characteristic pixels. These characteristic pixels or the pattern, which is two-dimensional in particular, of the characteristic pixels, thus the point cloud spanned by the coordinates of the characteristic pixels, are or is then found again in the search image. For this purpose, it is not necessary to specify a specific value of the differences of the pixels of the different directions per characteristic pixel. The respective position of the characteristic pixels and the respective associated direction are sufficient. The number of the pixels of the respective direction which are to be used for the difference value can either likewise be provided by the pattern vector or also specified as a general standard setting. In particular, it is provided that the first intensity overall value and/or the second intensity overall value is formed by a mean value, for example an arithmetic mean value or a median value. The success value can, for example, be formed directly from the difference values, for example by summing and optional norming. The success value can also be generated in dependence on the weighting function, however, in which the difference values are incorporated in weighted form. It is preferably provided that the maximum diff