CN108256566A - A kind of adaptive masterplate matching process and device based on cosine similarity - Google Patents

Abstract

The present application discloses an adaptive template matching method based on cosine similarity, which includes: matching and storing the similarity value of the image to be recognized from the matching starting point and the template image of the matching target according to the calculation; according to the increase of the similarity value Decrease the trend to update the translation distance with the reverse trend; update the matching starting point according to the translation distance and preset translation direction; judge whether the coordinates of the updated matching starting point meet the preset termination conditions; if not, continue to calculate the image to be recognized from the matching The similarity value matched with the template image of the matching target at the starting point is stored; if so, the maximum value of all similarity values is determined, and then the matching target in the image to be recognized is determined. The present application adaptively adjusts the translation distance according to the cosine similarity, which can improve matching accuracy and speed at the same time. The present application also discloses an adaptive template matching device, equipment and computer-readable storage medium based on cosine similarity, which also have the above beneficial effects.

Description

A method and device for adaptive template matching based on cosine similarity

technical field

The present application relates to the technical field of image recognition, in particular to an adaptive template matching method, device, device and computer-readable storage device based on cosine similarity.

Background technique

In the field of image recognition technology, template matching is the most basic and primitive pattern recognition method. It determines the matching target in the image to be recognized by selecting a local area in the image to be recognized that has the same shape and size as the template image for matching and comparison with the template image, and then judging the calculated dissimilarity or similarity.

In the process of template matching, matching speed and matching accuracy are two important indicators to measure the superiority of the algorithm. However, the template matching methods in the prior art either have low matching speed or low matching accuracy. This is because, in the prior art, there is no scientific and reasonable control strategy to adjust the translation speed of the local area in the image to be recognized. The translation speed mentioned here actually refers to the translation distance of each translation. If the translation distance is too small, the matching speed will be reduced. If the translation distance is too large, in some areas close to the matching target, the matching target will be missed.

It can be seen that which template matching method to use to effectively improve both matching accuracy and matching speed is an important technical problem to be solved by those skilled in the art.

Contents of the invention

The purpose of the present application is to provide an adaptive template matching method, device, equipment and computer-readable storage medium based on cosine similarity, so as to effectively improve matching accuracy and matching speed at the same time.

In order to solve the above technical problems, the present application provides an adaptive template matching method based on cosine similarity, including:

According to the preset cosine similarity function, calculate the similarity value of matching the template image of the image to be recognized from the matching starting point and the matching target and store it; the cosine similarity function is:

or

Wherein, the pixel size of the template image is a×b; p(i, j) is the gray value of the template image at point coordinates (i, j); a, b, i and j are all positive integers , and 1≤i≤a, 1≤j≤b; (x, y) is the coordinate of the matching starting point of the image to be recognized; q(x+i, y+j) is the image to be recognized Gray value at point coordinates (x+i, y+j); f(x, y) is the similarity value corresponding to the matching starting point (x, y);

Updating the translation distance in a reverse trend according to the increase or decrease trend of the similarity value corresponding to the matching starting point;

updating the coordinates of the matching starting point according to the updated translation distance and preset translation direction;

judging whether the updated coordinates of the matching start point satisfy a preset termination condition;

If not, then continue to perform the subsequent steps of calculating the similarity value of the image to be recognized from the matching starting point and matching the template image of the matching target and storing;

If yes, determine the maximum value among all the similarity values, and determine the matching target in the image to be recognized according to a preset recognition basis.

Optionally, the updating the translation distance in a reverse trend according to the increase or decrease trend of the similarity value corresponding to the matching starting point includes:

According to the formula Update the translation distance; where, h is the updated translation distance, h' is the translation distance before update, h and h' are both positive integers; c1 is the speed-up adjustment parameter, 1<c1 ;c2 is the deceleration adjustment parameter, 0<c2<1; Preset threshold; [] indicates the integer part.

Optionally, the preset threshold for:

Wherein, β is the mean value of similarity values obtained by matching the sample image containing the matching target with the template image; l is a threshold coefficient.

Optionally, the preset translation direction is column translation.

Optionally, the translation distance satisfies the condition:

1≤h≤a.

Optionally, the determining the matching target in the image to be recognized according to the preset recognition basis includes:

judging whether the distance between the matching start points corresponding to two adjacent maximum values is lower than a preset distance threshold;

If so, it is determined that there is one matching target in the image to be recognized at the matching starting point corresponding to the larger value among the two adjacent maximum values;

If not, it is determined that there is one matching target in the image to be recognized at the matching starting points respectively corresponding to the two adjacent maximum values.

Optionally, the preset distance threshold includes a preset lateral distance threshold and a preset longitudinal distance threshold;

The judging whether the distance between the matching starting points corresponding to two adjacent maximum values is lower than the preset distance threshold includes:

Judging whether the lateral distance between the matching starting points corresponding to the two adjacent maximum values is lower than the preset lateral distance threshold, or the distance between the matching starting points corresponding to the two adjacent maximum values The longitudinal distance is lower than the preset longitudinal distance threshold.

The present application also provides an adaptive template matching device based on cosine similarity, including:

Calculation module: used to calculate and store the similarity value of the image to be recognized from the matching starting point and the template image of the matching target according to the preset cosine similarity function; the cosine similarity function is:

or

Wherein, the pixel size of the template image is a×b; p(i, j) is the gray value of the template image at point coordinates (i, j); a, b, i and j are all positive integers , and 1≤i≤a, 1≤j≤b; (x, y) is the coordinate of the matching starting point of the image to be recognized; q(x+i, y+j) is the image to be recognized Gray value at point coordinates (x+i, y+j); f(x, y) is the similarity value corresponding to the matching starting point (x, y);

An update module: used to update the translation distance in a reverse trend according to the increase and decrease trend of the similarity value corresponding to the matching starting point; and start the matching according to the updated translation distance and preset translation direction The coordinates of the starting point are updated;

Judging module: used to judge whether the updated coordinates of the matching starting point meet the preset termination conditions; so that the calculation module continues when the updated coordinates of the matching starting point do not meet the preset termination conditions It is used to calculate and store the similarity value of matching the image to be recognized with the template image of the matching target from the matching starting point;

Identification module: used to determine the maximum value of all the similarity values when the coordinates of the matching starting point satisfy the preset termination condition, and determine the matching in the image to be identified according to a preset identification basis Target.

The present application also provides an adaptive template matching device based on cosine similarity, including:

memory: used to store computer instructions;

Processor: used to execute the computer instructions to implement the steps of any cosine similarity-based adaptive template matching method as described above.

The present application also provides a computer-readable storage medium, where a computer program is stored in the computer-readable storage medium, and when the computer program is executed by a processor, any one of the above-mentioned cosine similarity-based adaptive The steps of the template matching method.

The adaptive template matching method based on cosine similarity provided by this application includes:

According to the preset cosine similarity function, calculate the similarity value of matching the template image of the image to be recognized from the matching starting point and the matching target and store it; the cosine similarity function is:

or

Wherein, the pixel size of the template image is a×b; p(i, j) is the gray value of the template image at point coordinates (i, j); a, b, i and j are all positive integers , and 1≤i≤a, 1≤j≤b; (x, y) is the coordinate of the matching starting point of the image to be recognized; q(x+i, y+j) is the image to be recognized Gray value at point coordinates (x+i, y+j); f(x, y) is the similarity value corresponding to the matching starting point (x, y);

Update the size of the translation distance with a reverse trend according to the increase or decrease trend of the similarity value corresponding to the matching starting point; perform the coordinates of the matching starting point according to the updated translation distance and preset translation direction Update; judging whether the updated coordinates of the matching starting point meet the preset termination condition; if not, continue to perform the calculation of the similarity value of the image to be recognized from the matching starting point and the template image of the matching target and The subsequent step of storing; if yes, determining the maximum value among all the similarity values, and determining the matching target in the image to be recognized according to a preset recognition basis.

It can be seen that compared with the prior art, in the adaptive template matching method based on cosine similarity provided by the present application, the cosine similarity commonly used in statistics is used to judge the matching similarity between the local area of the image to be recognized and the template image. degree, and then according to the matching degree, that is, the size of the similarity value, the translation distance is adjusted against the trend when the matching starting point of the local area is translated, so that the translation distance can be reduced in the area with a high similarity value, while in the similarity value Low areas increase the translation distance. It can be seen that the present application can effectively improve matching accuracy and matching speed at the same time. The cosine similarity-based adaptive template matching device, equipment, and computer-readable storage medium provided in the present application can realize the above-mentioned cosine similarity-based adaptive template matching method, and also have the above-mentioned beneficial effects.

Description of drawings

In order to illustrate the prior art and the technical solutions in the embodiments of the present application more clearly, the following briefly introduces the drawings that need to be used in the description of the prior art and the embodiments of the present application. Of course, the following drawings related to the embodiments of the application describe only a part of the embodiments of the application, and those of ordinary skill in the art can obtain other The accompanying drawings, and other obtained drawings also belong to the protection scope of the present application.

Fig. 1 is a flow chart of an adaptive template matching method based on cosine similarity provided by the embodiment of the present application;

Fig. 2 is a schematic diagram of the row direction moving direction provided by the embodiment of the present application;

FIG. 3 is a schematic diagram of a row moving direction provided by an embodiment of the present application;

FIG. 4 is a flow chart of another cosine similarity-based adaptive template matching method provided by an embodiment of the present application;

Fig. 5 is a structural block diagram of an adaptive template matching device based on cosine similarity provided by an embodiment of the present application.

Detailed ways

The core of the present application is to provide an adaptive template matching method, device, device and computer-readable storage device based on cosine similarity, so as to effectively improve matching accuracy and matching speed at the same time.

In order to describe the technical solutions in the embodiments of the present application more clearly and completely, the technical solutions in the embodiments of the present application will be introduced below in conjunction with the drawings in the embodiments of the present application. Apparently, the described embodiments are only some of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the scope of protection of this application.

Please refer to FIG. 1. FIG. 1 is a flow chart of an adaptive template matching method based on cosine similarity provided by the embodiment of the present application, which mainly includes the following steps:

Step 11: According to the preset cosine similarity function, calculate and store the similarity value of matching the image to be recognized with the template image of the matching target from the matching starting point; the cosine similarity function is:

or

Among them, the pixel size of the template image is a×b; p(i, j) is the gray value of the template image at point coordinates (i, j); a, b, i and j are all positive integers, and 1≤ i≤a, 1≤j≤b; (x, y) is the coordinate of the matching starting point of the image to be recognized; q(x+i, y+j) is the point coordinate of the image to be recognized at (x+i, y+ The gray value at j); f(x, y) is the similarity value corresponding to the matching starting point (x, y).

Specifically, when the template matching method provided by this application is in the process of matching and comparing the image to be recognized and the template image, the similarity used is specifically the similarity calculated based on cosine, which is especially suitable for gray value angles Image recognition occasions with significant characteristics. When the similarity between the gray value of each pixel in the local area selected from the image to be recognized and the gray value of the corresponding pixel in the template image is calculated, the higher the possibility of matching the target in the local area .

It can also be seen from the calculation expression of the cosine similarity function that when calculating the similarity, it starts from the matching starting point and traverses the selected local area in units of 1 pixel. Cosine similarity calculation is performed between the gray values of all pixels and the gray values of the pixels of the template image. When performing template matching, in order to sequentially compare the local area selected from the image to be recognized with each pixel of the template image, it is necessary to set the origin and a unified coordinate axis direction for the image to be recognized and the template image respectively. Generally, the upper left vertices of the image to be recognized and the template image can be set as their respective origins, and then the horizontal to the right and vertical downward can be set to the abscissa direction and the ordinate direction respectively. Since the local area covers multiple pixel points, the matching starting point can generally be used to measure the position of the local area. Similarly, generally, the pixel point of the upper left vertex of the local area is used as the matching start point.

Step 12: Update the translation distance in a reverse trend according to the increase and decrease trend of the similarity value corresponding to the matching starting point.

Specifically, when the template matching method provided in the present application translates the matching starting point, that is, the local area, it adapts the anti-trend adjustment according to the matching result corresponding to the current matching starting point, that is, the similarity value. When the similarity value corresponding to the current matching starting point is large, it means that the selected local area is relatively close to the template image, and there is a high possibility of matching targets. The translation distance is used to prevent missing local areas with higher similarity, which can effectively improve the matching accuracy. When the similarity value corresponding to the current matching starting point is small, it means that the local area selected at this time is very different from the template image, and the possibility of saving the matching target is low. The translation distance in order to improve the matching speed.

Step 13: Update the coordinates of the matching starting point according to the updated translation distance and preset translation direction.

Specifically, the matching starting point is translated according to the translation distance and the preset translation direction. When the origin is selected as the upper left vertex, the preset translation direction can be either row-by-row right translation or column-by-column translation down, and the exact number of pixels to be translated depends on the updated translation distance.

Step 14: Determine whether the updated coordinates of the matching starting point meet the preset termination condition: if not, go to step 11; if yes, go to step 15.

Specifically, as mentioned above, during the execution of the template matching algorithm, the translation of the matching starting point, that is, the local area, is carried out in a planned and orderly manner according to the preset translation direction. The purpose of the translation is to make all occurrences The overlaid local areas can be superimposed to cover the image to be recognized as much as possible. After the preset translation direction is determined, the termination condition of the translation is also determined. The termination condition may be specifically related to the coordinates of the matching starting point, because the coordinates of the matching starting point represent the range of the corresponding local area. For example, when the pixel size of the template image is a×b, and the coordinates of the matching starting point of the image to be recognized are (x, y), the local area corresponding to the matching starting point is the pixel point (x, y) as The upper left vertex, the area whose vertical length is a and horizontal length is b.

For example, if the pixel size of the image to be recognized is A×B, in order to prevent the local area from exceeding the range of the image to be recognized, the matching starting point can be set within the valid range of (A-a)×(B-b) in the image to be recognized according to the preset Pan direction to pan. Specifically, the matching start point can be shifted column by column from the origin, or shifted row by row. Column-to-column translation means to translate the matching starting point in the column from top to bottom (with the upper left vertex as the origin), and when it reaches the bottom of the column, switch to the top of the next column and continue to translate downward, as shown in Figure 2 Shown; Row-to-row translation means that the matching starting point is translated from left to right (with the upper left vertex as the origin) in the row, and when it is translated to the right end of the column, it is switched to the next row and continues to translate to the right, as shown in Figure 3 shown. When the matching start point is translated within the valid range of (A-a)×(B-b), the coordinates after it is translated again must fall outside the valid range. within the effective range as the default termination condition.

Step 15: Determine the maximum value among all similarity values, and determine the matching target in the image to be recognized according to the preset recognition basis.

Specifically, after the translation of the matching starting point within the effective range is completed, the maximum value can be determined according to all the similarity values obtained during the translation process, so as to determine Matching objects in the image to be recognized.

It can be seen that in the adaptive template matching method based on cosine similarity provided in the embodiment of the present application, the cosine similarity commonly used in statistics is used to judge the matching similarity between the local area of the image to be recognized and the template image, and then according to the The degree of matching, that is, the size of the similarity value, comes from the anti-trend adjustment to adjust the translation distance when the matching starting point of the local area is translated, so that the translation distance can be reduced in areas with high similarity values and increased in areas with low similarity values. translation distance. It can be seen that the present application can effectively improve matching accuracy and matching speed at the same time.

The adaptive template matching method based on cosine similarity provided by this application, on the basis of the above-mentioned embodiments:

As a preferred embodiment, updating the translation distance according to the similarity value corresponding to the matching starting point includes:

According to the formula Update the translation distance.

Among them, h is the updated translation distance, h' is the translation distance before update, both h and h' are positive integers; c1 is the acceleration adjustment parameter, 1<c1; c2 is the deceleration adjustment parameter, 0<c2<1 ; Preset threshold; [] indicates the integer part.

Specifically, when the translation distance is adaptively adjusted against the trend according to the similarity value, the translation distance can be adjusted by using the acceleration adjustment parameter c1 and the deceleration adjustment parameter c2 respectively. As for the specific parameter values, those skilled in the art can choose and set them by themselves.

In addition, in the embodiment of the present application, when updating the translation distance, the judgment of whether the similarity value f(x, y) is large or not adopts two kinds of judgment criteria in combination: namely, with the preset threshold value Positive and negative comparisons for comparison and rate of change. When the similarity value is greater than the preset threshold Or when it is in the increasing state, it means that the possibility of matching the template image is very high at this time, or the possibility of matching the previous matching result is higher, so the translation distance can be reduced to improve the matching accuracy; and when The similarity value is less than the preset threshold Or when it is in a decreasing state, it means that the possibility of matching with the template image at this time is very low, or the possibility is lower compared with the last matching result, so the translation distance can be adjusted to increase to improve the matching speed.

As a preferred embodiment, the preset threshold for:

Among them, β is the mean value of the similarity value obtained by matching the sample image containing the matching target with the template image; l is the threshold coefficient.

Specifically, the preset threshold It can be selected in combination with the general average level of the matching results between the sample image and the template image. Of course, the said sample image is a sample image containing the matching target, and generally should be randomly selected.

As a preferred embodiment, the preset translation direction is column translation.

As mentioned above, according to the computer's habit of processing data column by column, the matching starting point can be shifted column-wise within the effective range, so that each corresponding local area covers the image to be recognized. Certainly, those skilled in the art may also adopt other manners, which are not limited in this embodiment of the present application.

As a preferred embodiment, the translation distance satisfies the condition:

1≤h≤a.

When the column-wise translation method is used, the translation distance h after each update can be limited in size, so as to prevent the translation distance h from being too large and skipping some regions containing the matching target, which affects the matching accuracy. Specifically, the translation distance h can be set to be not greater than the vertical length a of the template image, that is, 1≤h≤a. When ensuring the size of the translation distance h, a saturation function can be used specifically, that is:

Wherein, h _bef is the translation distance h before saturation processing, and h _aft is the translation distance h after saturation processing.

As a preferred embodiment, determining the matching target in the image to be recognized according to the preset recognition basis includes:

Judging whether the distance between matching starting points corresponding to two adjacent maximum values is lower than a preset distance threshold;

If so, it is determined that there is a matching target at the matching starting point corresponding to the larger value of the adjacent two maximum values in the image to be recognized;

If not, it is determined that there is a matching target in the image to be recognized at the matching start points corresponding to the two adjacent maximum values respectively.

Specifically, in the adaptive template matching method based on cosine similarity provided in the embodiment of the present application, when determining the matching target according to each similarity value, the determination is specifically based on the maximum value. According to the physical meaning of the maximum value, the image to be recognized may have a matching target at the matching starting point (local area) corresponding to a certain maximum value; it may also be a matching target corresponding to multiple maximum values. There is a matching target in the area between the starting points.

In order to distinguish this, the embodiment of the present application judges the distance between the matching starting points corresponding to two adjacent maximum values. If the distance is less than the preset distance threshold, it means that the two matching starting points are far apart. Very close, it should belong to the second situation above. The specific position of the matching target described by these two maximum values can be determined by the larger value of the two adjacent maximum values, that is, there is a matching target at the matching starting point corresponding to the larger value . Assuming that the coordinates of the matching starting point corresponding to the larger value are (x1, y1), the position coordinates of the center of the matching target can be regarded as (x1+a/2, y1+b/2).

If the distance between the matching starting points corresponding to the two adjacent maximum values is not less than the preset distance threshold, it means that the two matching starting points are far apart, which should belong to the first case above. Therefore, it can be determined that there is a matching target at the matching start points corresponding to the two maximum values in the image to be recognized.

In addition, the specific value of the preset distance threshold mentioned here can be selected and set by those skilled in the art, which is not limited in the embodiments of the present application.

As a preferred embodiment, the preset distance threshold includes a preset horizontal distance threshold and a preset longitudinal distance threshold;

Judging whether the distance between the matching start points corresponding to two adjacent maximum values is lower than the preset distance threshold includes:

Judging whether the horizontal distance between the matching start points corresponding to two adjacent maximum values is lower than the preset horizontal distance threshold, or whether the vertical distance between the matching starting points corresponding to two adjacent maximum values is lower than the preset Set the vertical distance threshold.

Specifically, the preset distance thresholds adopted in the embodiment of the present application include two preset horizontal distance thresholds and preset vertical distance thresholds, which are respectively used to judge the horizontal distance and vertical distance corresponding to the matching starting point of two maximum values. And, specifically, the preset horizontal distance threshold and the preset vertical distance threshold may be respectively set as the vertical length a and the horizontal length b of the template image. Certainly, those skilled in the art may also select and set other thresholds by themselves, which is not limited in the embodiments of the present application.

Please refer to FIG. 4. FIG. 4 is a flow chart of another cosine similarity-based adaptive template matching method provided by the embodiment of the present application; including:

Step 21: According to the formula Calculate and store the similarity value of matching the image to be recognized with the template image of the matching target from the matching starting point.

Among them, the pixel size of the template image is a×b; p(i, j) is the gray value of the template image at point coordinates (i, j); a, b, i and j are all positive integers, and 1≤ i≤a, 1≤j≤b; (x, y) is the coordinate of the matching starting point of the image to be recognized; q(x+i, y+j) is the point coordinate of the image to be recognized at (x+i, y+ The gray value at j); f(x, y) is the similarity value corresponding to the matching starting point (x, y).

Specifically, the adaptive template matching method provided by the embodiment of the present application is based on the formula to calculate the similarity. And, the initial value of the matching starting point (x,y) is (1,1).

Step 22: According to the formula Update the translation distance.

Among them, h is the updated translation distance, h' is the translation distance before update, both h and h' are positive integers; c1 is the acceleration adjustment parameter, 1<c1; c2 is the deceleration adjustment parameter, 0<c2<1 ; Preset threshold, [] indicates the integer part.

Step 23: Reassign x+h to x, and keep y unchanged.

Specifically, the adaptive template matching method provided by the embodiment of the present application adopts the translation method of the starting point of the column-wise translation matching, and the translation distance h of each translation to the right is adjusted according to the similarity value f(x, y) and the speed-up parameter c1 or deceleration adjustment parameter c2 is determined.

Step 24: Determine whether x>A-a holds: if yes, go to step 25; if not, go to step 26.

Specifically, after the coordinate column of the matching starting point is shifted downward, it may first be judged whether the vertical coordinate of the matching starting point after the downward translation is beyond the valid range, that is, whether the column needs to be changed. When x>A-a is established, it is necessary to change columns. When x>A-a does not hold, there is no need to change columns, and it is directly judged whether the preset termination condition is satisfied.

Step 25: Reassign y+1 to y, and take x as 1; go to step 26.

Specifically, when changing columns, add 1 to the abscissa of the matching starting point, and take the ordinate as an initial value of 1. After changing columns, it can be judged whether the preset termination condition is satisfied.

Step 26: Determine whether y>B-b holds: if not, go to step 21; if yes, go to step 27.

When y>B-b holds true, it means that the abscissa of the matching starting point is beyond the valid range, and it means that the translation process within the valid range has been completed at this time, and step 27 can be entered to determine the matching target. Otherwise, continue to return to step 21 to calculate the similarity of the new matching starting point.

Step 27: Determine the maximum value among all similarity values; go to step 28.

Step 28: Determine whether the distance between the matching starting points corresponding to two adjacent maximum values is lower than the preset distance threshold: if so, determine the distance between the larger value of the two adjacent maximum values of the image to be recognized There is a matching target at the corresponding matching starting point; if not, it is determined that there is a matching target at the corresponding matching starting points of the two adjacent maximum values in the image to be recognized.

The cosine similarity-based adaptive template matching device provided by the embodiment of the present application is introduced below.

Please refer to FIG. 5, which is a structural block diagram of an adaptive template matching device based on cosine similarity provided by the present application; including a calculation module 1, an update module 2, a judgment module 3 and an identification module 4;

Calculation module 1 is used for calculating the similarity value of matching the template image of the image to be recognized from the matching starting point and matching target according to the preset cosine similarity function and storing; the cosine similarity function is:

or

Among them, the pixel size of the template image is a×b; p(i, j) is the gray value of the template image at point coordinates (i, j); a, b, i and j are all positive integers, and 1≤ i≤a, 1≤j≤b; (x, y) is the coordinate of the matching starting point of the image to be recognized; q(x+i, y+j) is the point coordinate of the image to be recognized at (x+i, y+ j) at the gray value; f(x, y) is the similarity value corresponding to the matching starting point (x, y);

The update module 2 is used to update the translation distance with a reverse trend according to the increase and decrease trend of the similarity value corresponding to the matching starting point; and update the coordinates of the matching starting point according to the updated translation distance and preset translation direction;

Judgment module 3 is used for judging whether the coordinates of the matching starting point after updating meet the preset termination condition; so that the calculation module continues to be used to calculate the image to be recognized from the matching when the coordinates of the updated matching starting point do not meet the preset termination condition The similarity value of the matching target template image at the starting point is stored;

The recognition module 4 is used to determine the maximum value among all similarity values when the coordinates of the matching starting point satisfy the preset termination condition, and determine the matching target in the image to be recognized according to the preset recognition basis.

It can be seen that the adaptive template matching device based on cosine similarity provided by the present application adopts the cosine similarity commonly used in statistics to judge the matching similarity between the local area of the image to be recognized and the template image, and then according to the matching degree that is The size of the similarity value comes from the anti-trend adjustment to adjust the translation distance when the matching starting point of the local area is translated, so that the translation distance is reduced in the area with high similarity value, and the translation distance is increased in the area with low similarity value. It can be seen that the present application can effectively improve matching accuracy and matching speed at the same time.

The present application also provides an adaptive template matching device based on cosine similarity, including:

memory: used to store computer instructions;

Processor: for executing the computer instructions to implement the steps of any one of the cosine similarity-based adaptive template matching methods described above.

The present application also provides a computer-readable storage medium, where a computer program is stored in the computer-readable storage medium, and when the computer program is executed by a processor, any one of the above-mentioned adaptive template matching based on cosine similarity can be realized method steps.

The specific implementations of the cosine similarity-based adaptive template matching device, equipment, and computer-readable storage medium provided in this application and the cosine similarity-based adaptive template matching method described above can be referred to each other. No longer.

Each embodiment in the present application is described in a progressive manner, each embodiment focuses on the differences from other embodiments, and the same and similar parts of the various embodiments can be referred to each other. As for the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and for the related information, please refer to the description of the method part.

It should also be noted that in this application, relative terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between such entities or operations. Furthermore, the term "comprises", "comprises" or any other variation thereof is intended to cover a non-exclusive inclusion such that a process, method, article or apparatus comprising a set of elements includes not only those elements, but also includes elements not expressly listed. other elements of or also include elements inherent in such a process, method, article, or device. Without further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article or apparatus comprising said element.

The technical solution provided by the present application has been introduced in detail above. In this paper, specific examples are used to illustrate the principles and implementation methods of the present application, and the descriptions of the above embodiments are only used to help understand the methods and core ideas of the present application. It should be pointed out that those skilled in the art can make some improvements and modifications to the application without departing from the principles of the application, and these improvements and modifications also fall within the protection scope of the claims of the application.

Claims (10)

1. a kind of adaptive masterplate matching process based on cosine similarity, which is characterized in that including：

According to preset cosine similarity function, template image of the images to be recognized from matching starting point with matching target is calculated It carries out matched similarity value and stores；The cosine similarity function is：

Or

Wherein, the pixel size of the template image is a × b；P (i, j) is ash of the template image at point coordinates (i, j) Angle value；A, b, i and j are positive integer, and 1≤i≤a, 1≤j≤b；(x, y) is that the matching of the images to be recognized originates The coordinate of point；Q (x+i, y+j) is gray value of the images to be recognized at point coordinates (x+i, y+j)；F (x, y) be with it is described Match the corresponding similarity value of starting point (x, y)；

According to the growth trend with the corresponding similarity value of the matching starting point with reversed trend update adjustment translation away from From；

The coordinate of the matching starting point is updated according to the updated translation distance and default translation direction；

Judge whether the coordinate of the updated matching starting point meets preset termination condition；

If it is not, the calculating images to be recognized is then continued to execute from matching starting point and the template image progress for matching target The similarity value matched and the subsequent step stored；

If so, determining the maximum in all similarity values, and determined in images to be recognized according to default basis of characterization The matching target.

2. adaptive masterplate matching process according to claim 1, which is characterized in that the basis matches starting with described The growth trend of the corresponding similarity value of point is included with reversed trend update translation distance：

According to formulaThe translation distance is updated；Wherein, h is The updated translation distance, h ' are the translation distance before update, and h and h ' are positive integer；C1 adjusts ginseng for speedup Number, 1<c1；C2 be deceleration adjustment parameter, 0<c2<1；Predetermined threshold value；[] represents round numbers part.

3. adaptive masterplate matching process according to claim 2, which is characterized in that the predetermined threshold valueFor：

Wherein, β is that the sample image containing the matching target carries out matching obtained similarity value with the template image Mean value；L is threshold coefficient.

4. adaptive masterplate matching process according to claim 2, which is characterized in that the default translation direction for row to Translation.

5. adaptive masterplate matching process according to claim 4, which is characterized in that the translation distance meets condition：

1≤h≤a。

6. adaptive masterplate matching process according to any one of claims 1 to 5, which is characterized in that the basis is preset Basis of characterization determines that the matching target in images to be recognized includes：

Judge the distance between matching starting point corresponding to the two neighboring maximum whether less than pre-determined distance threshold value；

If so, judge the matching starting point corresponding to higher value of the images to be recognized in the two neighboring maximum There are a matching targets at place；

If it is not, then judge that the images to be recognized is deposited respectively in the two neighboring corresponding matching starting point of maximum In a matching target.

7. adaptive template matching process according to claim 6, which is characterized in that the pre-determined distance threshold value includes pre- If lateral distance threshold value and default fore-and-aft distance threshold value；

Whether the distance between described matching starting point judged corresponding to two neighboring maximum is less than pre-determined distance threshold value packet It includes：

Judge whether that the lateral distance between the matching starting point corresponding to the two neighboring maximum is less than the default horizontal stroke It is default less than described to the fore-and-aft distance between the matching starting point corresponding to distance threshold or the two neighboring maximum Fore-and-aft distance threshold value.

8. a kind of adaptive masterplate coalignment based on cosine similarity, which is characterized in that including：

Computing module：For according to preset cosine similarity function, calculating images to be recognized from matching starting point with matching The template image of target carries out matched similarity value and stores；The cosine similarity function is：

Or

Wherein, the pixel size of the template image is a × b；P (i, j) is ash of the template image at point coordinates (i, j) Angle value；A, b, i and j are positive integer, and 1≤i≤a, 1≤j≤b；(x, y) is that the matching of the images to be recognized originates The coordinate of point；Q (x+i, y+j) is gray value of the images to be recognized at point coordinates (x+i, y+j)；F (x, y) be with it is described Match the corresponding similarity value of starting point (x, y)；

Update module：For according to the growth trend with the corresponding similarity value of the matching starting point with reversed trend more New translation distance；And the coordinate of the matching starting point is carried out according to the updated translation distance and default translation direction Update；

Judgment module：Whether the coordinate for judging the updated matching starting point meets preset termination condition；With toilet When stating the coordinate of the matching starting point of computing module in the updated and being unsatisfactory for the preset termination condition, continue on for calculating Images to be recognized is from matching starting point with matching the matched similarity value of template image progress of target and storing；

Identification module：For when the coordinate of the matching starting point meets the preset termination condition, determining all phases Like the maximum in angle value, and determine according to default basis of characterization the matching target in images to be recognized.

9. a kind of adaptive masterplate matching unit based on cosine similarity, which is characterized in that including：

Memory：For storing computer instruction；

Processor：It is as described in any one of claim 1 to 7 similar based on cosine to realize for performing the computer instruction The step of adaptive masterplate matching process of degree.

10. a kind of computer readable storage medium, which is characterized in that be stored with computer in the computer readable storage medium Program, the computer program are realized as described in any one of claim 1 to 7 based on cosine similarity when being executed by processor Adaptive masterplate matching process the step of.